2023-01-05 10:33:54,359 INFO [train.py:934] Training started
2023-01-05 10:33:54,360 INFO [train.py:944] Device: cuda:0
2023-01-05 10:33:54,540 INFO [lexicon.py:168] Loading pre-compiled data/lang_phone/Linv.pt
2023-01-05 10:33:55,973 INFO [train.py:960] { 'activation': 'doubleswish',
  'am_scale': 0.0,
  'average_period': 200,
  'batch_idx_train': 0,
  'beam_size': 10,
  'best_train_epoch': -1,
  'best_train_loss': inf,
  'best_valid_epoch': -1,
  'best_valid_loss': inf,
  'blank_id': 0,
  'bucketing_sampler': True,
  'channels': 300,
  'concatenate_cuts': False,
  'context_size': 2,
  'conv_layers': 18,
  'ctc_loss_scale': 0.7,
  'decoder_dim': 256,
  'drop_last': True,
  'duration_factor': 1.0,
  'enable_musan': True,
  'enable_spec_aug': False,
  'encoder_dim': 256,
  'env_info': { 'IP address': '127.0.1.1',
                'hostname': 'kao-dgxa-f12-u17',
                'icefall-git-branch': 'tiny',
                'icefall-git-date': 'Mon Jan 2 00:08:32 2023',
                'icefall-git-sha1': '2fd970b-dirty',
                'icefall-path': '/home/jsong/git/icefall',
                'k2-build-type': 'Release',
                'k2-git-date': 'Fri Nov 25 08:23:51 2022',
                'k2-git-sha1': '1feafa064cf3b6c243e6b33b0192601224210937',
                'k2-path': '/home/jsong/miniconda3/envs/k2/lib/python3.9/site-packages/k2/__init__.py',
                'k2-version': '1.23.2',
                'k2-with-cuda': True,
                'lhotse-path': '/home/jsong/miniconda3/envs/k2/lib/python3.9/site-packages/lhotse/__init__.py',
                'lhotse-version': '1.7.0',
                'python-version': '3.9',
                'torch-cuda-available': True,
                'torch-cuda-version': '11.3',
                'torch-version': '1.12.0'},
  'exp_dir': PosixPath('tiny_transducer_ctc/exp_2m_phone_halfdelay'),
  'feature_dim': 80,
  'full_libri': True,
  'gap': 1.0,
  'inf_check': False,
  'initial_lr': 0.003,
  'input_strategy': 'PrecomputedFeatures',
  'joiner_dim': 256,
  'keep_last_k': 5,
  'lang_dir': 'data/lang_phone',
  'lm_scale': 0.25,
  'log_interval': 500,
  'manifest_dir': PosixPath('data/fbank'),
  'master_port': 12354,
  'max_duration': 600,
  'num_buckets': 30,
  'num_epochs': 30,
  'num_workers': 2,
  'on_the_fly_feats': False,
  'print_diagnostics': False,
  'prune_range': 5,
  'reset_interval': 200,
  'return_cuts': True,
  'save_every_n': 10000,
  'seed': 42,
  'shuffle': True,
  'simple_loss_scale': 0.5,
  'skip_add': True,
  'spec_aug_time_warp_factor': 80,
  'start_batch': 0,
  'start_epoch': 1,
  'subsampling_factor': 4,
  'tensorboard': True,
  'use_double_scores': True,
  'use_dscnn': True,
  'use_fp16': False,
  'valid_interval': 9000,
  'vocab_size': 72,
  'warm_step': 5000,
  'world_size': 1}
2023-01-05 10:33:55,973 INFO [train.py:964] About to create model
2023-01-05 10:33:58,436 INFO [train.py:425] Encoder MAC ops for 10 seconds of audio is 501.07M
2023-01-05 10:33:58,439 INFO [train.py:974] Number of model parameters: 2186242
2023-01-05 10:33:58,439 INFO [train.py:975] Number of encoder parameters: 1960162
2023-01-05 10:33:58,439 INFO [train.py:976] Number of decoder parameters: 20480
2023-01-05 10:33:58,439 INFO [train.py:977] Number of joiner parameters: 150088
2023-01-05 10:33:58,439 INFO [train.py:978] Number of ctc parameters: 18504
2023-01-05 10:33:58,487 INFO [asr_datamodule.py:419] About to get the shuffled train-clean-100,             train-clean-360 and train-other-500 cuts
2023-01-05 10:33:58,488 INFO [asr_datamodule.py:224] Enable MUSAN
2023-01-05 10:33:58,488 INFO [asr_datamodule.py:225] About to get Musan cuts
2023-01-05 10:33:59,979 INFO [asr_datamodule.py:271] Disable SpecAugment
2023-01-05 10:33:59,980 INFO [asr_datamodule.py:273] About to create train dataset
2023-01-05 10:33:59,980 INFO [asr_datamodule.py:300] Using DynamicBucketingSampler.
2023-01-05 10:34:01,452 INFO [asr_datamodule.py:315] About to create train dataloader
2023-01-05 10:34:01,453 INFO [asr_datamodule.py:429] About to get dev-clean cuts
2023-01-05 10:34:01,453 INFO [asr_datamodule.py:436] About to get dev-other cuts
2023-01-05 10:34:01,454 INFO [asr_datamodule.py:346] About to create dev dataset
2023-01-05 10:34:01,629 INFO [asr_datamodule.py:363] About to create dev dataloader
2023-01-05 10:34:06,342 INFO [train.py:862] Epoch 1, batch 0, loss[loss=4.045, simple_loss=5.07, pruned_loss=5.523, ctc_loss=3.369, over 14501.00 frames. ], tot_loss[loss=4.045, simple_loss=5.07, pruned_loss=5.523, ctc_loss=3.369, over 14501.00 frames. ], batch size: 34, lr: 3.00e-03, 
2023-01-05 10:35:20,900 INFO [train.py:862] Epoch 1, batch 500, loss[loss=0.6665, simple_loss=0.8019, pruned_loss=0.4626, ctc_loss=0.588, over 14678.00 frames. ], tot_loss[loss=0.8172, simple_loss=0.8696, pruned_loss=0.6707, ctc_loss=0.7629, over 2626367.87 frames. ], batch size: 44, lr: 3.00e-03, 
2023-01-05 10:36:29,168 INFO [train.py:862] Epoch 1, batch 1000, loss[loss=0.4997, simple_loss=0.6492, pruned_loss=0.3253, ctc_loss=0.4243, over 14703.00 frames. ], tot_loss[loss=0.5844, simple_loss=0.7075, pruned_loss=0.4043, ctc_loss=0.5143, over 2842778.44 frames. ], batch size: 39, lr: 3.00e-03, 
2023-01-05 10:37:37,092 INFO [train.py:862] Epoch 1, batch 1500, loss[loss=0.528, simple_loss=0.6736, pruned_loss=0.346, ctc_loss=0.4541, over 10306.00 frames. ], tot_loss[loss=0.5114, simple_loss=0.6553, pruned_loss=0.3351, ctc_loss=0.4386, over 2873661.73 frames. ], batch size: 105, lr: 3.00e-03, 
2023-01-05 10:38:44,331 INFO [train.py:862] Epoch 1, batch 2000, loss[loss=0.508, simple_loss=0.6548, pruned_loss=0.3226, ctc_loss=0.4377, over 14851.00 frames. ], tot_loss[loss=0.4674, simple_loss=0.622, pruned_loss=0.2979, ctc_loss=0.3952, over 2861939.72 frames. ], batch size: 42, lr: 3.00e-03, 
2023-01-05 10:39:52,461 INFO [train.py:862] Epoch 1, batch 2500, loss[loss=0.3828, simple_loss=0.5762, pruned_loss=0.2308, ctc_loss=0.3073, over 14838.00 frames. ], tot_loss[loss=0.4366, simple_loss=0.6012, pruned_loss=0.2726, ctc_loss=0.3654, over 2867315.44 frames. ], batch size: 41, lr: 3.00e-03, 
2023-01-05 10:40:59,584 INFO [train.py:862] Epoch 1, batch 3000, loss[loss=0.3813, simple_loss=0.5509, pruned_loss=0.2381, ctc_loss=0.3142, over 14515.00 frames. ], tot_loss[loss=0.4222, simple_loss=0.5927, pruned_loss=0.2621, ctc_loss=0.3524, over 2867822.88 frames. ], batch size: 38, lr: 3.00e-03, 
2023-01-05 10:42:06,483 INFO [train.py:862] Epoch 1, batch 3500, loss[loss=0.3904, simple_loss=0.5987, pruned_loss=0.2315, ctc_loss=0.3185, over 14510.00 frames. ], tot_loss[loss=0.4041, simple_loss=0.58, pruned_loss=0.2495, ctc_loss=0.3365, over 2863347.37 frames. ], batch size: 50, lr: 3.00e-03, 
2023-01-05 10:43:13,876 INFO [train.py:862] Epoch 1, batch 4000, loss[loss=0.4274, simple_loss=0.6067, pruned_loss=0.2636, ctc_loss=0.3619, over 14230.00 frames. ], tot_loss[loss=0.3841, simple_loss=0.5699, pruned_loss=0.2347, ctc_loss=0.3185, over 2877110.95 frames. ], batch size: 52, lr: 3.00e-03, 
2023-01-05 10:44:20,257 INFO [train.py:862] Epoch 1, batch 4500, loss[loss=0.4169, simple_loss=0.6062, pruned_loss=0.2664, ctc_loss=0.3488, over 10014.00 frames. ], tot_loss[loss=0.374, simple_loss=0.5646, pruned_loss=0.2283, ctc_loss=0.3109, over 2863328.42 frames. ], batch size: 104, lr: 3.00e-03, 
2023-01-05 10:45:26,890 INFO [train.py:862] Epoch 1, batch 5000, loss[loss=0.3416, simple_loss=0.5504, pruned_loss=0.2093, ctc_loss=0.2804, over 14863.00 frames. ], tot_loss[loss=0.3627, simple_loss=0.5589, pruned_loss=0.2215, ctc_loss=0.3022, over 2868372.93 frames. ], batch size: 40, lr: 3.00e-03, 
2023-01-05 10:46:33,813 INFO [train.py:862] Epoch 1, batch 5500, loss[loss=0.4029, simple_loss=0.5848, pruned_loss=0.2483, ctc_loss=0.3438, over 13681.00 frames. ], tot_loss[loss=0.3544, simple_loss=0.5549, pruned_loss=0.2148, ctc_loss=0.2953, over 2879118.72 frames. ], batch size: 63, lr: 3.00e-03, 
2023-01-05 10:47:41,071 INFO [train.py:862] Epoch 1, batch 6000, loss[loss=0.327, simple_loss=0.5612, pruned_loss=0.1918, ctc_loss=0.2647, over 14851.00 frames. ], tot_loss[loss=0.348, simple_loss=0.551, pruned_loss=0.2096, ctc_loss=0.2892, over 2877183.25 frames. ], batch size: 42, lr: 3.00e-03, 
2023-01-05 10:48:47,587 INFO [train.py:862] Epoch 1, batch 6500, loss[loss=0.3563, simple_loss=0.5717, pruned_loss=0.2169, ctc_loss=0.2935, over 14675.00 frames. ], tot_loss[loss=0.3424, simple_loss=0.5485, pruned_loss=0.2052, ctc_loss=0.2837, over 2862333.20 frames. ], batch size: 47, lr: 3.00e-03, 
2023-01-05 10:49:53,915 INFO [train.py:862] Epoch 1, batch 7000, loss[loss=0.3733, simple_loss=0.5274, pruned_loss=0.2358, ctc_loss=0.3191, over 14517.00 frames. ], tot_loss[loss=0.3361, simple_loss=0.5424, pruned_loss=0.2005, ctc_loss=0.2779, over 2879169.38 frames. ], batch size: 34, lr: 3.00e-03, 
2023-01-05 10:50:59,927 INFO [train.py:862] Epoch 1, batch 7500, loss[loss=0.3782, simple_loss=0.5786, pruned_loss=0.2351, ctc_loss=0.3156, over 14713.00 frames. ], tot_loss[loss=0.3302, simple_loss=0.5397, pruned_loss=0.1962, ctc_loss=0.2719, over 2864188.48 frames. ], batch size: 47, lr: 3.00e-03, 
2023-01-05 10:52:06,521 INFO [train.py:862] Epoch 1, batch 8000, loss[loss=0.2404, simple_loss=0.4498, pruned_loss=0.1356, ctc_loss=0.1889, over 14521.00 frames. ], tot_loss[loss=0.326, simple_loss=0.5367, pruned_loss=0.1929, ctc_loss=0.268, over 2866657.41 frames. ], batch size: 36, lr: 3.00e-03, 
2023-01-05 10:53:12,389 INFO [train.py:862] Epoch 1, batch 8500, loss[loss=0.431, simple_loss=0.6246, pruned_loss=0.2624, ctc_loss=0.3694, over 14737.00 frames. ], tot_loss[loss=0.3233, simple_loss=0.5364, pruned_loss=0.1905, ctc_loss=0.2652, over 2873587.94 frames. ], batch size: 45, lr: 3.00e-03, 
2023-01-05 10:54:18,442 INFO [train.py:862] Epoch 1, batch 9000, loss[loss=0.3379, simple_loss=0.5498, pruned_loss=0.2013, ctc_loss=0.2786, over 14666.00 frames. ], tot_loss[loss=0.3197, simple_loss=0.5322, pruned_loss=0.1882, ctc_loss=0.262, over 2869561.78 frames. ], batch size: 41, lr: 3.00e-03, 
2023-01-05 10:54:18,442 INFO [train.py:887] Computing validation loss
2023-01-05 10:54:43,748 INFO [train.py:897] Epoch 1, validation: loss=0.3083, simple_loss=0.5372, pruned_loss=0.1797, ctc_loss=0.2483, over 944034.00 frames. 
2023-01-05 10:54:43,748 INFO [train.py:898] Maximum memory allocated so far is 7283MB
2023-01-05 10:55:49,404 INFO [train.py:862] Epoch 1, batch 9500, loss[loss=0.3002, simple_loss=0.5008, pruned_loss=0.1743, ctc_loss=0.2469, over 14688.00 frames. ], tot_loss[loss=0.3149, simple_loss=0.53, pruned_loss=0.1844, ctc_loss=0.2573, over 2866365.69 frames. ], batch size: 35, lr: 3.00e-03, 
2023-01-05 10:56:55,296 INFO [checkpoint.py:75] Saving checkpoint to tiny_transducer_ctc/exp_2m_phone_halfdelay/checkpoint-10000.pt
2023-01-05 10:56:55,546 INFO [train.py:862] Epoch 1, batch 10000, loss[loss=0.2694, simple_loss=0.5217, pruned_loss=0.1507, ctc_loss=0.2085, over 14657.00 frames. ], tot_loss[loss=0.3121, simple_loss=0.5277, pruned_loss=0.1827, ctc_loss=0.2545, over 2860498.86 frames. ], batch size: 41, lr: 3.00e-03, 
2023-01-05 10:58:01,088 INFO [train.py:862] Epoch 1, batch 10500, loss[loss=0.2963, simple_loss=0.5179, pruned_loss=0.1694, ctc_loss=0.2397, over 14688.00 frames. ], tot_loss[loss=0.3069, simple_loss=0.5244, pruned_loss=0.1791, ctc_loss=0.2493, over 2852880.21 frames. ], batch size: 44, lr: 3.00e-03, 
2023-01-05 10:59:07,072 INFO [train.py:862] Epoch 1, batch 11000, loss[loss=0.3515, simple_loss=0.5751, pruned_loss=0.2111, ctc_loss=0.2884, over 14659.00 frames. ], tot_loss[loss=0.3044, simple_loss=0.5231, pruned_loss=0.177, ctc_loss=0.2469, over 2844449.71 frames. ], batch size: 47, lr: 3.00e-03, 
2023-01-05 11:00:12,933 INFO [train.py:862] Epoch 1, batch 11500, loss[loss=0.3017, simple_loss=0.5427, pruned_loss=0.1741, ctc_loss=0.2401, over 14654.00 frames. ], tot_loss[loss=0.3045, simple_loss=0.5227, pruned_loss=0.177, ctc_loss=0.2472, over 2846653.90 frames. ], batch size: 42, lr: 3.00e-03, 
2023-01-05 11:01:19,462 INFO [train.py:862] Epoch 1, batch 12000, loss[loss=0.2755, simple_loss=0.4879, pruned_loss=0.159, ctc_loss=0.2208, over 14694.00 frames. ], tot_loss[loss=0.3018, simple_loss=0.5201, pruned_loss=0.1751, ctc_loss=0.2447, over 2852131.11 frames. ], batch size: 38, lr: 3.00e-03, 
2023-01-05 11:02:25,055 INFO [train.py:862] Epoch 1, batch 12500, loss[loss=0.2249, simple_loss=0.4426, pruned_loss=0.1222, ctc_loss=0.1741, over 14782.00 frames. ], tot_loss[loss=0.2997, simple_loss=0.5179, pruned_loss=0.1734, ctc_loss=0.2429, over 2853513.45 frames. ], batch size: 36, lr: 3.00e-03, 
2023-01-05 11:03:31,049 INFO [train.py:862] Epoch 1, batch 13000, loss[loss=0.3375, simple_loss=0.5709, pruned_loss=0.1974, ctc_loss=0.2751, over 14750.00 frames. ], tot_loss[loss=0.2944, simple_loss=0.5164, pruned_loss=0.1697, ctc_loss=0.2372, over 2875110.29 frames. ], batch size: 45, lr: 3.00e-03, 
2023-01-05 11:04:37,077 INFO [train.py:862] Epoch 1, batch 13500, loss[loss=0.2955, simple_loss=0.5005, pruned_loss=0.1727, ctc_loss=0.2409, over 14728.00 frames. ], tot_loss[loss=0.2957, simple_loss=0.5169, pruned_loss=0.1705, ctc_loss=0.2386, over 2868572.18 frames. ], batch size: 37, lr: 3.00e-03, 
2023-01-05 11:05:42,677 INFO [train.py:862] Epoch 1, batch 14000, loss[loss=0.3039, simple_loss=0.5438, pruned_loss=0.1704, ctc_loss=0.2446, over 14572.00 frames. ], tot_loss[loss=0.2898, simple_loss=0.5121, pruned_loss=0.1661, ctc_loss=0.2331, over 2864455.13 frames. ], batch size: 43, lr: 3.00e-03, 
2023-01-05 11:06:48,717 INFO [train.py:862] Epoch 1, batch 14500, loss[loss=0.306, simple_loss=0.5315, pruned_loss=0.1753, ctc_loss=0.2481, over 14546.00 frames. ], tot_loss[loss=0.2915, simple_loss=0.5134, pruned_loss=0.1671, ctc_loss=0.2348, over 2856540.18 frames. ], batch size: 53, lr: 3.00e-03, 
2023-01-05 11:07:54,214 INFO [train.py:862] Epoch 1, batch 15000, loss[loss=0.2594, simple_loss=0.5002, pruned_loss=0.1435, ctc_loss=0.2019, over 14649.00 frames. ], tot_loss[loss=0.2858, simple_loss=0.5098, pruned_loss=0.1631, ctc_loss=0.2291, over 2849394.92 frames. ], batch size: 44, lr: 3.00e-03, 
2023-01-05 11:09:00,527 INFO [train.py:862] Epoch 1, batch 15500, loss[loss=0.2546, simple_loss=0.478, pruned_loss=0.1437, ctc_loss=0.1996, over 14711.00 frames. ], tot_loss[loss=0.2865, simple_loss=0.5102, pruned_loss=0.1636, ctc_loss=0.2299, over 2846436.73 frames. ], batch size: 38, lr: 3.00e-03, 
2023-01-05 11:10:06,600 INFO [train.py:862] Epoch 1, batch 16000, loss[loss=0.3195, simple_loss=0.5446, pruned_loss=0.1827, ctc_loss=0.2614, over 14541.00 frames. ], tot_loss[loss=0.2826, simple_loss=0.5076, pruned_loss=0.1608, ctc_loss=0.2261, over 2869791.19 frames. ], batch size: 50, lr: 3.00e-03, 
2023-01-05 11:11:12,275 INFO [train.py:862] Epoch 1, batch 16500, loss[loss=0.2696, simple_loss=0.473, pruned_loss=0.1579, ctc_loss=0.2161, over 14020.00 frames. ], tot_loss[loss=0.2807, simple_loss=0.5061, pruned_loss=0.1597, ctc_loss=0.2241, over 2869547.53 frames. ], batch size: 31, lr: 3.00e-03, 
2023-01-05 11:12:18,231 INFO [train.py:862] Epoch 1, batch 17000, loss[loss=0.3634, simple_loss=0.5465, pruned_loss=0.2173, ctc_loss=0.3089, over 10020.00 frames. ], tot_loss[loss=0.2796, simple_loss=0.5054, pruned_loss=0.159, ctc_loss=0.223, over 2872501.14 frames. ], batch size: 104, lr: 3.00e-03, 
2023-01-05 11:13:24,232 INFO [train.py:862] Epoch 1, batch 17500, loss[loss=0.2625, simple_loss=0.4686, pruned_loss=0.1573, ctc_loss=0.2071, over 14674.00 frames. ], tot_loss[loss=0.2789, simple_loss=0.5051, pruned_loss=0.1584, ctc_loss=0.2223, over 2851580.59 frames. ], batch size: 35, lr: 3.00e-03, 
2023-01-05 11:14:28,798 INFO [train.py:862] Epoch 1, batch 18000, loss[loss=0.2995, simple_loss=0.5041, pruned_loss=0.1736, ctc_loss=0.2454, over 14695.00 frames. ], tot_loss[loss=0.276, simple_loss=0.5027, pruned_loss=0.1565, ctc_loss=0.2196, over 2857997.32 frames. ], batch size: 39, lr: 3.00e-03, 
2023-01-05 11:14:28,799 INFO [train.py:887] Computing validation loss
2023-01-05 11:14:54,016 INFO [train.py:897] Epoch 1, validation: loss=0.2761, simple_loss=0.5147, pruned_loss=0.1568, ctc_loss=0.2169, over 944034.00 frames. 
2023-01-05 11:14:54,017 INFO [train.py:898] Maximum memory allocated so far is 7283MB
2023-01-05 11:15:21,330 INFO [checkpoint.py:75] Saving checkpoint to tiny_transducer_ctc/exp_2m_phone_halfdelay/epoch-1.pt
2023-01-05 11:15:24,047 INFO [train.py:862] Epoch 2, batch 0, loss[loss=0.3508, simple_loss=0.5599, pruned_loss=0.2061, ctc_loss=0.2928, over 14249.00 frames. ], tot_loss[loss=0.3508, simple_loss=0.5599, pruned_loss=0.2061, ctc_loss=0.2928, over 14249.00 frames. ], batch size: 52, lr: 3.00e-03, 
2023-01-05 11:16:30,048 INFO [train.py:862] Epoch 2, batch 500, loss[loss=0.2753, simple_loss=0.5166, pruned_loss=0.1558, ctc_loss=0.2158, over 14794.00 frames. ], tot_loss[loss=0.2773, simple_loss=0.5048, pruned_loss=0.1571, ctc_loss=0.2207, over 2638144.53 frames. ], batch size: 43, lr: 3.00e-03, 
2023-01-05 11:17:36,009 INFO [train.py:862] Epoch 2, batch 1000, loss[loss=0.2752, simple_loss=0.5055, pruned_loss=0.1548, ctc_loss=0.2185, over 14833.00 frames. ], tot_loss[loss=0.2718, simple_loss=0.4993, pruned_loss=0.1531, ctc_loss=0.2158, over 2857134.22 frames. ], batch size: 42, lr: 3.00e-03, 
2023-01-05 11:18:42,154 INFO [train.py:862] Epoch 2, batch 1500, loss[loss=0.2626, simple_loss=0.5026, pruned_loss=0.1404, ctc_loss=0.2072, over 14520.00 frames. ], tot_loss[loss=0.2729, simple_loss=0.5015, pruned_loss=0.1535, ctc_loss=0.2167, over 2874945.04 frames. ], batch size: 48, lr: 3.00e-03, 
2023-01-05 11:19:19,637 INFO [checkpoint.py:75] Saving checkpoint to tiny_transducer_ctc/exp_2m_phone_halfdelay/checkpoint-20000.pt
2023-01-05 11:19:48,721 INFO [train.py:862] Epoch 2, batch 2000, loss[loss=0.2536, simple_loss=0.5061, pruned_loss=0.1381, ctc_loss=0.1947, over 14080.00 frames. ], tot_loss[loss=0.2718, simple_loss=0.5001, pruned_loss=0.1534, ctc_loss=0.2154, over 2867077.35 frames. ], batch size: 56, lr: 3.00e-03, 
2023-01-05 11:20:54,284 INFO [train.py:862] Epoch 2, batch 2500, loss[loss=0.2562, simple_loss=0.5251, pruned_loss=0.1367, ctc_loss=0.1949, over 14587.00 frames. ], tot_loss[loss=0.2728, simple_loss=0.5012, pruned_loss=0.1539, ctc_loss=0.2164, over 2861041.26 frames. ], batch size: 43, lr: 3.00e-03, 
2023-01-05 11:22:00,388 INFO [train.py:862] Epoch 2, batch 3000, loss[loss=0.2384, simple_loss=0.4843, pruned_loss=0.1258, ctc_loss=0.1828, over 14652.00 frames. ], tot_loss[loss=0.27, simple_loss=0.4998, pruned_loss=0.1517, ctc_loss=0.2136, over 2865107.20 frames. ], batch size: 42, lr: 3.00e-03, 
2023-01-05 11:23:05,521 INFO [train.py:862] Epoch 2, batch 3500, loss[loss=0.2192, simple_loss=0.4331, pruned_loss=0.1168, ctc_loss=0.1703, over 14388.00 frames. ], tot_loss[loss=0.2672, simple_loss=0.4974, pruned_loss=0.1498, ctc_loss=0.211, over 2885257.99 frames. ], batch size: 32, lr: 3.00e-03, 
2023-01-05 11:24:10,694 INFO [train.py:862] Epoch 2, batch 4000, loss[loss=0.3093, simple_loss=0.5391, pruned_loss=0.1738, ctc_loss=0.2519, over 14521.00 frames. ], tot_loss[loss=0.2646, simple_loss=0.4951, pruned_loss=0.1477, ctc_loss=0.2087, over 2871149.29 frames. ], batch size: 53, lr: 3.00e-03, 
2023-01-05 11:25:16,250 INFO [train.py:862] Epoch 2, batch 4500, loss[loss=0.2283, simple_loss=0.4763, pruned_loss=0.1193, ctc_loss=0.173, over 14583.00 frames. ], tot_loss[loss=0.267, simple_loss=0.4961, pruned_loss=0.1494, ctc_loss=0.2111, over 2860132.98 frames. ], batch size: 43, lr: 3.00e-03, 
2023-01-05 11:26:21,667 INFO [train.py:862] Epoch 2, batch 5000, loss[loss=0.2838, simple_loss=0.5275, pruned_loss=0.167, ctc_loss=0.2209, over 10076.00 frames. ], tot_loss[loss=0.2625, simple_loss=0.4929, pruned_loss=0.1463, ctc_loss=0.2066, over 2862576.95 frames. ], batch size: 104, lr: 3.00e-03, 
2023-01-05 11:27:27,458 INFO [train.py:862] Epoch 2, batch 5500, loss[loss=0.2516, simple_loss=0.502, pruned_loss=0.1371, ctc_loss=0.1931, over 12871.00 frames. ], tot_loss[loss=0.2629, simple_loss=0.4938, pruned_loss=0.1465, ctc_loss=0.2069, over 2877417.09 frames. ], batch size: 75, lr: 3.00e-03, 
2023-01-05 11:28:32,946 INFO [train.py:862] Epoch 2, batch 6000, loss[loss=0.2876, simple_loss=0.4983, pruned_loss=0.1646, ctc_loss=0.2336, over 14707.00 frames. ], tot_loss[loss=0.2635, simple_loss=0.4946, pruned_loss=0.147, ctc_loss=0.2074, over 2865382.62 frames. ], batch size: 38, lr: 3.00e-03, 
2023-01-05 11:29:38,003 INFO [train.py:862] Epoch 2, batch 6500, loss[loss=0.2311, simple_loss=0.4801, pruned_loss=0.1199, ctc_loss=0.1759, over 14740.00 frames. ], tot_loss[loss=0.2625, simple_loss=0.4938, pruned_loss=0.1462, ctc_loss=0.2065, over 2858978.40 frames. ], batch size: 45, lr: 3.00e-03, 
2023-01-05 11:30:43,527 INFO [train.py:862] Epoch 2, batch 7000, loss[loss=0.3195, simple_loss=0.5318, pruned_loss=0.1815, ctc_loss=0.2647, over 14711.00 frames. ], tot_loss[loss=0.2577, simple_loss=0.4893, pruned_loss=0.1431, ctc_loss=0.2019, over 2866478.69 frames. ], batch size: 46, lr: 3.00e-03, 
2023-01-05 11:31:48,325 INFO [train.py:862] Epoch 2, batch 7500, loss[loss=0.2362, simple_loss=0.465, pruned_loss=0.1296, ctc_loss=0.1823, over 14712.00 frames. ], tot_loss[loss=0.257, simple_loss=0.4897, pruned_loss=0.1426, ctc_loss=0.2011, over 2881227.85 frames. ], batch size: 37, lr: 3.00e-03, 
2023-01-05 11:32:53,605 INFO [train.py:862] Epoch 2, batch 8000, loss[loss=0.2311, simple_loss=0.476, pruned_loss=0.1279, ctc_loss=0.1733, over 14529.00 frames. ], tot_loss[loss=0.2584, simple_loss=0.4912, pruned_loss=0.1436, ctc_loss=0.2024, over 2867786.59 frames. ], batch size: 37, lr: 3.00e-03, 
2023-01-05 11:33:58,727 INFO [train.py:862] Epoch 2, batch 8500, loss[loss=0.1995, simple_loss=0.4451, pruned_loss=0.1077, ctc_loss=0.1435, over 14698.00 frames. ], tot_loss[loss=0.2575, simple_loss=0.4901, pruned_loss=0.1426, ctc_loss=0.2017, over 2864715.87 frames. ], batch size: 38, lr: 3.00e-03, 
2023-01-05 11:35:03,328 INFO [train.py:862] Epoch 2, batch 9000, loss[loss=0.2479, simple_loss=0.4392, pruned_loss=0.1458, ctc_loss=0.1975, over 14827.00 frames. ], tot_loss[loss=0.2572, simple_loss=0.4892, pruned_loss=0.1427, ctc_loss=0.2014, over 2863283.01 frames. ], batch size: 34, lr: 3.00e-03, 
2023-01-05 11:35:03,328 INFO [train.py:887] Computing validation loss
2023-01-05 11:35:29,061 INFO [train.py:897] Epoch 2, validation: loss=0.2556, simple_loss=0.5013, pruned_loss=0.1411, ctc_loss=0.1973, over 944034.00 frames. 
2023-01-05 11:35:29,061 INFO [train.py:898] Maximum memory allocated so far is 7283MB
2023-01-05 11:36:34,326 INFO [train.py:862] Epoch 2, batch 9500, loss[loss=0.2028, simple_loss=0.4486, pruned_loss=0.1035, ctc_loss=0.1492, over 14883.00 frames. ], tot_loss[loss=0.2551, simple_loss=0.488, pruned_loss=0.1409, ctc_loss=0.1995, over 2859722.95 frames. ], batch size: 40, lr: 3.00e-03, 
2023-01-05 11:37:39,716 INFO [train.py:862] Epoch 2, batch 10000, loss[loss=0.2351, simple_loss=0.4734, pruned_loss=0.1265, ctc_loss=0.1802, over 14653.00 frames. ], tot_loss[loss=0.2552, simple_loss=0.4893, pruned_loss=0.1411, ctc_loss=0.1993, over 2838644.09 frames. ], batch size: 40, lr: 3.00e-03, 
2023-01-05 11:38:45,544 INFO [train.py:862] Epoch 2, batch 10500, loss[loss=0.2093, simple_loss=0.4494, pruned_loss=0.1106, ctc_loss=0.1553, over 14713.00 frames. ], tot_loss[loss=0.2548, simple_loss=0.4891, pruned_loss=0.1406, ctc_loss=0.199, over 2845960.32 frames. ], batch size: 38, lr: 3.00e-03, 
2023-01-05 11:39:51,632 INFO [train.py:862] Epoch 2, batch 11000, loss[loss=0.2293, simple_loss=0.4961, pruned_loss=0.1214, ctc_loss=0.1693, over 14726.00 frames. ], tot_loss[loss=0.2544, simple_loss=0.4874, pruned_loss=0.1406, ctc_loss=0.1986, over 2860836.01 frames. ], batch size: 45, lr: 3.00e-03, 
2023-01-05 11:40:58,125 INFO [train.py:862] Epoch 2, batch 11500, loss[loss=0.2571, simple_loss=0.4846, pruned_loss=0.1426, ctc_loss=0.2023, over 14657.00 frames. ], tot_loss[loss=0.2536, simple_loss=0.4869, pruned_loss=0.1401, ctc_loss=0.1979, over 2863661.81 frames. ], batch size: 40, lr: 3.00e-03, 
2023-01-05 11:41:35,040 INFO [checkpoint.py:75] Saving checkpoint to tiny_transducer_ctc/exp_2m_phone_halfdelay/checkpoint-30000.pt
2023-01-05 11:42:04,298 INFO [train.py:862] Epoch 2, batch 12000, loss[loss=0.278, simple_loss=0.4963, pruned_loss=0.1532, ctc_loss=0.2252, over 14696.00 frames. ], tot_loss[loss=0.2536, simple_loss=0.4883, pruned_loss=0.1397, ctc_loss=0.1979, over 2859481.54 frames. ], batch size: 38, lr: 3.00e-03, 
2023-01-05 11:43:10,338 INFO [train.py:862] Epoch 2, batch 12500, loss[loss=0.2162, simple_loss=0.4296, pruned_loss=0.117, ctc_loss=0.1666, over 14521.00 frames. ], tot_loss[loss=0.2489, simple_loss=0.4845, pruned_loss=0.1365, ctc_loss=0.1933, over 2864608.11 frames. ], batch size: 36, lr: 3.00e-03, 
2023-01-05 11:44:15,911 INFO [train.py:862] Epoch 2, batch 13000, loss[loss=0.229, simple_loss=0.4855, pruned_loss=0.1204, ctc_loss=0.1714, over 14712.00 frames. ], tot_loss[loss=0.2503, simple_loss=0.4863, pruned_loss=0.1373, ctc_loss=0.1946, over 2866167.69 frames. ], batch size: 46, lr: 3.00e-03, 
2023-01-05 11:45:21,205 INFO [train.py:862] Epoch 2, batch 13500, loss[loss=0.2699, simple_loss=0.4886, pruned_loss=0.1533, ctc_loss=0.2151, over 14863.00 frames. ], tot_loss[loss=0.2522, simple_loss=0.4873, pruned_loss=0.1387, ctc_loss=0.1965, over 2860841.44 frames. ], batch size: 40, lr: 3.00e-03, 
2023-01-05 11:46:26,936 INFO [train.py:862] Epoch 2, batch 14000, loss[loss=0.2395, simple_loss=0.4701, pruned_loss=0.1326, ctc_loss=0.1847, over 14709.00 frames. ], tot_loss[loss=0.2482, simple_loss=0.4846, pruned_loss=0.136, ctc_loss=0.1924, over 2869520.32 frames. ], batch size: 38, lr: 3.00e-03, 
2023-01-05 11:47:32,244 INFO [train.py:862] Epoch 2, batch 14500, loss[loss=0.274, simple_loss=0.51, pruned_loss=0.1474, ctc_loss=0.219, over 14008.00 frames. ], tot_loss[loss=0.2448, simple_loss=0.4812, pruned_loss=0.1335, ctc_loss=0.1894, over 2845761.96 frames. ], batch size: 56, lr: 3.00e-03, 
2023-01-05 11:48:38,484 INFO [train.py:862] Epoch 2, batch 15000, loss[loss=0.2941, simple_loss=0.4801, pruned_loss=0.1783, ctc_loss=0.2409, over 14408.00 frames. ], tot_loss[loss=0.2481, simple_loss=0.4834, pruned_loss=0.1362, ctc_loss=0.1925, over 2863791.70 frames. ], batch size: 33, lr: 3.00e-03, 
2023-01-05 11:49:43,860 INFO [train.py:862] Epoch 2, batch 15500, loss[loss=0.2838, simple_loss=0.5107, pruned_loss=0.1525, ctc_loss=0.2306, over 14693.00 frames. ], tot_loss[loss=0.2475, simple_loss=0.4839, pruned_loss=0.1355, ctc_loss=0.1918, over 2870616.33 frames. ], batch size: 44, lr: 3.00e-03, 
2023-01-05 11:50:49,221 INFO [train.py:862] Epoch 2, batch 16000, loss[loss=0.2071, simple_loss=0.4457, pruned_loss=0.1056, ctc_loss=0.155, over 14725.00 frames. ], tot_loss[loss=0.2423, simple_loss=0.4792, pruned_loss=0.1318, ctc_loss=0.187, over 2870488.33 frames. ], batch size: 37, lr: 3.00e-03, 
2023-01-05 11:51:55,260 INFO [train.py:862] Epoch 2, batch 16500, loss[loss=0.2173, simple_loss=0.4479, pruned_loss=0.1163, ctc_loss=0.1647, over 14699.00 frames. ], tot_loss[loss=0.2439, simple_loss=0.4809, pruned_loss=0.1328, ctc_loss=0.1884, over 2869475.15 frames. ], batch size: 37, lr: 3.00e-03, 
2023-01-05 11:53:00,653 INFO [train.py:862] Epoch 2, batch 17000, loss[loss=0.3074, simple_loss=0.5296, pruned_loss=0.1807, ctc_loss=0.2482, over 14658.00 frames. ], tot_loss[loss=0.2463, simple_loss=0.4825, pruned_loss=0.1349, ctc_loss=0.1907, over 2850856.20 frames. ], batch size: 40, lr: 3.00e-03, 
2023-01-05 11:54:06,297 INFO [train.py:862] Epoch 2, batch 17500, loss[loss=0.2583, simple_loss=0.5203, pruned_loss=0.1412, ctc_loss=0.197, over 14801.00 frames. ], tot_loss[loss=0.2444, simple_loss=0.4804, pruned_loss=0.1334, ctc_loss=0.1891, over 2861905.80 frames. ], batch size: 43, lr: 3.00e-03, 
2023-01-05 11:55:10,306 INFO [train.py:862] Epoch 2, batch 18000, loss[loss=0.2613, simple_loss=0.512, pruned_loss=0.1452, ctc_loss=0.2014, over 14651.00 frames. ], tot_loss[loss=0.2449, simple_loss=0.4812, pruned_loss=0.1333, ctc_loss=0.1896, over 2859820.44 frames. ], batch size: 42, lr: 3.00e-03, 
2023-01-05 11:55:10,307 INFO [train.py:887] Computing validation loss
2023-01-05 11:55:35,326 INFO [train.py:897] Epoch 2, validation: loss=0.2524, simple_loss=0.4975, pruned_loss=0.139, ctc_loss=0.1944, over 944034.00 frames. 
2023-01-05 11:55:35,326 INFO [train.py:898] Maximum memory allocated so far is 7283MB
2023-01-05 11:56:02,811 INFO [checkpoint.py:75] Saving checkpoint to tiny_transducer_ctc/exp_2m_phone_halfdelay/epoch-2.pt
2023-01-05 11:56:05,512 INFO [train.py:862] Epoch 3, batch 0, loss[loss=0.2463, simple_loss=0.4421, pruned_loss=0.1397, ctc_loss=0.1973, over 14527.00 frames. ], tot_loss[loss=0.2463, simple_loss=0.4421, pruned_loss=0.1397, ctc_loss=0.1973, over 14527.00 frames. ], batch size: 34, lr: 3.00e-03, 
2023-01-05 11:57:10,572 INFO [train.py:862] Epoch 3, batch 500, loss[loss=0.2921, simple_loss=0.521, pruned_loss=0.1645, ctc_loss=0.2352, over 14117.00 frames. ], tot_loss[loss=0.245, simple_loss=0.4824, pruned_loss=0.1337, ctc_loss=0.1894, over 2631299.08 frames. ], batch size: 56, lr: 3.00e-03, 
2023-01-05 11:58:15,438 INFO [train.py:862] Epoch 3, batch 1000, loss[loss=0.2635, simple_loss=0.4562, pruned_loss=0.1578, ctc_loss=0.211, over 14055.00 frames. ], tot_loss[loss=0.2447, simple_loss=0.4815, pruned_loss=0.133, ctc_loss=0.1893, over 2855410.97 frames. ], batch size: 31, lr: 3.00e-03, 
2023-01-05 11:59:21,002 INFO [train.py:862] Epoch 3, batch 1500, loss[loss=0.2352, simple_loss=0.4901, pruned_loss=0.1214, ctc_loss=0.179, over 14654.00 frames. ], tot_loss[loss=0.2428, simple_loss=0.4804, pruned_loss=0.1318, ctc_loss=0.1875, over 2872654.73 frames. ], batch size: 41, lr: 3.00e-03, 
2023-01-05 12:00:25,976 INFO [train.py:862] Epoch 3, batch 2000, loss[loss=0.1999, simple_loss=0.4358, pruned_loss=0.1037, ctc_loss=0.1478, over 14518.00 frames. ], tot_loss[loss=0.2431, simple_loss=0.4795, pruned_loss=0.1322, ctc_loss=0.1878, over 2871443.14 frames. ], batch size: 36, lr: 3.00e-03, 
2023-01-05 12:01:30,999 INFO [train.py:862] Epoch 3, batch 2500, loss[loss=0.2579, simple_loss=0.5102, pruned_loss=0.14, ctc_loss=0.1991, over 14009.00 frames. ], tot_loss[loss=0.2407, simple_loss=0.4788, pruned_loss=0.1305, ctc_loss=0.1853, over 2860619.03 frames. ], batch size: 56, lr: 3.00e-03, 
2023-01-05 12:02:36,202 INFO [train.py:862] Epoch 3, batch 3000, loss[loss=0.2563, simple_loss=0.5148, pruned_loss=0.1348, ctc_loss=0.1981, over 14543.00 frames. ], tot_loss[loss=0.2408, simple_loss=0.4791, pruned_loss=0.1302, ctc_loss=0.1855, over 2873246.86 frames. ], batch size: 48, lr: 3.00e-03, 
2023-01-05 12:03:41,635 INFO [train.py:862] Epoch 3, batch 3500, loss[loss=0.2272, simple_loss=0.4878, pruned_loss=0.1155, ctc_loss=0.1706, over 14701.00 frames. ], tot_loss[loss=0.2444, simple_loss=0.4801, pruned_loss=0.1332, ctc_loss=0.1892, over 2869440.06 frames. ], batch size: 44, lr: 3.00e-03, 
2023-01-05 12:03:50,717 INFO [checkpoint.py:75] Saving checkpoint to tiny_transducer_ctc/exp_2m_phone_halfdelay/checkpoint-40000.pt
2023-01-05 12:04:48,628 INFO [train.py:862] Epoch 3, batch 4000, loss[loss=0.1924, simple_loss=0.4239, pruned_loss=0.09842, ctc_loss=0.1418, over 14095.00 frames. ], tot_loss[loss=0.2415, simple_loss=0.4788, pruned_loss=0.1312, ctc_loss=0.1861, over 2876916.26 frames. ], batch size: 31, lr: 3.00e-03, 
2023-01-05 12:05:55,133 INFO [train.py:862] Epoch 3, batch 4500, loss[loss=0.2514, simple_loss=0.5019, pruned_loss=0.1364, ctc_loss=0.1932, over 14668.00 frames. ], tot_loss[loss=0.2413, simple_loss=0.4782, pruned_loss=0.1308, ctc_loss=0.1862, over 2861745.19 frames. ], batch size: 42, lr: 3.00e-03, 
2023-01-05 12:07:01,154 INFO [train.py:862] Epoch 3, batch 5000, loss[loss=0.1881, simple_loss=0.4294, pruned_loss=0.1013, ctc_loss=0.1333, over 13914.00 frames. ], tot_loss[loss=0.2381, simple_loss=0.4774, pruned_loss=0.1285, ctc_loss=0.1828, over 2872687.60 frames. ], batch size: 31, lr: 3.00e-03, 
2023-01-05 12:08:08,005 INFO [train.py:862] Epoch 3, batch 5500, loss[loss=0.263, simple_loss=0.5157, pruned_loss=0.1469, ctc_loss=0.2023, over 14556.00 frames. ], tot_loss[loss=0.2403, simple_loss=0.4779, pruned_loss=0.1303, ctc_loss=0.1851, over 2846224.84 frames. ], batch size: 48, lr: 3.00e-03, 
2023-01-05 12:09:14,439 INFO [train.py:862] Epoch 3, batch 6000, loss[loss=0.2398, simple_loss=0.4926, pruned_loss=0.1298, ctc_loss=0.1815, over 14118.00 frames. ], tot_loss[loss=0.2395, simple_loss=0.4769, pruned_loss=0.1297, ctc_loss=0.1844, over 2875069.78 frames. ], batch size: 56, lr: 3.00e-03, 
2023-01-05 12:10:21,379 INFO [train.py:862] Epoch 3, batch 6500, loss[loss=0.2431, simple_loss=0.4958, pruned_loss=0.1311, ctc_loss=0.1848, over 13804.00 frames. ], tot_loss[loss=0.2366, simple_loss=0.475, pruned_loss=0.1279, ctc_loss=0.1815, over 2877690.48 frames. ], batch size: 63, lr: 3.00e-03, 
2023-01-05 12:11:27,672 INFO [train.py:862] Epoch 3, batch 7000, loss[loss=0.2206, simple_loss=0.4889, pruned_loss=0.1113, ctc_loss=0.1626, over 14683.00 frames. ], tot_loss[loss=0.2365, simple_loss=0.475, pruned_loss=0.1272, ctc_loss=0.1815, over 2870662.88 frames. ], batch size: 47, lr: 3.00e-03, 
2023-01-05 12:12:34,068 INFO [train.py:862] Epoch 3, batch 7500, loss[loss=0.1908, simple_loss=0.4544, pruned_loss=0.09341, ctc_loss=0.1352, over 14595.00 frames. ], tot_loss[loss=0.2341, simple_loss=0.4739, pruned_loss=0.1258, ctc_loss=0.179, over 2857243.18 frames. ], batch size: 43, lr: 3.00e-03, 
2023-01-05 12:13:41,012 INFO [train.py:862] Epoch 3, batch 8000, loss[loss=0.1841, simple_loss=0.4506, pruned_loss=0.09189, ctc_loss=0.127, over 14823.00 frames. ], tot_loss[loss=0.2359, simple_loss=0.4746, pruned_loss=0.1271, ctc_loss=0.1808, over 2871165.78 frames. ], batch size: 41, lr: 3.00e-03, 
2023-01-05 12:14:47,453 INFO [train.py:862] Epoch 3, batch 8500, loss[loss=0.2278, simple_loss=0.4804, pruned_loss=0.1234, ctc_loss=0.1696, over 14844.00 frames. ], tot_loss[loss=0.2361, simple_loss=0.4753, pruned_loss=0.1274, ctc_loss=0.1809, over 2857486.43 frames. ], batch size: 42, lr: 3.00e-03, 
2023-01-05 12:15:54,253 INFO [train.py:862] Epoch 3, batch 9000, loss[loss=0.28, simple_loss=0.5351, pruned_loss=0.1514, ctc_loss=0.2205, over 14728.00 frames. ], tot_loss[loss=0.2367, simple_loss=0.4747, pruned_loss=0.1275, ctc_loss=0.1817, over 2863389.25 frames. ], batch size: 46, lr: 3.00e-03, 
2023-01-05 12:15:54,253 INFO [train.py:887] Computing validation loss
2023-01-05 12:16:19,733 INFO [train.py:897] Epoch 3, validation: loss=0.2392, simple_loss=0.4883, pruned_loss=0.1291, ctc_loss=0.1818, over 944034.00 frames. 
2023-01-05 12:16:19,733 INFO [train.py:898] Maximum memory allocated so far is 7283MB
2023-01-05 12:17:26,584 INFO [train.py:862] Epoch 3, batch 9500, loss[loss=0.2538, simple_loss=0.4746, pruned_loss=0.14, ctc_loss=0.2009, over 14510.00 frames. ], tot_loss[loss=0.2337, simple_loss=0.4727, pruned_loss=0.1254, ctc_loss=0.1788, over 2863722.65 frames. ], batch size: 37, lr: 3.00e-03, 
2023-01-05 12:18:33,902 INFO [train.py:862] Epoch 3, batch 10000, loss[loss=0.2173, simple_loss=0.442, pruned_loss=0.1184, ctc_loss=0.165, over 14524.00 frames. ], tot_loss[loss=0.2351, simple_loss=0.4739, pruned_loss=0.1263, ctc_loss=0.1802, over 2865344.83 frames. ], batch size: 38, lr: 3.00e-03, 
2023-01-05 12:19:40,377 INFO [train.py:862] Epoch 3, batch 10500, loss[loss=0.2296, simple_loss=0.4804, pruned_loss=0.1215, ctc_loss=0.173, over 14529.00 frames. ], tot_loss[loss=0.2335, simple_loss=0.4731, pruned_loss=0.1252, ctc_loss=0.1786, over 2874768.71 frames. ], batch size: 53, lr: 3.00e-03, 
2023-01-05 12:20:46,767 INFO [train.py:862] Epoch 3, batch 11000, loss[loss=0.2168, simple_loss=0.4733, pruned_loss=0.1135, ctc_loss=0.1596, over 13742.00 frames. ], tot_loss[loss=0.2337, simple_loss=0.475, pruned_loss=0.125, ctc_loss=0.1784, over 2862346.95 frames. ], batch size: 63, lr: 3.00e-03, 
2023-01-05 12:21:53,184 INFO [train.py:862] Epoch 3, batch 11500, loss[loss=0.2553, simple_loss=0.4963, pruned_loss=0.1424, ctc_loss=0.1973, over 13013.00 frames. ], tot_loss[loss=0.2334, simple_loss=0.4734, pruned_loss=0.1252, ctc_loss=0.1783, over 2873110.33 frames. ], batch size: 76, lr: 3.00e-03, 
2023-01-05 12:22:59,334 INFO [train.py:862] Epoch 3, batch 12000, loss[loss=0.2216, simple_loss=0.4827, pruned_loss=0.1169, ctc_loss=0.163, over 13645.00 frames. ], tot_loss[loss=0.2308, simple_loss=0.473, pruned_loss=0.1233, ctc_loss=0.1755, over 2859942.84 frames. ], batch size: 63, lr: 3.00e-03, 
2023-01-05 12:24:06,294 INFO [train.py:862] Epoch 3, batch 12500, loss[loss=0.2404, simple_loss=0.4788, pruned_loss=0.1366, ctc_loss=0.1822, over 12900.00 frames. ], tot_loss[loss=0.2314, simple_loss=0.4723, pruned_loss=0.1235, ctc_loss=0.1765, over 2848734.31 frames. ], batch size: 76, lr: 3.00e-03, 
2023-01-05 12:25:12,744 INFO [train.py:862] Epoch 3, batch 13000, loss[loss=0.2425, simple_loss=0.4427, pruned_loss=0.1352, ctc_loss=0.1936, over 14414.00 frames. ], tot_loss[loss=0.232, simple_loss=0.4736, pruned_loss=0.1238, ctc_loss=0.1769, over 2869118.86 frames. ], batch size: 33, lr: 3.00e-03, 
2023-01-05 12:26:19,776 INFO [train.py:862] Epoch 3, batch 13500, loss[loss=0.2328, simple_loss=0.4917, pruned_loss=0.1247, ctc_loss=0.1738, over 14477.00 frames. ], tot_loss[loss=0.2333, simple_loss=0.4735, pruned_loss=0.1249, ctc_loss=0.1783, over 2854099.12 frames. ], batch size: 50, lr: 3.00e-03, 
2023-01-05 12:26:28,188 INFO [checkpoint.py:75] Saving checkpoint to tiny_transducer_ctc/exp_2m_phone_halfdelay/checkpoint-50000.pt
2023-01-05 12:27:25,906 INFO [train.py:862] Epoch 3, batch 14000, loss[loss=0.2258, simple_loss=0.4661, pruned_loss=0.1154, ctc_loss=0.1732, over 14028.00 frames. ], tot_loss[loss=0.2319, simple_loss=0.4736, pruned_loss=0.124, ctc_loss=0.1767, over 2846003.78 frames. ], batch size: 56, lr: 3.00e-03, 
2023-01-05 12:28:32,815 INFO [train.py:862] Epoch 3, batch 14500, loss[loss=0.1882, simple_loss=0.4474, pruned_loss=0.09648, ctc_loss=0.1316, over 14684.00 frames. ], tot_loss[loss=0.2315, simple_loss=0.4718, pruned_loss=0.1236, ctc_loss=0.1766, over 2856259.56 frames. ], batch size: 44, lr: 3.00e-03, 
2023-01-05 12:29:39,217 INFO [train.py:862] Epoch 3, batch 15000, loss[loss=0.1815, simple_loss=0.4375, pruned_loss=0.09102, ctc_loss=0.1265, over 14705.00 frames. ], tot_loss[loss=0.2275, simple_loss=0.4686, pruned_loss=0.1207, ctc_loss=0.1729, over 2857684.50 frames. ], batch size: 38, lr: 3.00e-03, 
2023-01-05 12:30:45,545 INFO [train.py:862] Epoch 3, batch 15500, loss[loss=0.2192, simple_loss=0.4702, pruned_loss=0.1131, ctc_loss=0.164, over 14863.00 frames. ], tot_loss[loss=0.23, simple_loss=0.4716, pruned_loss=0.1227, ctc_loss=0.1749, over 2864780.58 frames. ], batch size: 40, lr: 3.00e-03, 
2023-01-05 12:31:52,192 INFO [train.py:862] Epoch 3, batch 16000, loss[loss=0.2137, simple_loss=0.4536, pruned_loss=0.1126, ctc_loss=0.1598, over 14718.00 frames. ], tot_loss[loss=0.2278, simple_loss=0.4682, pruned_loss=0.1212, ctc_loss=0.1732, over 2881613.24 frames. ], batch size: 37, lr: 3.00e-03, 
2023-01-05 12:32:58,484 INFO [train.py:862] Epoch 3, batch 16500, loss[loss=0.2634, simple_loss=0.5013, pruned_loss=0.1369, ctc_loss=0.2102, over 14736.00 frames. ], tot_loss[loss=0.2281, simple_loss=0.4699, pruned_loss=0.1214, ctc_loss=0.1731, over 2875318.92 frames. ], batch size: 46, lr: 3.00e-03, 
2023-01-05 12:34:04,796 INFO [train.py:862] Epoch 3, batch 17000, loss[loss=0.2488, simple_loss=0.4383, pruned_loss=0.1408, ctc_loss=0.2012, over 14523.00 frames. ], tot_loss[loss=0.2294, simple_loss=0.4702, pruned_loss=0.1225, ctc_loss=0.1744, over 2862321.37 frames. ], batch size: 34, lr: 3.00e-03, 
2023-01-05 12:35:11,524 INFO [train.py:862] Epoch 3, batch 17500, loss[loss=0.2105, simple_loss=0.4753, pruned_loss=0.1071, ctc_loss=0.153, over 14834.00 frames. ], tot_loss[loss=0.2287, simple_loss=0.4706, pruned_loss=0.1217, ctc_loss=0.1736, over 2859145.68 frames. ], batch size: 42, lr: 3.00e-03, 
2023-01-05 12:36:16,401 INFO [train.py:862] Epoch 3, batch 18000, loss[loss=0.1911, simple_loss=0.4327, pruned_loss=0.1027, ctc_loss=0.1363, over 14405.00 frames. ], tot_loss[loss=0.2289, simple_loss=0.4708, pruned_loss=0.1219, ctc_loss=0.1738, over 2855542.66 frames. ], batch size: 33, lr: 3.00e-03, 
2023-01-05 12:36:16,402 INFO [train.py:887] Computing validation loss
2023-01-05 12:36:42,148 INFO [train.py:897] Epoch 3, validation: loss=0.235, simple_loss=0.4854, pruned_loss=0.1269, ctc_loss=0.1773, over 944034.00 frames. 
2023-01-05 12:36:42,149 INFO [train.py:898] Maximum memory allocated so far is 7283MB
2023-01-05 12:37:09,060 INFO [checkpoint.py:75] Saving checkpoint to tiny_transducer_ctc/exp_2m_phone_halfdelay/epoch-3.pt
2023-01-05 12:37:12,494 INFO [train.py:862] Epoch 4, batch 0, loss[loss=0.2619, simple_loss=0.5099, pruned_loss=0.1459, ctc_loss=0.2023, over 14673.00 frames. ], tot_loss[loss=0.2619, simple_loss=0.5099, pruned_loss=0.1459, ctc_loss=0.2023, over 14673.00 frames. ], batch size: 47, lr: 3.00e-03, 
2023-01-05 12:38:18,544 INFO [train.py:862] Epoch 4, batch 500, loss[loss=0.2174, simple_loss=0.4652, pruned_loss=0.1129, ctc_loss=0.1624, over 14661.00 frames. ], tot_loss[loss=0.2293, simple_loss=0.471, pruned_loss=0.1221, ctc_loss=0.1744, over 2633296.85 frames. ], batch size: 40, lr: 3.00e-03, 
2023-01-05 12:39:25,331 INFO [train.py:862] Epoch 4, batch 1000, loss[loss=0.2484, simple_loss=0.4927, pruned_loss=0.1341, ctc_loss=0.1918, over 14560.00 frames. ], tot_loss[loss=0.2252, simple_loss=0.4673, pruned_loss=0.1189, ctc_loss=0.1707, over 2858774.56 frames. ], batch size: 48, lr: 3.00e-03, 
2023-01-05 12:40:31,782 INFO [train.py:862] Epoch 4, batch 1500, loss[loss=0.2266, simple_loss=0.5024, pruned_loss=0.1167, ctc_loss=0.166, over 14757.00 frames. ], tot_loss[loss=0.2289, simple_loss=0.4711, pruned_loss=0.1219, ctc_loss=0.1738, over 2855450.30 frames. ], batch size: 45, lr: 3.00e-03, 
2023-01-05 12:41:37,337 INFO [train.py:862] Epoch 4, batch 2000, loss[loss=0.1784, simple_loss=0.4015, pruned_loss=0.08826, ctc_loss=0.131, over 14671.00 frames. ], tot_loss[loss=0.2292, simple_loss=0.4704, pruned_loss=0.122, ctc_loss=0.1743, over 2871912.65 frames. ], batch size: 35, lr: 3.00e-03, 
2023-01-05 12:42:42,915 INFO [train.py:862] Epoch 4, batch 2500, loss[loss=0.2489, simple_loss=0.4976, pruned_loss=0.1297, ctc_loss=0.1933, over 14690.00 frames. ], tot_loss[loss=0.2271, simple_loss=0.4703, pruned_loss=0.1205, ctc_loss=0.1721, over 2865853.84 frames. ], batch size: 44, lr: 3.00e-03, 
2023-01-05 12:43:47,662 INFO [train.py:862] Epoch 4, batch 3000, loss[loss=0.1942, simple_loss=0.4056, pruned_loss=0.1031, ctc_loss=0.1463, over 14526.00 frames. ], tot_loss[loss=0.2244, simple_loss=0.4673, pruned_loss=0.1186, ctc_loss=0.1696, over 2862415.12 frames. ], batch size: 34, lr: 3.00e-03, 
2023-01-05 12:44:53,460 INFO [train.py:862] Epoch 4, batch 3500, loss[loss=0.2341, simple_loss=0.4279, pruned_loss=0.1326, ctc_loss=0.1859, over 14532.00 frames. ], tot_loss[loss=0.2279, simple_loss=0.468, pruned_loss=0.1212, ctc_loss=0.1734, over 2877966.91 frames. ], batch size: 34, lr: 3.00e-03, 
2023-01-05 12:45:58,771 INFO [train.py:862] Epoch 4, batch 4000, loss[loss=0.2155, simple_loss=0.4429, pruned_loss=0.1172, ctc_loss=0.1627, over 14532.00 frames. ], tot_loss[loss=0.2274, simple_loss=0.4686, pruned_loss=0.1207, ctc_loss=0.1727, over 2873414.37 frames. ], batch size: 36, lr: 3.00e-03, 
2023-01-05 12:47:03,978 INFO [train.py:862] Epoch 4, batch 4500, loss[loss=0.2218, simple_loss=0.4467, pruned_loss=0.1164, ctc_loss=0.1713, over 14706.00 frames. ], tot_loss[loss=0.2277, simple_loss=0.4685, pruned_loss=0.121, ctc_loss=0.173, over 2872774.67 frames. ], batch size: 37, lr: 3.00e-03, 
2023-01-05 12:48:09,996 INFO [train.py:862] Epoch 4, batch 5000, loss[loss=0.2591, simple_loss=0.4963, pruned_loss=0.1395, ctc_loss=0.2039, over 14477.00 frames. ], tot_loss[loss=0.2259, simple_loss=0.4683, pruned_loss=0.1197, ctc_loss=0.171, over 2863048.19 frames. ], batch size: 50, lr: 3.00e-03, 
2023-01-05 12:48:55,399 INFO [checkpoint.py:75] Saving checkpoint to tiny_transducer_ctc/exp_2m_phone_halfdelay/checkpoint-60000.pt
2023-01-05 12:49:15,685 INFO [train.py:862] Epoch 4, batch 5500, loss[loss=0.2054, simple_loss=0.4576, pruned_loss=0.104, ctc_loss=0.1508, over 14849.00 frames. ], tot_loss[loss=0.2246, simple_loss=0.468, pruned_loss=0.1184, ctc_loss=0.1699, over 2852681.04 frames. ], batch size: 42, lr: 3.00e-03, 
2023-01-05 12:50:21,349 INFO [train.py:862] Epoch 4, batch 6000, loss[loss=0.2505, simple_loss=0.4918, pruned_loss=0.1339, ctc_loss=0.1951, over 14571.00 frames. ], tot_loss[loss=0.2265, simple_loss=0.469, pruned_loss=0.1202, ctc_loss=0.1715, over 2849674.24 frames. ], batch size: 43, lr: 3.00e-03, 
2023-01-05 12:51:27,016 INFO [train.py:862] Epoch 4, batch 6500, loss[loss=0.2053, simple_loss=0.4349, pruned_loss=0.1088, ctc_loss=0.1534, over 14522.00 frames. ], tot_loss[loss=0.224, simple_loss=0.4664, pruned_loss=0.1186, ctc_loss=0.1692, over 2870343.24 frames. ], batch size: 36, lr: 3.00e-03, 
2023-01-05 12:52:32,322 INFO [train.py:862] Epoch 4, batch 7000, loss[loss=0.2123, simple_loss=0.4526, pruned_loss=0.1117, ctc_loss=0.1584, over 14517.00 frames. ], tot_loss[loss=0.2254, simple_loss=0.4678, pruned_loss=0.1195, ctc_loss=0.1706, over 2861829.44 frames. ], batch size: 38, lr: 3.00e-03, 
2023-01-05 12:53:38,248 INFO [train.py:862] Epoch 4, batch 7500, loss[loss=0.2442, simple_loss=0.4872, pruned_loss=0.1264, ctc_loss=0.1903, over 14794.00 frames. ], tot_loss[loss=0.225, simple_loss=0.4684, pruned_loss=0.1188, ctc_loss=0.1702, over 2850076.85 frames. ], batch size: 43, lr: 3.00e-03, 
2023-01-05 12:54:43,262 INFO [train.py:862] Epoch 4, batch 8000, loss[loss=0.1953, simple_loss=0.44, pruned_loss=0.09685, ctc_loss=0.1432, over 14701.00 frames. ], tot_loss[loss=0.222, simple_loss=0.4658, pruned_loss=0.1169, ctc_loss=0.1672, over 2856250.98 frames. ], batch size: 38, lr: 3.00e-03, 
2023-01-05 12:55:48,599 INFO [train.py:862] Epoch 4, batch 8500, loss[loss=0.1892, simple_loss=0.4213, pruned_loss=0.09575, ctc_loss=0.1389, over 14424.00 frames. ], tot_loss[loss=0.2232, simple_loss=0.4665, pruned_loss=0.118, ctc_loss=0.1683, over 2873001.25 frames. ], batch size: 33, lr: 3.00e-03, 
2023-01-05 12:56:54,573 INFO [train.py:862] Epoch 4, batch 9000, loss[loss=0.2418, simple_loss=0.488, pruned_loss=0.126, ctc_loss=0.1868, over 14730.00 frames. ], tot_loss[loss=0.2217, simple_loss=0.4649, pruned_loss=0.1165, ctc_loss=0.1672, over 2843458.61 frames. ], batch size: 46, lr: 3.00e-03, 
2023-01-05 12:56:54,573 INFO [train.py:887] Computing validation loss
2023-01-05 12:57:19,485 INFO [train.py:897] Epoch 4, validation: loss=0.2294, simple_loss=0.4812, pruned_loss=0.1207, ctc_loss=0.1729, over 944034.00 frames. 
2023-01-05 12:57:19,485 INFO [train.py:898] Maximum memory allocated so far is 7283MB
2023-01-05 12:58:25,768 INFO [train.py:862] Epoch 4, batch 9500, loss[loss=0.2636, simple_loss=0.4968, pruned_loss=0.1497, ctc_loss=0.206, over 10212.00 frames. ], tot_loss[loss=0.2235, simple_loss=0.4659, pruned_loss=0.1183, ctc_loss=0.1688, over 2856736.69 frames. ], batch size: 104, lr: 3.00e-03, 
2023-01-05 12:59:31,373 INFO [train.py:862] Epoch 4, batch 10000, loss[loss=0.2308, simple_loss=0.4888, pruned_loss=0.1118, ctc_loss=0.177, over 14676.00 frames. ], tot_loss[loss=0.2223, simple_loss=0.4649, pruned_loss=0.1173, ctc_loss=0.1676, over 2875076.09 frames. ], batch size: 44, lr: 3.00e-03, 
2023-01-05 13:00:37,314 INFO [train.py:862] Epoch 4, batch 10500, loss[loss=0.2991, simple_loss=0.5243, pruned_loss=0.1704, ctc_loss=0.2419, over 9738.00 frames. ], tot_loss[loss=0.2236, simple_loss=0.4661, pruned_loss=0.1181, ctc_loss=0.1689, over 2861063.92 frames. ], batch size: 105, lr: 3.00e-03, 
2023-01-05 13:01:42,597 INFO [train.py:862] Epoch 4, batch 11000, loss[loss=0.1925, simple_loss=0.4173, pruned_loss=0.09999, ctc_loss=0.1427, over 14517.00 frames. ], tot_loss[loss=0.2215, simple_loss=0.4648, pruned_loss=0.1166, ctc_loss=0.1668, over 2874392.33 frames. ], batch size: 37, lr: 3.00e-03, 
2023-01-05 13:02:48,169 INFO [train.py:862] Epoch 4, batch 11500, loss[loss=0.239, simple_loss=0.4703, pruned_loss=0.1296, ctc_loss=0.1851, over 14673.00 frames. ], tot_loss[loss=0.2227, simple_loss=0.4664, pruned_loss=0.1171, ctc_loss=0.1679, over 2861088.72 frames. ], batch size: 39, lr: 3.00e-03, 
2023-01-05 13:03:53,991 INFO [train.py:862] Epoch 4, batch 12000, loss[loss=0.2467, simple_loss=0.439, pruned_loss=0.1397, ctc_loss=0.1985, over 14763.00 frames. ], tot_loss[loss=0.221, simple_loss=0.4642, pruned_loss=0.1163, ctc_loss=0.1665, over 2869279.19 frames. ], batch size: 36, lr: 3.00e-03, 
2023-01-05 13:04:59,455 INFO [train.py:862] Epoch 4, batch 12500, loss[loss=0.2025, simple_loss=0.4484, pruned_loss=0.1051, ctc_loss=0.1481, over 14524.00 frames. ], tot_loss[loss=0.2205, simple_loss=0.4634, pruned_loss=0.1158, ctc_loss=0.1661, over 2872383.15 frames. ], batch size: 37, lr: 3.00e-03, 
2023-01-05 13:06:04,379 INFO [train.py:862] Epoch 4, batch 13000, loss[loss=0.2406, simple_loss=0.4875, pruned_loss=0.1249, ctc_loss=0.1856, over 14546.00 frames. ], tot_loss[loss=0.2208, simple_loss=0.4656, pruned_loss=0.116, ctc_loss=0.1659, over 2867784.01 frames. ], batch size: 48, lr: 3.00e-03, 
2023-01-05 13:07:10,466 INFO [train.py:862] Epoch 4, batch 13500, loss[loss=0.1928, simple_loss=0.4237, pruned_loss=0.1001, ctc_loss=0.1417, over 14675.00 frames. ], tot_loss[loss=0.2238, simple_loss=0.4671, pruned_loss=0.118, ctc_loss=0.169, over 2849421.84 frames. ], batch size: 35, lr: 3.00e-03, 
2023-01-05 13:08:15,740 INFO [train.py:862] Epoch 4, batch 14000, loss[loss=0.1973, simple_loss=0.472, pruned_loss=0.09443, ctc_loss=0.1403, over 14729.00 frames. ], tot_loss[loss=0.2221, simple_loss=0.466, pruned_loss=0.1165, ctc_loss=0.1675, over 2831625.13 frames. ], batch size: 46, lr: 3.00e-03, 
2023-01-05 13:09:21,819 INFO [train.py:862] Epoch 4, batch 14500, loss[loss=0.2227, simple_loss=0.447, pruned_loss=0.1173, ctc_loss=0.1721, over 14528.00 frames. ], tot_loss[loss=0.2215, simple_loss=0.4648, pruned_loss=0.1164, ctc_loss=0.1669, over 2850314.57 frames. ], batch size: 37, lr: 3.00e-03, 
2023-01-05 13:10:27,108 INFO [train.py:862] Epoch 4, batch 15000, loss[loss=0.2471, simple_loss=0.4926, pruned_loss=0.1361, ctc_loss=0.1891, over 14666.00 frames. ], tot_loss[loss=0.2245, simple_loss=0.4677, pruned_loss=0.1184, ctc_loss=0.1697, over 2844092.48 frames. ], batch size: 42, lr: 3.00e-03, 
2023-01-05 13:11:12,936 INFO [checkpoint.py:75] Saving checkpoint to tiny_transducer_ctc/exp_2m_phone_halfdelay/checkpoint-70000.pt
2023-01-05 13:11:33,365 INFO [train.py:862] Epoch 4, batch 15500, loss[loss=0.1939, simple_loss=0.4153, pruned_loss=0.1009, ctc_loss=0.1447, over 14020.00 frames. ], tot_loss[loss=0.2217, simple_loss=0.4651, pruned_loss=0.1166, ctc_loss=0.1672, over 2851220.68 frames. ], batch size: 31, lr: 3.00e-03, 
2023-01-05 13:12:39,156 INFO [train.py:862] Epoch 4, batch 16000, loss[loss=0.266, simple_loss=0.5332, pruned_loss=0.1428, ctc_loss=0.2046, over 14518.00 frames. ], tot_loss[loss=0.2197, simple_loss=0.4635, pruned_loss=0.1153, ctc_loss=0.1651, over 2875744.71 frames. ], batch size: 53, lr: 3.00e-03, 
2023-01-05 13:13:45,121 INFO [train.py:862] Epoch 4, batch 16500, loss[loss=0.2245, simple_loss=0.4773, pruned_loss=0.1217, ctc_loss=0.1662, over 14824.00 frames. ], tot_loss[loss=0.2188, simple_loss=0.4633, pruned_loss=0.1148, ctc_loss=0.1641, over 2859413.42 frames. ], batch size: 41, lr: 3.00e-03, 
2023-01-05 13:14:51,024 INFO [train.py:862] Epoch 4, batch 17000, loss[loss=0.2027, simple_loss=0.4563, pruned_loss=0.1047, ctc_loss=0.1469, over 14708.00 frames. ], tot_loss[loss=0.2191, simple_loss=0.4632, pruned_loss=0.1148, ctc_loss=0.1645, over 2872657.43 frames. ], batch size: 39, lr: 3.00e-03, 
2023-01-05 13:15:56,983 INFO [train.py:862] Epoch 4, batch 17500, loss[loss=0.1903, simple_loss=0.4352, pruned_loss=0.09281, ctc_loss=0.1388, over 14883.00 frames. ], tot_loss[loss=0.2193, simple_loss=0.4635, pruned_loss=0.115, ctc_loss=0.1646, over 2866474.28 frames. ], batch size: 40, lr: 3.00e-03, 
2023-01-05 13:17:01,822 INFO [train.py:862] Epoch 4, batch 18000, loss[loss=0.1979, simple_loss=0.466, pruned_loss=0.09491, ctc_loss=0.1422, over 14649.00 frames. ], tot_loss[loss=0.2225, simple_loss=0.4655, pruned_loss=0.1171, ctc_loss=0.1679, over 2856952.63 frames. ], batch size: 41, lr: 3.00e-03, 
2023-01-05 13:17:01,823 INFO [train.py:887] Computing validation loss
2023-01-05 13:17:27,217 INFO [train.py:897] Epoch 4, validation: loss=0.2299, simple_loss=0.4813, pruned_loss=0.1223, ctc_loss=0.1728, over 944034.00 frames. 
2023-01-05 13:17:27,218 INFO [train.py:898] Maximum memory allocated so far is 7283MB
2023-01-05 13:17:54,180 INFO [checkpoint.py:75] Saving checkpoint to tiny_transducer_ctc/exp_2m_phone_halfdelay/epoch-4.pt
2023-01-05 13:17:57,263 INFO [train.py:862] Epoch 5, batch 0, loss[loss=0.2967, simple_loss=0.5027, pruned_loss=0.1648, ctc_loss=0.2456, over 14631.00 frames. ], tot_loss[loss=0.2967, simple_loss=0.5027, pruned_loss=0.1648, ctc_loss=0.2456, over 14631.00 frames. ], batch size: 47, lr: 3.00e-03, 
2023-01-05 13:19:02,596 INFO [train.py:862] Epoch 5, batch 500, loss[loss=0.2323, simple_loss=0.4967, pruned_loss=0.1202, ctc_loss=0.1739, over 14544.00 frames. ], tot_loss[loss=0.2194, simple_loss=0.4638, pruned_loss=0.1148, ctc_loss=0.1649, over 2645600.18 frames. ], batch size: 48, lr: 3.00e-03, 
2023-01-05 13:20:07,185 INFO [train.py:862] Epoch 5, batch 1000, loss[loss=0.1958, simple_loss=0.4749, pruned_loss=0.09675, ctc_loss=0.1365, over 14869.00 frames. ], tot_loss[loss=0.2184, simple_loss=0.4635, pruned_loss=0.1141, ctc_loss=0.1638, over 2850869.22 frames. ], batch size: 40, lr: 3.00e-03, 
2023-01-05 13:21:12,483 INFO [train.py:862] Epoch 5, batch 1500, loss[loss=0.2282, simple_loss=0.4455, pruned_loss=0.1215, ctc_loss=0.1785, over 14513.00 frames. ], tot_loss[loss=0.2189, simple_loss=0.4639, pruned_loss=0.1146, ctc_loss=0.1642, over 2869172.86 frames. ], batch size: 37, lr: 3.00e-03, 
2023-01-05 13:22:17,648 INFO [train.py:862] Epoch 5, batch 2000, loss[loss=0.1546, simple_loss=0.3942, pruned_loss=0.07394, ctc_loss=0.1046, over 14403.00 frames. ], tot_loss[loss=0.2203, simple_loss=0.4634, pruned_loss=0.1154, ctc_loss=0.1659, over 2887528.48 frames. ], batch size: 33, lr: 3.00e-03, 
2023-01-05 13:23:22,692 INFO [train.py:862] Epoch 5, batch 2500, loss[loss=0.1891, simple_loss=0.4325, pruned_loss=0.09752, ctc_loss=0.1356, over 14865.00 frames. ], tot_loss[loss=0.2176, simple_loss=0.4614, pruned_loss=0.1137, ctc_loss=0.1633, over 2879598.62 frames. ], batch size: 40, lr: 3.00e-03, 
2023-01-05 13:24:27,900 INFO [train.py:862] Epoch 5, batch 3000, loss[loss=0.1951, simple_loss=0.4488, pruned_loss=0.09943, ctc_loss=0.1399, over 14665.00 frames. ], tot_loss[loss=0.2194, simple_loss=0.4643, pruned_loss=0.1147, ctc_loss=0.1647, over 2872750.84 frames. ], batch size: 40, lr: 3.00e-03, 
2023-01-05 13:25:32,922 INFO [train.py:862] Epoch 5, batch 3500, loss[loss=0.2369, simple_loss=0.4862, pruned_loss=0.1252, ctc_loss=0.1806, over 14795.00 frames. ], tot_loss[loss=0.219, simple_loss=0.4633, pruned_loss=0.1146, ctc_loss=0.1645, over 2861368.73 frames. ], batch size: 43, lr: 3.00e-03, 
2023-01-05 13:26:38,177 INFO [train.py:862] Epoch 5, batch 4000, loss[loss=0.2161, simple_loss=0.4685, pruned_loss=0.1146, ctc_loss=0.1592, over 14664.00 frames. ], tot_loss[loss=0.219, simple_loss=0.4637, pruned_loss=0.1146, ctc_loss=0.1644, over 2873878.61 frames. ], batch size: 42, lr: 3.00e-03, 
2023-01-05 13:27:44,376 INFO [train.py:862] Epoch 5, batch 4500, loss[loss=0.219, simple_loss=0.4759, pruned_loss=0.1127, ctc_loss=0.1626, over 14143.00 frames. ], tot_loss[loss=0.2159, simple_loss=0.4613, pruned_loss=0.1124, ctc_loss=0.1614, over 2860984.27 frames. ], batch size: 56, lr: 3.00e-03, 
2023-01-05 13:28:49,949 INFO [train.py:862] Epoch 5, batch 5000, loss[loss=0.2372, simple_loss=0.4946, pruned_loss=0.1275, ctc_loss=0.1782, over 14542.00 frames. ], tot_loss[loss=0.2187, simple_loss=0.4632, pruned_loss=0.1143, ctc_loss=0.1642, over 2858386.61 frames. ], batch size: 50, lr: 3.00e-03, 
2023-01-05 13:29:56,281 INFO [train.py:862] Epoch 5, batch 5500, loss[loss=0.2566, simple_loss=0.4987, pruned_loss=0.1395, ctc_loss=0.2, over 14515.00 frames. ], tot_loss[loss=0.2195, simple_loss=0.4637, pruned_loss=0.1149, ctc_loss=0.165, over 2867764.13 frames. ], batch size: 50, lr: 3.00e-03, 
2023-01-05 13:31:02,294 INFO [train.py:862] Epoch 5, batch 6000, loss[loss=0.1661, simple_loss=0.4118, pruned_loss=0.08089, ctc_loss=0.1144, over 14551.00 frames. ], tot_loss[loss=0.2183, simple_loss=0.4622, pruned_loss=0.114, ctc_loss=0.1639, over 2871603.30 frames. ], batch size: 34, lr: 3.00e-03, 
2023-01-05 13:32:08,312 INFO [train.py:862] Epoch 5, batch 6500, loss[loss=0.1891, simple_loss=0.4696, pruned_loss=0.09052, ctc_loss=0.1307, over 14654.00 frames. ], tot_loss[loss=0.2174, simple_loss=0.4628, pruned_loss=0.1135, ctc_loss=0.1627, over 2879666.90 frames. ], batch size: 42, lr: 3.00e-03, 
2023-01-05 13:33:14,230 INFO [train.py:862] Epoch 5, batch 7000, loss[loss=0.2209, simple_loss=0.4726, pruned_loss=0.1158, ctc_loss=0.1647, over 14701.00 frames. ], tot_loss[loss=0.2189, simple_loss=0.4627, pruned_loss=0.1147, ctc_loss=0.1645, over 2857700.53 frames. ], batch size: 39, lr: 3.00e-03, 
2023-01-05 13:33:31,442 INFO [checkpoint.py:75] Saving checkpoint to tiny_transducer_ctc/exp_2m_phone_halfdelay/checkpoint-80000.pt
2023-01-05 13:34:20,267 INFO [train.py:862] Epoch 5, batch 7500, loss[loss=0.2129, simple_loss=0.4937, pruned_loss=0.1058, ctc_loss=0.1531, over 14509.00 frames. ], tot_loss[loss=0.2162, simple_loss=0.4608, pruned_loss=0.1128, ctc_loss=0.1618, over 2863989.57 frames. ], batch size: 53, lr: 3.00e-03, 
2023-01-05 13:35:26,540 INFO [train.py:862] Epoch 5, batch 8000, loss[loss=0.199, simple_loss=0.3982, pruned_loss=0.1013, ctc_loss=0.1556, over 14684.00 frames. ], tot_loss[loss=0.2185, simple_loss=0.4625, pruned_loss=0.1141, ctc_loss=0.1641, over 2855522.73 frames. ], batch size: 35, lr: 3.00e-03, 
2023-01-05 13:36:32,564 INFO [train.py:862] Epoch 5, batch 8500, loss[loss=0.233, simple_loss=0.4811, pruned_loss=0.1259, ctc_loss=0.1758, over 14069.00 frames. ], tot_loss[loss=0.217, simple_loss=0.4612, pruned_loss=0.1132, ctc_loss=0.1627, over 2859029.13 frames. ], batch size: 56, lr: 3.00e-03, 
2023-01-05 13:37:38,704 INFO [train.py:862] Epoch 5, batch 9000, loss[loss=0.3027, simple_loss=0.4951, pruned_loss=0.1759, ctc_loss=0.2509, over 14864.00 frames. ], tot_loss[loss=0.2152, simple_loss=0.4597, pruned_loss=0.1121, ctc_loss=0.1609, over 2864324.04 frames. ], batch size: 40, lr: 3.00e-03, 
2023-01-05 13:37:38,705 INFO [train.py:887] Computing validation loss
2023-01-05 13:38:04,456 INFO [train.py:897] Epoch 5, validation: loss=0.2225, simple_loss=0.4761, pruned_loss=0.117, ctc_loss=0.1656, over 944034.00 frames. 
2023-01-05 13:38:04,457 INFO [train.py:898] Maximum memory allocated so far is 7283MB
2023-01-05 13:39:10,920 INFO [train.py:862] Epoch 5, batch 9500, loss[loss=0.1705, simple_loss=0.3933, pruned_loss=0.08972, ctc_loss=0.1209, over 14668.00 frames. ], tot_loss[loss=0.2172, simple_loss=0.462, pruned_loss=0.1132, ctc_loss=0.1628, over 2851153.90 frames. ], batch size: 35, lr: 3.00e-03, 
2023-01-05 13:40:18,015 INFO [train.py:862] Epoch 5, batch 10000, loss[loss=0.1975, simple_loss=0.4325, pruned_loss=0.09899, ctc_loss=0.147, over 14689.00 frames. ], tot_loss[loss=0.2161, simple_loss=0.4602, pruned_loss=0.1125, ctc_loss=0.1619, over 2874922.40 frames. ], batch size: 38, lr: 3.00e-03, 
2023-01-05 13:41:23,887 INFO [train.py:862] Epoch 5, batch 10500, loss[loss=0.2232, simple_loss=0.4792, pruned_loss=0.1117, ctc_loss=0.1683, over 14685.00 frames. ], tot_loss[loss=0.2162, simple_loss=0.4617, pruned_loss=0.1126, ctc_loss=0.1617, over 2867047.67 frames. ], batch size: 47, lr: 3.00e-03, 
2023-01-05 13:42:29,845 INFO [train.py:862] Epoch 5, batch 11000, loss[loss=0.1889, simple_loss=0.4727, pruned_loss=0.08812, ctc_loss=0.1307, over 14673.00 frames. ], tot_loss[loss=0.2146, simple_loss=0.4602, pruned_loss=0.1114, ctc_loss=0.1603, over 2849607.69 frames. ], batch size: 47, lr: 3.00e-03, 
2023-01-05 13:43:36,585 INFO [train.py:862] Epoch 5, batch 11500, loss[loss=0.1974, simple_loss=0.4648, pruned_loss=0.09912, ctc_loss=0.1399, over 14643.00 frames. ], tot_loss[loss=0.2142, simple_loss=0.46, pruned_loss=0.1114, ctc_loss=0.1596, over 2864283.35 frames. ], batch size: 41, lr: 3.00e-03, 
2023-01-05 13:44:42,820 INFO [train.py:862] Epoch 5, batch 12000, loss[loss=0.1925, simple_loss=0.4159, pruned_loss=0.1008, ctc_loss=0.1427, over 14783.00 frames. ], tot_loss[loss=0.2159, simple_loss=0.4609, pruned_loss=0.1124, ctc_loss=0.1615, over 2866147.88 frames. ], batch size: 36, lr: 3.00e-03, 
2023-01-05 13:45:49,499 INFO [train.py:862] Epoch 5, batch 12500, loss[loss=0.2464, simple_loss=0.48, pruned_loss=0.132, ctc_loss=0.1925, over 14156.00 frames. ], tot_loss[loss=0.215, simple_loss=0.4609, pruned_loss=0.1114, ctc_loss=0.1606, over 2866654.51 frames. ], batch size: 56, lr: 3.00e-03, 
2023-01-05 13:46:55,729 INFO [train.py:862] Epoch 5, batch 13000, loss[loss=0.2392, simple_loss=0.4804, pruned_loss=0.1259, ctc_loss=0.1848, over 14658.00 frames. ], tot_loss[loss=0.2153, simple_loss=0.4615, pruned_loss=0.1117, ctc_loss=0.1607, over 2862304.69 frames. ], batch size: 40, lr: 3.00e-03, 
2023-01-05 13:48:02,263 INFO [train.py:862] Epoch 5, batch 13500, loss[loss=0.2316, simple_loss=0.4799, pruned_loss=0.1199, ctc_loss=0.1766, over 14726.00 frames. ], tot_loss[loss=0.2149, simple_loss=0.4614, pruned_loss=0.1112, ctc_loss=0.1605, over 2862981.50 frames. ], batch size: 46, lr: 3.00e-03, 
2023-01-05 13:49:09,290 INFO [train.py:862] Epoch 5, batch 14000, loss[loss=0.2431, simple_loss=0.4728, pruned_loss=0.1295, ctc_loss=0.1905, over 14656.00 frames. ], tot_loss[loss=0.2124, simple_loss=0.4578, pruned_loss=0.11, ctc_loss=0.1582, over 2876822.40 frames. ], batch size: 40, lr: 3.00e-03, 
2023-01-05 13:50:15,574 INFO [train.py:862] Epoch 5, batch 14500, loss[loss=0.1752, simple_loss=0.4098, pruned_loss=0.08916, ctc_loss=0.1242, over 14398.00 frames. ], tot_loss[loss=0.2154, simple_loss=0.4602, pruned_loss=0.1119, ctc_loss=0.1612, over 2849185.96 frames. ], batch size: 33, lr: 3.00e-03, 
2023-01-05 13:51:22,393 INFO [train.py:862] Epoch 5, batch 15000, loss[loss=0.2209, simple_loss=0.4412, pruned_loss=0.116, ctc_loss=0.1713, over 14725.00 frames. ], tot_loss[loss=0.2143, simple_loss=0.4599, pruned_loss=0.1111, ctc_loss=0.16, over 2853910.25 frames. ], batch size: 37, lr: 3.00e-03, 
2023-01-05 13:52:28,607 INFO [train.py:862] Epoch 5, batch 15500, loss[loss=0.1686, simple_loss=0.3941, pruned_loss=0.08428, ctc_loss=0.1202, over 14522.00 frames. ], tot_loss[loss=0.2132, simple_loss=0.4586, pruned_loss=0.1104, ctc_loss=0.1589, over 2868074.67 frames. ], batch size: 34, lr: 3.00e-03, 
2023-01-05 13:53:35,484 INFO [train.py:862] Epoch 5, batch 16000, loss[loss=0.225, simple_loss=0.4764, pruned_loss=0.1191, ctc_loss=0.1683, over 14654.00 frames. ], tot_loss[loss=0.2138, simple_loss=0.459, pruned_loss=0.111, ctc_loss=0.1595, over 2853738.05 frames. ], batch size: 42, lr: 3.00e-03, 
2023-01-05 13:54:41,768 INFO [train.py:862] Epoch 5, batch 16500, loss[loss=0.1992, simple_loss=0.4218, pruned_loss=0.1056, ctc_loss=0.149, over 14667.00 frames. ], tot_loss[loss=0.2144, simple_loss=0.4603, pruned_loss=0.1115, ctc_loss=0.1599, over 2866639.66 frames. ], batch size: 35, lr: 3.00e-03, 
2023-01-05 13:55:48,046 INFO [train.py:862] Epoch 5, batch 17000, loss[loss=0.1879, simple_loss=0.4154, pruned_loss=0.09606, ctc_loss=0.1383, over 14538.00 frames. ], tot_loss[loss=0.2118, simple_loss=0.4576, pruned_loss=0.1096, ctc_loss=0.1575, over 2862763.07 frames. ], batch size: 36, lr: 3.00e-03, 
2023-01-05 13:56:06,032 INFO [checkpoint.py:75] Saving checkpoint to tiny_transducer_ctc/exp_2m_phone_halfdelay/checkpoint-90000.pt
2023-01-05 13:56:55,195 INFO [train.py:862] Epoch 5, batch 17500, loss[loss=0.2081, simple_loss=0.4693, pruned_loss=0.1047, ctc_loss=0.1518, over 14547.00 frames. ], tot_loss[loss=0.2127, simple_loss=0.4586, pruned_loss=0.1099, ctc_loss=0.1585, over 2850173.14 frames. ], batch size: 53, lr: 3.00e-03, 
2023-01-05 13:58:00,243 INFO [train.py:862] Epoch 5, batch 18000, loss[loss=0.2665, simple_loss=0.4999, pruned_loss=0.1448, ctc_loss=0.2115, over 9873.00 frames. ], tot_loss[loss=0.2152, simple_loss=0.4608, pruned_loss=0.1117, ctc_loss=0.1609, over 2862424.09 frames. ], batch size: 105, lr: 3.00e-03, 
2023-01-05 13:58:00,244 INFO [train.py:887] Computing validation loss
2023-01-05 13:58:25,967 INFO [train.py:897] Epoch 5, validation: loss=0.2225, simple_loss=0.4754, pruned_loss=0.1176, ctc_loss=0.1655, over 944034.00 frames. 
2023-01-05 13:58:25,968 INFO [train.py:898] Maximum memory allocated so far is 7283MB
2023-01-05 13:58:53,305 INFO [checkpoint.py:75] Saving checkpoint to tiny_transducer_ctc/exp_2m_phone_halfdelay/epoch-5.pt
2023-01-05 13:58:56,296 INFO [train.py:862] Epoch 6, batch 0, loss[loss=0.2517, simple_loss=0.4877, pruned_loss=0.1387, ctc_loss=0.1956, over 14694.00 frames. ], tot_loss[loss=0.2517, simple_loss=0.4877, pruned_loss=0.1387, ctc_loss=0.1956, over 14694.00 frames. ], batch size: 39, lr: 3.00e-03, 
2023-01-05 14:00:02,235 INFO [train.py:862] Epoch 6, batch 500, loss[loss=0.2069, simple_loss=0.4839, pruned_loss=0.1079, ctc_loss=0.1456, over 14661.00 frames. ], tot_loss[loss=0.2142, simple_loss=0.4605, pruned_loss=0.1109, ctc_loss=0.1599, over 2647350.96 frames. ], batch size: 41, lr: 3.00e-03, 
2023-01-05 14:01:07,848 INFO [train.py:862] Epoch 6, batch 1000, loss[loss=0.1552, simple_loss=0.3931, pruned_loss=0.07896, ctc_loss=0.1036, over 14660.00 frames. ], tot_loss[loss=0.2119, simple_loss=0.4591, pruned_loss=0.1093, ctc_loss=0.1574, over 2851882.20 frames. ], batch size: 35, lr: 3.00e-03, 
2023-01-05 14:02:14,001 INFO [train.py:862] Epoch 6, batch 1500, loss[loss=0.2474, simple_loss=0.4773, pruned_loss=0.1331, ctc_loss=0.1941, over 10331.00 frames. ], tot_loss[loss=0.2134, simple_loss=0.4596, pruned_loss=0.1105, ctc_loss=0.159, over 2859773.10 frames. ], batch size: 104, lr: 3.00e-03, 
2023-01-05 14:03:20,122 INFO [train.py:862] Epoch 6, batch 2000, loss[loss=0.1765, simple_loss=0.4365, pruned_loss=0.08871, ctc_loss=0.1206, over 14711.00 frames. ], tot_loss[loss=0.2122, simple_loss=0.4596, pruned_loss=0.1094, ctc_loss=0.1578, over 2877040.02 frames. ], batch size: 38, lr: 3.00e-03, 
2023-01-05 14:04:26,398 INFO [train.py:862] Epoch 6, batch 2500, loss[loss=0.2114, simple_loss=0.4881, pruned_loss=0.1069, ctc_loss=0.1516, over 14654.00 frames. ], tot_loss[loss=0.2117, simple_loss=0.4592, pruned_loss=0.1092, ctc_loss=0.1573, over 2880167.66 frames. ], batch size: 41, lr: 3.00e-03, 
2023-01-05 14:05:32,057 INFO [train.py:862] Epoch 6, batch 3000, loss[loss=0.2029, simple_loss=0.4299, pruned_loss=0.1068, ctc_loss=0.152, over 14645.00 frames. ], tot_loss[loss=0.212, simple_loss=0.459, pruned_loss=0.1095, ctc_loss=0.1575, over 2870515.54 frames. ], batch size: 35, lr: 3.00e-03, 
2023-01-05 14:06:37,523 INFO [train.py:862] Epoch 6, batch 3500, loss[loss=0.2295, simple_loss=0.4874, pruned_loss=0.1175, ctc_loss=0.1731, over 14522.00 frames. ], tot_loss[loss=0.2127, simple_loss=0.4593, pruned_loss=0.1098, ctc_loss=0.1583, over 2869887.03 frames. ], batch size: 48, lr: 3.00e-03, 
2023-01-05 14:07:43,304 INFO [train.py:862] Epoch 6, batch 4000, loss[loss=0.2263, simple_loss=0.4772, pruned_loss=0.1182, ctc_loss=0.1704, over 14563.00 frames. ], tot_loss[loss=0.2117, simple_loss=0.4573, pruned_loss=0.1094, ctc_loss=0.1576, over 2874134.49 frames. ], batch size: 48, lr: 3.00e-03, 
2023-01-05 14:08:48,825 INFO [train.py:862] Epoch 6, batch 4500, loss[loss=0.2335, simple_loss=0.4859, pruned_loss=0.1191, ctc_loss=0.1784, over 14120.00 frames. ], tot_loss[loss=0.2144, simple_loss=0.4599, pruned_loss=0.1112, ctc_loss=0.1601, over 2869597.01 frames. ], batch size: 56, lr: 3.00e-03, 
2023-01-05 14:09:54,078 INFO [train.py:862] Epoch 6, batch 5000, loss[loss=0.2138, simple_loss=0.4541, pruned_loss=0.1087, ctc_loss=0.1615, over 14844.00 frames. ], tot_loss[loss=0.2091, simple_loss=0.4552, pruned_loss=0.1076, ctc_loss=0.1551, over 2854497.74 frames. ], batch size: 42, lr: 3.00e-03, 
2023-01-05 14:11:00,553 INFO [train.py:862] Epoch 6, batch 5500, loss[loss=0.1517, simple_loss=0.4055, pruned_loss=0.07176, ctc_loss=0.09905, over 14705.00 frames. ], tot_loss[loss=0.2096, simple_loss=0.4576, pruned_loss=0.1078, ctc_loss=0.1552, over 2856727.58 frames. ], batch size: 37, lr: 3.00e-03, 
2023-01-05 14:12:06,297 INFO [train.py:862] Epoch 6, batch 6000, loss[loss=0.2457, simple_loss=0.4829, pruned_loss=0.132, ctc_loss=0.1909, over 14723.00 frames. ], tot_loss[loss=0.2095, simple_loss=0.4567, pruned_loss=0.1076, ctc_loss=0.1553, over 2867663.57 frames. ], batch size: 45, lr: 3.00e-03, 
2023-01-05 14:13:12,281 INFO [train.py:862] Epoch 6, batch 6500, loss[loss=0.1554, simple_loss=0.4032, pruned_loss=0.07267, ctc_loss=0.1044, over 14515.00 frames. ], tot_loss[loss=0.2109, simple_loss=0.4568, pruned_loss=0.1089, ctc_loss=0.1568, over 2856925.73 frames. ], batch size: 38, lr: 3.00e-03, 
2023-01-05 14:14:17,552 INFO [train.py:862] Epoch 6, batch 7000, loss[loss=0.2305, simple_loss=0.4395, pruned_loss=0.1309, ctc_loss=0.1789, over 14390.00 frames. ], tot_loss[loss=0.2104, simple_loss=0.4573, pruned_loss=0.1083, ctc_loss=0.1562, over 2857070.40 frames. ], batch size: 33, lr: 3.00e-03, 
2023-01-05 14:15:23,291 INFO [train.py:862] Epoch 6, batch 7500, loss[loss=0.2551, simple_loss=0.4981, pruned_loss=0.1451, ctc_loss=0.1956, over 14536.00 frames. ], tot_loss[loss=0.2096, simple_loss=0.4569, pruned_loss=0.1081, ctc_loss=0.1552, over 2860948.64 frames. ], batch size: 48, lr: 3.00e-03, 
2023-01-05 14:16:29,035 INFO [train.py:862] Epoch 6, batch 8000, loss[loss=0.2134, simple_loss=0.47, pruned_loss=0.1124, ctc_loss=0.1559, over 14702.00 frames. ], tot_loss[loss=0.2105, simple_loss=0.4569, pruned_loss=0.1086, ctc_loss=0.1563, over 2858352.98 frames. ], batch size: 47, lr: 3.00e-03, 
2023-01-05 14:17:34,290 INFO [train.py:862] Epoch 6, batch 8500, loss[loss=0.2001, simple_loss=0.4402, pruned_loss=0.101, ctc_loss=0.1482, over 14896.00 frames. ], tot_loss[loss=0.2099, simple_loss=0.4572, pruned_loss=0.1082, ctc_loss=0.1556, over 2860818.22 frames. ], batch size: 39, lr: 3.00e-03, 
2023-01-05 14:18:28,603 INFO [checkpoint.py:75] Saving checkpoint to tiny_transducer_ctc/exp_2m_phone_halfdelay/checkpoint-100000.pt
2023-01-05 14:18:39,952 INFO [train.py:862] Epoch 6, batch 9000, loss[loss=0.2123, simple_loss=0.4604, pruned_loss=0.1101, ctc_loss=0.1575, over 14834.00 frames. ], tot_loss[loss=0.2098, simple_loss=0.4565, pruned_loss=0.108, ctc_loss=0.1556, over 2871433.85 frames. ], batch size: 42, lr: 3.00e-03, 
2023-01-05 14:18:39,953 INFO [train.py:887] Computing validation loss
2023-01-05 14:19:05,305 INFO [train.py:897] Epoch 6, validation: loss=0.2188, simple_loss=0.4732, pruned_loss=0.1136, ctc_loss=0.1625, over 944034.00 frames. 
2023-01-05 14:19:05,306 INFO [train.py:898] Maximum memory allocated so far is 7283MB
2023-01-05 14:20:10,556 INFO [train.py:862] Epoch 6, batch 9500, loss[loss=0.2183, simple_loss=0.4752, pruned_loss=0.1048, ctc_loss=0.1651, over 14128.00 frames. ], tot_loss[loss=0.2105, simple_loss=0.4575, pruned_loss=0.1082, ctc_loss=0.1564, over 2869535.36 frames. ], batch size: 56, lr: 3.00e-03, 
2023-01-05 14:21:16,519 INFO [train.py:862] Epoch 6, batch 10000, loss[loss=0.1824, simple_loss=0.4385, pruned_loss=0.08773, ctc_loss=0.129, over 14664.00 frames. ], tot_loss[loss=0.2097, simple_loss=0.4568, pruned_loss=0.1076, ctc_loss=0.1556, over 2867631.64 frames. ], batch size: 40, lr: 3.00e-03, 
2023-01-05 14:22:22,114 INFO [train.py:862] Epoch 6, batch 10500, loss[loss=0.2126, simple_loss=0.4721, pruned_loss=0.1068, ctc_loss=0.1567, over 14498.00 frames. ], tot_loss[loss=0.209, simple_loss=0.4561, pruned_loss=0.1073, ctc_loss=0.1549, over 2854176.50 frames. ], batch size: 53, lr: 3.00e-03, 
2023-01-05 14:23:27,371 INFO [train.py:862] Epoch 6, batch 11000, loss[loss=0.2368, simple_loss=0.4878, pruned_loss=0.1235, ctc_loss=0.1809, over 14647.00 frames. ], tot_loss[loss=0.2094, simple_loss=0.4564, pruned_loss=0.1075, ctc_loss=0.1552, over 2855120.49 frames. ], batch size: 41, lr: 3.00e-03, 
2023-01-05 14:24:33,353 INFO [train.py:862] Epoch 6, batch 11500, loss[loss=0.1925, simple_loss=0.4593, pruned_loss=0.09726, ctc_loss=0.1349, over 14733.00 frames. ], tot_loss[loss=0.2097, simple_loss=0.4575, pruned_loss=0.1079, ctc_loss=0.1553, over 2865434.25 frames. ], batch size: 46, lr: 3.00e-03, 
2023-01-05 14:25:38,531 INFO [train.py:862] Epoch 6, batch 12000, loss[loss=0.1956, simple_loss=0.435, pruned_loss=0.09625, ctc_loss=0.145, over 14524.00 frames. ], tot_loss[loss=0.2089, simple_loss=0.4564, pruned_loss=0.1069, ctc_loss=0.1548, over 2858287.02 frames. ], batch size: 36, lr: 3.00e-03, 
2023-01-05 14:26:44,416 INFO [train.py:862] Epoch 6, batch 12500, loss[loss=0.2328, simple_loss=0.5012, pruned_loss=0.119, ctc_loss=0.1741, over 14722.00 frames. ], tot_loss[loss=0.21, simple_loss=0.4578, pruned_loss=0.108, ctc_loss=0.1555, over 2851004.78 frames. ], batch size: 45, lr: 3.00e-03, 
2023-01-05 14:27:49,994 INFO [train.py:862] Epoch 6, batch 13000, loss[loss=0.2057, simple_loss=0.475, pruned_loss=0.1039, ctc_loss=0.1475, over 14499.00 frames. ], tot_loss[loss=0.2075, simple_loss=0.4554, pruned_loss=0.1063, ctc_loss=0.1533, over 2866377.35 frames. ], batch size: 50, lr: 3.00e-03, 
2023-01-05 14:28:55,647 INFO [train.py:862] Epoch 6, batch 13500, loss[loss=0.2164, simple_loss=0.4259, pruned_loss=0.1163, ctc_loss=0.168, over 14411.00 frames. ], tot_loss[loss=0.2087, simple_loss=0.4577, pruned_loss=0.1071, ctc_loss=0.1542, over 2840539.01 frames. ], batch size: 33, lr: 3.00e-03, 
2023-01-05 14:30:01,616 INFO [train.py:862] Epoch 6, batch 14000, loss[loss=0.2178, simple_loss=0.4284, pruned_loss=0.12, ctc_loss=0.1679, over 14409.00 frames. ], tot_loss[loss=0.2097, simple_loss=0.4569, pruned_loss=0.1075, ctc_loss=0.1556, over 2864943.84 frames. ], batch size: 33, lr: 3.00e-03, 
2023-01-05 14:31:07,329 INFO [train.py:862] Epoch 6, batch 14500, loss[loss=0.219, simple_loss=0.422, pruned_loss=0.1185, ctc_loss=0.1717, over 14000.00 frames. ], tot_loss[loss=0.2071, simple_loss=0.4535, pruned_loss=0.1064, ctc_loss=0.1532, over 2879030.15 frames. ], batch size: 31, lr: 3.00e-03, 
2023-01-05 14:32:13,435 INFO [train.py:862] Epoch 6, batch 15000, loss[loss=0.2548, simple_loss=0.4923, pruned_loss=0.136, ctc_loss=0.2003, over 10316.00 frames. ], tot_loss[loss=0.2079, simple_loss=0.4548, pruned_loss=0.1063, ctc_loss=0.1539, over 2871298.08 frames. ], batch size: 104, lr: 3.00e-03, 
2023-01-05 14:33:18,911 INFO [train.py:862] Epoch 6, batch 15500, loss[loss=0.2548, simple_loss=0.4793, pruned_loss=0.1398, ctc_loss=0.2014, over 14802.00 frames. ], tot_loss[loss=0.2084, simple_loss=0.4558, pruned_loss=0.1068, ctc_loss=0.1542, over 2854584.99 frames. ], batch size: 41, lr: 3.00e-03, 
2023-01-05 14:34:25,124 INFO [train.py:862] Epoch 6, batch 16000, loss[loss=0.2468, simple_loss=0.5007, pruned_loss=0.1231, ctc_loss=0.1926, over 14525.00 frames. ], tot_loss[loss=0.21, simple_loss=0.4561, pruned_loss=0.1079, ctc_loss=0.156, over 2856729.88 frames. ], batch size: 48, lr: 3.00e-03, 
2023-01-05 14:35:30,435 INFO [train.py:862] Epoch 6, batch 16500, loss[loss=0.1738, simple_loss=0.4204, pruned_loss=0.08845, ctc_loss=0.1203, over 14497.00 frames. ], tot_loss[loss=0.2063, simple_loss=0.4544, pruned_loss=0.1054, ctc_loss=0.1521, over 2849857.22 frames. ], batch size: 38, lr: 3.00e-03, 
2023-01-05 14:36:36,399 INFO [train.py:862] Epoch 6, batch 17000, loss[loss=0.2026, simple_loss=0.4611, pruned_loss=0.103, ctc_loss=0.1466, over 14821.00 frames. ], tot_loss[loss=0.2061, simple_loss=0.4533, pruned_loss=0.1055, ctc_loss=0.1521, over 2844171.94 frames. ], batch size: 41, lr: 3.00e-03, 
2023-01-05 14:37:41,659 INFO [train.py:862] Epoch 6, batch 17500, loss[loss=0.2185, simple_loss=0.4867, pruned_loss=0.1095, ctc_loss=0.1609, over 14742.00 frames. ], tot_loss[loss=0.2075, simple_loss=0.4557, pruned_loss=0.1062, ctc_loss=0.1533, over 2865197.28 frames. ], batch size: 45, lr: 3.00e-03, 
2023-01-05 14:38:45,201 INFO [train.py:862] Epoch 6, batch 18000, loss[loss=0.2215, simple_loss=0.4612, pruned_loss=0.1143, ctc_loss=0.1687, over 14552.00 frames. ], tot_loss[loss=0.2077, simple_loss=0.4555, pruned_loss=0.1061, ctc_loss=0.1537, over 2868368.20 frames. ], batch size: 48, lr: 3.00e-03, 
2023-01-05 14:38:45,202 INFO [train.py:887] Computing validation loss
2023-01-05 14:39:11,024 INFO [train.py:897] Epoch 6, validation: loss=0.2192, simple_loss=0.4741, pruned_loss=0.1145, ctc_loss=0.1625, over 944034.00 frames. 
2023-01-05 14:39:11,025 INFO [train.py:898] Maximum memory allocated so far is 7283MB
2023-01-05 14:39:37,583 INFO [checkpoint.py:75] Saving checkpoint to tiny_transducer_ctc/exp_2m_phone_halfdelay/epoch-6.pt
2023-01-05 14:39:40,570 INFO [train.py:862] Epoch 7, batch 0, loss[loss=0.2366, simple_loss=0.4515, pruned_loss=0.127, ctc_loss=0.1868, over 14779.00 frames. ], tot_loss[loss=0.2366, simple_loss=0.4515, pruned_loss=0.127, ctc_loss=0.1868, over 14779.00 frames. ], batch size: 36, lr: 3.00e-03, 
2023-01-05 14:40:46,554 INFO [train.py:862] Epoch 7, batch 500, loss[loss=0.2083, simple_loss=0.4403, pruned_loss=0.1086, ctc_loss=0.1566, over 14533.00 frames. ], tot_loss[loss=0.2092, simple_loss=0.4568, pruned_loss=0.1071, ctc_loss=0.155, over 2633483.34 frames. ], batch size: 36, lr: 3.00e-03, 
2023-01-05 14:41:12,578 INFO [checkpoint.py:75] Saving checkpoint to tiny_transducer_ctc/exp_2m_phone_halfdelay/checkpoint-110000.pt
2023-01-05 14:41:52,231 INFO [train.py:862] Epoch 7, batch 1000, loss[loss=0.1658, simple_loss=0.3909, pruned_loss=0.08239, ctc_loss=0.1179, over 14522.00 frames. ], tot_loss[loss=0.2059, simple_loss=0.4537, pruned_loss=0.105, ctc_loss=0.152, over 2846089.20 frames. ], batch size: 34, lr: 3.00e-03, 
2023-01-05 14:42:57,565 INFO [train.py:862] Epoch 7, batch 1500, loss[loss=0.2035, simple_loss=0.462, pruned_loss=0.1061, ctc_loss=0.1462, over 14797.00 frames. ], tot_loss[loss=0.2075, simple_loss=0.4561, pruned_loss=0.106, ctc_loss=0.1533, over 2862820.19 frames. ], batch size: 43, lr: 3.00e-03, 
2023-01-05 14:44:03,115 INFO [train.py:862] Epoch 7, batch 2000, loss[loss=0.1979, simple_loss=0.4763, pruned_loss=0.0949, ctc_loss=0.14, over 14676.00 frames. ], tot_loss[loss=0.2063, simple_loss=0.4555, pruned_loss=0.1052, ctc_loss=0.1521, over 2869421.92 frames. ], batch size: 44, lr: 3.00e-03, 
2023-01-05 14:45:08,821 INFO [train.py:862] Epoch 7, batch 2500, loss[loss=0.1522, simple_loss=0.4004, pruned_loss=0.07321, ctc_loss=0.1003, over 14722.00 frames. ], tot_loss[loss=0.2075, simple_loss=0.4553, pruned_loss=0.1063, ctc_loss=0.1533, over 2858290.49 frames. ], batch size: 37, lr: 3.00e-03, 
2023-01-05 14:46:14,251 INFO [train.py:862] Epoch 7, batch 3000, loss[loss=0.2022, simple_loss=0.4702, pruned_loss=0.09499, ctc_loss=0.1473, over 14701.00 frames. ], tot_loss[loss=0.2083, simple_loss=0.4565, pruned_loss=0.1064, ctc_loss=0.1541, over 2875657.15 frames. ], batch size: 44, lr: 3.00e-03, 
2023-01-05 14:47:19,760 INFO [train.py:862] Epoch 7, batch 3500, loss[loss=0.184, simple_loss=0.4451, pruned_loss=0.08583, ctc_loss=0.1307, over 14816.00 frames. ], tot_loss[loss=0.2067, simple_loss=0.4547, pruned_loss=0.1053, ctc_loss=0.1527, over 2875625.70 frames. ], batch size: 41, lr: 3.00e-03, 
2023-01-05 14:48:25,087 INFO [train.py:862] Epoch 7, batch 4000, loss[loss=0.2031, simple_loss=0.4483, pruned_loss=0.1055, ctc_loss=0.1488, over 14604.00 frames. ], tot_loss[loss=0.2059, simple_loss=0.4538, pruned_loss=0.1049, ctc_loss=0.1519, over 2865804.54 frames. ], batch size: 43, lr: 3.00e-03, 
2023-01-05 14:49:29,928 INFO [train.py:862] Epoch 7, batch 4500, loss[loss=0.1814, simple_loss=0.4085, pruned_loss=0.09171, ctc_loss=0.1323, over 14406.00 frames. ], tot_loss[loss=0.2055, simple_loss=0.4526, pruned_loss=0.1048, ctc_loss=0.1516, over 2868236.91 frames. ], batch size: 33, lr: 3.00e-03, 
2023-01-05 14:50:35,442 INFO [train.py:862] Epoch 7, batch 5000, loss[loss=0.2274, simple_loss=0.4622, pruned_loss=0.1234, ctc_loss=0.173, over 14847.00 frames. ], tot_loss[loss=0.2057, simple_loss=0.4538, pruned_loss=0.1047, ctc_loss=0.1518, over 2856155.81 frames. ], batch size: 40, lr: 3.00e-03, 
2023-01-05 14:51:40,575 INFO [train.py:862] Epoch 7, batch 5500, loss[loss=0.1889, simple_loss=0.4038, pruned_loss=0.09501, ctc_loss=0.1426, over 14784.00 frames. ], tot_loss[loss=0.2061, simple_loss=0.4538, pruned_loss=0.105, ctc_loss=0.1522, over 2869147.30 frames. ], batch size: 36, lr: 3.00e-03, 
2023-01-05 14:52:46,401 INFO [train.py:862] Epoch 7, batch 6000, loss[loss=0.2155, simple_loss=0.468, pruned_loss=0.1103, ctc_loss=0.1603, over 14118.00 frames. ], tot_loss[loss=0.2078, simple_loss=0.4557, pruned_loss=0.1063, ctc_loss=0.1537, over 2866647.64 frames. ], batch size: 56, lr: 3.00e-03, 
2023-01-05 14:53:51,528 INFO [train.py:862] Epoch 7, batch 6500, loss[loss=0.2017, simple_loss=0.457, pruned_loss=0.1015, ctc_loss=0.1468, over 13514.00 frames. ], tot_loss[loss=0.2051, simple_loss=0.4537, pruned_loss=0.1045, ctc_loss=0.1509, over 2855984.14 frames. ], batch size: 63, lr: 3.00e-03, 
2023-01-05 14:54:56,969 INFO [train.py:862] Epoch 7, batch 7000, loss[loss=0.2092, simple_loss=0.4275, pruned_loss=0.1064, ctc_loss=0.1617, over 14698.00 frames. ], tot_loss[loss=0.2049, simple_loss=0.4526, pruned_loss=0.1046, ctc_loss=0.151, over 2871494.75 frames. ], batch size: 35, lr: 3.00e-03, 
2023-01-05 14:56:02,382 INFO [train.py:862] Epoch 7, batch 7500, loss[loss=0.1725, simple_loss=0.435, pruned_loss=0.08403, ctc_loss=0.1173, over 14851.00 frames. ], tot_loss[loss=0.2036, simple_loss=0.4518, pruned_loss=0.1034, ctc_loss=0.1497, over 2868398.56 frames. ], batch size: 42, lr: 3.00e-03, 
2023-01-05 14:57:07,842 INFO [train.py:862] Epoch 7, batch 8000, loss[loss=0.225, simple_loss=0.4377, pruned_loss=0.1169, ctc_loss=0.1775, over 14729.00 frames. ], tot_loss[loss=0.2034, simple_loss=0.4508, pruned_loss=0.1034, ctc_loss=0.1496, over 2878586.87 frames. ], batch size: 38, lr: 3.00e-03, 
2023-01-05 14:58:13,097 INFO [train.py:862] Epoch 7, batch 8500, loss[loss=0.167, simple_loss=0.3996, pruned_loss=0.08342, ctc_loss=0.1172, over 14380.00 frames. ], tot_loss[loss=0.2038, simple_loss=0.4519, pruned_loss=0.1036, ctc_loss=0.1499, over 2876628.16 frames. ], batch size: 32, lr: 3.00e-03, 
2023-01-05 14:59:18,920 INFO [train.py:862] Epoch 7, batch 9000, loss[loss=0.1599, simple_loss=0.4332, pruned_loss=0.07197, ctc_loss=0.1047, over 14520.00 frames. ], tot_loss[loss=0.2072, simple_loss=0.4548, pruned_loss=0.1059, ctc_loss=0.1531, over 2871090.90 frames. ], batch size: 38, lr: 3.00e-03, 
2023-01-05 14:59:18,921 INFO [train.py:887] Computing validation loss
2023-01-05 14:59:43,987 INFO [train.py:897] Epoch 7, validation: loss=0.2169, simple_loss=0.4718, pruned_loss=0.1113, ctc_loss=0.1611, over 944034.00 frames. 
2023-01-05 14:59:43,987 INFO [train.py:898] Maximum memory allocated so far is 7283MB
2023-01-05 15:00:49,748 INFO [train.py:862] Epoch 7, batch 9500, loss[loss=0.2889, simple_loss=0.5208, pruned_loss=0.1652, ctc_loss=0.2303, over 14799.00 frames. ], tot_loss[loss=0.2038, simple_loss=0.4523, pruned_loss=0.1037, ctc_loss=0.1497, over 2865710.49 frames. ], batch size: 43, lr: 3.00e-03, 
2023-01-05 15:01:55,016 INFO [train.py:862] Epoch 7, batch 10000, loss[loss=0.1806, simple_loss=0.4283, pruned_loss=0.08641, ctc_loss=0.1292, over 14689.00 frames. ], tot_loss[loss=0.2041, simple_loss=0.4532, pruned_loss=0.1037, ctc_loss=0.1499, over 2870573.70 frames. ], batch size: 39, lr: 3.00e-03, 
2023-01-05 15:03:00,337 INFO [train.py:862] Epoch 7, batch 10500, loss[loss=0.1661, simple_loss=0.4363, pruned_loss=0.0806, ctc_loss=0.1093, over 14820.00 frames. ], tot_loss[loss=0.2037, simple_loss=0.4535, pruned_loss=0.1031, ctc_loss=0.1497, over 2869174.27 frames. ], batch size: 41, lr: 3.00e-03, 
2023-01-05 15:03:26,214 INFO [checkpoint.py:75] Saving checkpoint to tiny_transducer_ctc/exp_2m_phone_halfdelay/checkpoint-120000.pt
2023-01-05 15:04:05,768 INFO [train.py:862] Epoch 7, batch 11000, loss[loss=0.1961, simple_loss=0.4625, pruned_loss=0.09738, ctc_loss=0.1393, over 14534.00 frames. ], tot_loss[loss=0.2045, simple_loss=0.4533, pruned_loss=0.1037, ctc_loss=0.1505, over 2868621.12 frames. ], batch size: 48, lr: 3.00e-03, 
2023-01-05 15:05:11,038 INFO [train.py:862] Epoch 7, batch 11500, loss[loss=0.1561, simple_loss=0.4037, pruned_loss=0.07734, ctc_loss=0.1033, over 14809.00 frames. ], tot_loss[loss=0.2038, simple_loss=0.4531, pruned_loss=0.1036, ctc_loss=0.1496, over 2858484.45 frames. ], batch size: 36, lr: 3.00e-03, 
2023-01-05 15:06:16,855 INFO [train.py:862] Epoch 7, batch 12000, loss[loss=0.1795, simple_loss=0.4475, pruned_loss=0.08726, ctc_loss=0.1232, over 14673.00 frames. ], tot_loss[loss=0.2042, simple_loss=0.4528, pruned_loss=0.1036, ctc_loss=0.1503, over 2865563.61 frames. ], batch size: 39, lr: 3.00e-03, 
2023-01-05 15:07:22,413 INFO [train.py:862] Epoch 7, batch 12500, loss[loss=0.2213, simple_loss=0.4838, pruned_loss=0.1144, ctc_loss=0.1634, over 14566.00 frames. ], tot_loss[loss=0.2053, simple_loss=0.4541, pruned_loss=0.1043, ctc_loss=0.1513, over 2857548.22 frames. ], batch size: 50, lr: 3.00e-03, 
2023-01-05 15:08:27,675 INFO [train.py:862] Epoch 7, batch 13000, loss[loss=0.1778, simple_loss=0.4196, pruned_loss=0.08514, ctc_loss=0.1276, over 14527.00 frames. ], tot_loss[loss=0.2041, simple_loss=0.4534, pruned_loss=0.1036, ctc_loss=0.1501, over 2869695.81 frames. ], batch size: 37, lr: 3.00e-03, 
2023-01-05 15:09:33,626 INFO [train.py:862] Epoch 7, batch 13500, loss[loss=0.192, simple_loss=0.4095, pruned_loss=0.1055, ctc_loss=0.1413, over 14105.00 frames. ], tot_loss[loss=0.2032, simple_loss=0.4521, pruned_loss=0.1033, ctc_loss=0.1491, over 2854979.33 frames. ], batch size: 31, lr: 3.00e-03, 
2023-01-05 15:10:39,175 INFO [train.py:862] Epoch 7, batch 14000, loss[loss=0.2182, simple_loss=0.4641, pruned_loss=0.1093, ctc_loss=0.1654, over 14527.00 frames. ], tot_loss[loss=0.2051, simple_loss=0.4536, pruned_loss=0.104, ctc_loss=0.1513, over 2858584.72 frames. ], batch size: 48, lr: 3.00e-03, 
2023-01-05 15:11:45,253 INFO [train.py:862] Epoch 7, batch 14500, loss[loss=0.1908, simple_loss=0.4104, pruned_loss=0.0949, ctc_loss=0.1439, over 13654.00 frames. ], tot_loss[loss=0.2036, simple_loss=0.4511, pruned_loss=0.1034, ctc_loss=0.1499, over 2867630.06 frames. ], batch size: 30, lr: 3.00e-03, 
2023-01-05 15:12:50,321 INFO [train.py:862] Epoch 7, batch 15000, loss[loss=0.1665, simple_loss=0.4136, pruned_loss=0.08242, ctc_loss=0.1139, over 14518.00 frames. ], tot_loss[loss=0.2047, simple_loss=0.4528, pruned_loss=0.1038, ctc_loss=0.151, over 2859808.85 frames. ], batch size: 37, lr: 3.00e-03, 
2023-01-05 15:13:55,728 INFO [train.py:862] Epoch 7, batch 15500, loss[loss=0.1604, simple_loss=0.4184, pruned_loss=0.0765, ctc_loss=0.1067, over 14669.00 frames. ], tot_loss[loss=0.2025, simple_loss=0.4509, pruned_loss=0.1025, ctc_loss=0.1487, over 2857797.85 frames. ], batch size: 40, lr: 3.00e-03, 
2023-01-05 15:15:01,326 INFO [train.py:862] Epoch 7, batch 16000, loss[loss=0.1933, simple_loss=0.4336, pruned_loss=0.09707, ctc_loss=0.1416, over 14698.00 frames. ], tot_loss[loss=0.2041, simple_loss=0.4532, pruned_loss=0.1038, ctc_loss=0.15, over 2842591.65 frames. ], batch size: 38, lr: 3.00e-03, 
2023-01-05 15:16:06,834 INFO [train.py:862] Epoch 7, batch 16500, loss[loss=0.1692, simple_loss=0.4166, pruned_loss=0.07964, ctc_loss=0.1184, over 14718.00 frames. ], tot_loss[loss=0.2024, simple_loss=0.4524, pruned_loss=0.1023, ctc_loss=0.1483, over 2832421.51 frames. ], batch size: 37, lr: 3.00e-03, 
2023-01-05 15:17:12,437 INFO [train.py:862] Epoch 7, batch 17000, loss[loss=0.1804, simple_loss=0.4556, pruned_loss=0.08394, ctc_loss=0.1242, over 14719.00 frames. ], tot_loss[loss=0.2043, simple_loss=0.4526, pruned_loss=0.1037, ctc_loss=0.1505, over 2867525.12 frames. ], batch size: 46, lr: 3.00e-03, 
2023-01-05 15:18:18,116 INFO [train.py:862] Epoch 7, batch 17500, loss[loss=0.2261, simple_loss=0.45, pruned_loss=0.1224, ctc_loss=0.1741, over 14706.00 frames. ], tot_loss[loss=0.2043, simple_loss=0.4515, pruned_loss=0.1036, ctc_loss=0.1507, over 2865313.10 frames. ], batch size: 39, lr: 3.00e-03, 
2023-01-05 15:19:22,085 INFO [train.py:862] Epoch 7, batch 18000, loss[loss=0.1839, simple_loss=0.4239, pruned_loss=0.09209, ctc_loss=0.1325, over 14727.00 frames. ], tot_loss[loss=0.2017, simple_loss=0.4511, pruned_loss=0.102, ctc_loss=0.1478, over 2842703.59 frames. ], batch size: 37, lr: 3.00e-03, 
2023-01-05 15:19:22,085 INFO [train.py:887] Computing validation loss
2023-01-05 15:19:47,564 INFO [train.py:897] Epoch 7, validation: loss=0.2151, simple_loss=0.4706, pruned_loss=0.1108, ctc_loss=0.159, over 944034.00 frames. 
2023-01-05 15:19:47,565 INFO [train.py:898] Maximum memory allocated so far is 7283MB
2023-01-05 15:20:14,521 INFO [checkpoint.py:75] Saving checkpoint to tiny_transducer_ctc/exp_2m_phone_halfdelay/epoch-7.pt
2023-01-05 15:20:17,200 INFO [train.py:862] Epoch 8, batch 0, loss[loss=0.217, simple_loss=0.4638, pruned_loss=0.1096, ctc_loss=0.1637, over 14687.00 frames. ], tot_loss[loss=0.217, simple_loss=0.4638, pruned_loss=0.1096, ctc_loss=0.1637, over 14687.00 frames. ], batch size: 44, lr: 3.00e-03, 
2023-01-05 15:21:22,605 INFO [train.py:862] Epoch 8, batch 500, loss[loss=0.1784, simple_loss=0.4304, pruned_loss=0.09, ctc_loss=0.1241, over 14865.00 frames. ], tot_loss[loss=0.2052, simple_loss=0.4533, pruned_loss=0.1042, ctc_loss=0.1513, over 2637777.48 frames. ], batch size: 40, lr: 3.00e-03, 
2023-01-05 15:22:27,597 INFO [train.py:862] Epoch 8, batch 1000, loss[loss=0.1957, simple_loss=0.439, pruned_loss=0.09446, ctc_loss=0.145, over 14695.00 frames. ], tot_loss[loss=0.2053, simple_loss=0.4547, pruned_loss=0.104, ctc_loss=0.1513, over 2842466.42 frames. ], batch size: 39, lr: 3.00e-03, 
2023-01-05 15:23:32,642 INFO [train.py:862] Epoch 8, batch 1500, loss[loss=0.2319, simple_loss=0.4458, pruned_loss=0.1283, ctc_loss=0.1808, over 14792.00 frames. ], tot_loss[loss=0.202, simple_loss=0.4521, pruned_loss=0.1018, ctc_loss=0.148, over 2876730.55 frames. ], batch size: 36, lr: 3.00e-03, 
2023-01-05 15:24:38,106 INFO [train.py:862] Epoch 8, batch 2000, loss[loss=0.2375, simple_loss=0.4685, pruned_loss=0.1234, ctc_loss=0.186, over 14816.00 frames. ], tot_loss[loss=0.2042, simple_loss=0.4516, pruned_loss=0.1037, ctc_loss=0.1505, over 2867614.26 frames. ], batch size: 41, lr: 3.00e-03, 
2023-01-05 15:25:41,014 INFO [checkpoint.py:75] Saving checkpoint to tiny_transducer_ctc/exp_2m_phone_halfdelay/checkpoint-130000.pt
2023-01-05 15:25:43,167 INFO [train.py:862] Epoch 8, batch 2500, loss[loss=0.1978, simple_loss=0.4731, pruned_loss=0.09544, ctc_loss=0.1403, over 14694.00 frames. ], tot_loss[loss=0.2026, simple_loss=0.4503, pruned_loss=0.1027, ctc_loss=0.149, over 2880879.10 frames. ], batch size: 44, lr: 3.00e-03, 
2023-01-05 15:26:49,105 INFO [train.py:862] Epoch 8, batch 3000, loss[loss=0.178, simple_loss=0.4278, pruned_loss=0.08976, ctc_loss=0.1241, over 14701.00 frames. ], tot_loss[loss=0.2032, simple_loss=0.4514, pruned_loss=0.1032, ctc_loss=0.1494, over 2873886.16 frames. ], batch size: 38, lr: 3.00e-03, 
2023-01-05 15:27:54,452 INFO [train.py:862] Epoch 8, batch 3500, loss[loss=0.1993, simple_loss=0.471, pruned_loss=0.1006, ctc_loss=0.1406, over 14544.00 frames. ], tot_loss[loss=0.203, simple_loss=0.4536, pruned_loss=0.1024, ctc_loss=0.1489, over 2867951.27 frames. ], batch size: 48, lr: 3.00e-03, 
2023-01-05 15:28:59,442 INFO [train.py:862] Epoch 8, batch 4000, loss[loss=0.1541, simple_loss=0.4145, pruned_loss=0.07191, ctc_loss=0.1005, over 14529.00 frames. ], tot_loss[loss=0.203, simple_loss=0.4526, pruned_loss=0.1026, ctc_loss=0.1491, over 2867822.97 frames. ], batch size: 36, lr: 3.00e-03, 
2023-01-05 15:30:05,672 INFO [train.py:862] Epoch 8, batch 4500, loss[loss=0.2111, simple_loss=0.4601, pruned_loss=0.1052, ctc_loss=0.158, over 12781.00 frames. ], tot_loss[loss=0.2027, simple_loss=0.4514, pruned_loss=0.1025, ctc_loss=0.149, over 2862384.98 frames. ], batch size: 76, lr: 3.00e-03, 
2023-01-05 15:31:10,804 INFO [train.py:862] Epoch 8, batch 5000, loss[loss=0.195, simple_loss=0.4647, pruned_loss=0.098, ctc_loss=0.137, over 13874.00 frames. ], tot_loss[loss=0.2034, simple_loss=0.4535, pruned_loss=0.1028, ctc_loss=0.1494, over 2846956.74 frames. ], batch size: 63, lr: 3.00e-03, 
2023-01-05 15:32:16,884 INFO [train.py:862] Epoch 8, batch 5500, loss[loss=0.2131, simple_loss=0.4341, pruned_loss=0.1082, ctc_loss=0.165, over 14710.00 frames. ], tot_loss[loss=0.2026, simple_loss=0.4522, pruned_loss=0.1023, ctc_loss=0.1487, over 2878412.25 frames. ], batch size: 37, lr: 3.00e-03, 
2023-01-05 15:33:22,157 INFO [train.py:862] Epoch 8, batch 6000, loss[loss=0.1919, simple_loss=0.4239, pruned_loss=0.09405, ctc_loss=0.143, over 14691.00 frames. ], tot_loss[loss=0.2026, simple_loss=0.4507, pruned_loss=0.1025, ctc_loss=0.1489, over 2880593.09 frames. ], batch size: 39, lr: 3.00e-03, 
2023-01-05 15:34:27,191 INFO [train.py:862] Epoch 8, batch 6500, loss[loss=0.2384, simple_loss=0.4829, pruned_loss=0.1304, ctc_loss=0.1812, over 14736.00 frames. ], tot_loss[loss=0.2044, simple_loss=0.4526, pruned_loss=0.1039, ctc_loss=0.1505, over 2853183.53 frames. ], batch size: 46, lr: 3.00e-03, 
2023-01-05 15:35:32,441 INFO [train.py:862] Epoch 8, batch 7000, loss[loss=0.2149, simple_loss=0.4959, pruned_loss=0.1072, ctc_loss=0.1547, over 14686.00 frames. ], tot_loss[loss=0.202, simple_loss=0.4508, pruned_loss=0.1018, ctc_loss=0.1483, over 2878946.51 frames. ], batch size: 47, lr: 3.00e-03, 
2023-01-05 15:36:37,509 INFO [train.py:862] Epoch 8, batch 7500, loss[loss=0.2092, simple_loss=0.4526, pruned_loss=0.1094, ctc_loss=0.155, over 14840.00 frames. ], tot_loss[loss=0.2021, simple_loss=0.4513, pruned_loss=0.102, ctc_loss=0.1483, over 2860782.74 frames. ], batch size: 42, lr: 3.00e-03, 
2023-01-05 15:37:42,546 INFO [train.py:862] Epoch 8, batch 8000, loss[loss=0.2047, simple_loss=0.4667, pruned_loss=0.1053, ctc_loss=0.1473, over 14590.00 frames. ], tot_loss[loss=0.2017, simple_loss=0.4506, pruned_loss=0.1018, ctc_loss=0.148, over 2876501.03 frames. ], batch size: 43, lr: 3.00e-03, 
2023-01-05 15:38:48,247 INFO [train.py:862] Epoch 8, batch 8500, loss[loss=0.1744, simple_loss=0.4158, pruned_loss=0.08421, ctc_loss=0.124, over 14530.00 frames. ], tot_loss[loss=0.2009, simple_loss=0.4505, pruned_loss=0.1013, ctc_loss=0.147, over 2871467.24 frames. ], batch size: 36, lr: 3.00e-03, 
2023-01-05 15:39:52,828 INFO [train.py:862] Epoch 8, batch 9000, loss[loss=0.2257, simple_loss=0.4762, pruned_loss=0.117, ctc_loss=0.1703, over 13016.00 frames. ], tot_loss[loss=0.2011, simple_loss=0.4502, pruned_loss=0.1012, ctc_loss=0.1474, over 2864825.89 frames. ], batch size: 76, lr: 3.00e-03, 
2023-01-05 15:39:52,829 INFO [train.py:887] Computing validation loss
2023-01-05 15:40:18,655 INFO [train.py:897] Epoch 8, validation: loss=0.2118, simple_loss=0.4682, pruned_loss=0.1072, ctc_loss=0.1563, over 944034.00 frames. 
2023-01-05 15:40:18,655 INFO [train.py:898] Maximum memory allocated so far is 7283MB
2023-01-05 15:41:23,919 INFO [train.py:862] Epoch 8, batch 9500, loss[loss=0.1703, simple_loss=0.3949, pruned_loss=0.08364, ctc_loss=0.1228, over 14383.00 frames. ], tot_loss[loss=0.2032, simple_loss=0.4525, pruned_loss=0.1028, ctc_loss=0.1492, over 2846730.19 frames. ], batch size: 33, lr: 3.00e-03, 
2023-01-05 15:42:28,975 INFO [train.py:862] Epoch 8, batch 10000, loss[loss=0.2497, simple_loss=0.4907, pruned_loss=0.1328, ctc_loss=0.1947, over 14843.00 frames. ], tot_loss[loss=0.2022, simple_loss=0.4512, pruned_loss=0.1018, ctc_loss=0.1485, over 2861953.62 frames. ], batch size: 42, lr: 3.00e-03, 
2023-01-05 15:43:35,117 INFO [train.py:862] Epoch 8, batch 10500, loss[loss=0.2181, simple_loss=0.4516, pruned_loss=0.1141, ctc_loss=0.1659, over 14667.00 frames. ], tot_loss[loss=0.2025, simple_loss=0.4515, pruned_loss=0.1019, ctc_loss=0.1489, over 2858887.30 frames. ], batch size: 40, lr: 3.00e-03, 
2023-01-05 15:44:39,933 INFO [train.py:862] Epoch 8, batch 11000, loss[loss=0.247, simple_loss=0.4938, pruned_loss=0.1355, ctc_loss=0.189, over 9979.00 frames. ], tot_loss[loss=0.2004, simple_loss=0.4502, pruned_loss=0.1006, ctc_loss=0.1467, over 2841297.29 frames. ], batch size: 104, lr: 3.00e-03, 
2023-01-05 15:45:45,613 INFO [train.py:862] Epoch 8, batch 11500, loss[loss=0.1919, simple_loss=0.4383, pruned_loss=0.09468, ctc_loss=0.1396, over 14643.00 frames. ], tot_loss[loss=0.2024, simple_loss=0.452, pruned_loss=0.1022, ctc_loss=0.1485, over 2841865.91 frames. ], batch size: 41, lr: 3.00e-03, 
2023-01-05 15:46:50,853 INFO [train.py:862] Epoch 8, batch 12000, loss[loss=0.1697, simple_loss=0.4052, pruned_loss=0.0809, ctc_loss=0.121, over 14378.00 frames. ], tot_loss[loss=0.199, simple_loss=0.4487, pruned_loss=0.09972, ctc_loss=0.1454, over 2877760.83 frames. ], batch size: 33, lr: 3.00e-03, 
2023-01-05 15:47:53,576 INFO [checkpoint.py:75] Saving checkpoint to tiny_transducer_ctc/exp_2m_phone_halfdelay/checkpoint-140000.pt
2023-01-05 15:47:55,943 INFO [train.py:862] Epoch 8, batch 12500, loss[loss=0.1641, simple_loss=0.4232, pruned_loss=0.07998, ctc_loss=0.1095, over 14715.00 frames. ], tot_loss[loss=0.202, simple_loss=0.4514, pruned_loss=0.1018, ctc_loss=0.1482, over 2865008.34 frames. ], batch size: 37, lr: 3.00e-03, 
2023-01-05 15:49:01,448 INFO [train.py:862] Epoch 8, batch 13000, loss[loss=0.208, simple_loss=0.4804, pruned_loss=0.1018, ctc_loss=0.1506, over 14510.00 frames. ], tot_loss[loss=0.2011, simple_loss=0.4514, pruned_loss=0.1008, ctc_loss=0.1474, over 2853277.08 frames. ], batch size: 50, lr: 3.00e-03, 
2023-01-05 15:50:06,931 INFO [train.py:862] Epoch 8, batch 13500, loss[loss=0.228, simple_loss=0.4782, pruned_loss=0.1151, ctc_loss=0.1738, over 13019.00 frames. ], tot_loss[loss=0.2019, simple_loss=0.452, pruned_loss=0.1012, ctc_loss=0.1482, over 2863762.07 frames. ], batch size: 76, lr: 3.00e-03, 
2023-01-05 15:51:12,520 INFO [train.py:862] Epoch 8, batch 14000, loss[loss=0.203, simple_loss=0.4302, pruned_loss=0.1037, ctc_loss=0.1534, over 13582.00 frames. ], tot_loss[loss=0.2014, simple_loss=0.4514, pruned_loss=0.101, ctc_loss=0.1478, over 2856221.83 frames. ], batch size: 30, lr: 3.00e-03, 
2023-01-05 15:52:17,850 INFO [train.py:862] Epoch 8, batch 14500, loss[loss=0.1983, simple_loss=0.4705, pruned_loss=0.09859, ctc_loss=0.1402, over 14801.00 frames. ], tot_loss[loss=0.2018, simple_loss=0.4503, pruned_loss=0.1017, ctc_loss=0.1482, over 2864765.47 frames. ], batch size: 43, lr: 3.00e-03, 
2023-01-05 15:53:23,626 INFO [train.py:862] Epoch 8, batch 15000, loss[loss=0.1872, simple_loss=0.4606, pruned_loss=0.0867, ctc_loss=0.1316, over 14536.00 frames. ], tot_loss[loss=0.2002, simple_loss=0.4499, pruned_loss=0.1002, ctc_loss=0.1466, over 2855784.14 frames. ], batch size: 48, lr: 3.00e-03, 
2023-01-05 15:54:29,059 INFO [train.py:862] Epoch 8, batch 15500, loss[loss=0.2383, simple_loss=0.4802, pruned_loss=0.121, ctc_loss=0.1858, over 12931.00 frames. ], tot_loss[loss=0.2014, simple_loss=0.4494, pruned_loss=0.1014, ctc_loss=0.1479, over 2859359.45 frames. ], batch size: 76, lr: 3.00e-03, 
2023-01-05 15:55:34,641 INFO [train.py:862] Epoch 8, batch 16000, loss[loss=0.2224, simple_loss=0.4686, pruned_loss=0.1156, ctc_loss=0.1678, over 14872.00 frames. ], tot_loss[loss=0.2001, simple_loss=0.4477, pruned_loss=0.1005, ctc_loss=0.1469, over 2850784.88 frames. ], batch size: 40, lr: 3.00e-03, 
2023-01-05 15:56:40,652 INFO [train.py:862] Epoch 8, batch 16500, loss[loss=0.2631, simple_loss=0.4542, pruned_loss=0.1473, ctc_loss=0.2154, over 14648.00 frames. ], tot_loss[loss=0.1973, simple_loss=0.4472, pruned_loss=0.09859, ctc_loss=0.1438, over 2867590.50 frames. ], batch size: 35, lr: 3.00e-03, 
2023-01-05 15:57:45,891 INFO [train.py:862] Epoch 8, batch 17000, loss[loss=0.2317, simple_loss=0.4581, pruned_loss=0.1251, ctc_loss=0.1792, over 14706.00 frames. ], tot_loss[loss=0.1994, simple_loss=0.4465, pruned_loss=0.1004, ctc_loss=0.1461, over 2865809.15 frames. ], batch size: 37, lr: 3.00e-03, 
2023-01-05 15:58:51,718 INFO [train.py:862] Epoch 8, batch 17500, loss[loss=0.2047, simple_loss=0.4503, pruned_loss=0.1074, ctc_loss=0.1499, over 14560.00 frames. ], tot_loss[loss=0.2002, simple_loss=0.4467, pruned_loss=0.101, ctc_loss=0.1471, over 2869683.66 frames. ], batch size: 53, lr: 3.00e-03, 
2023-01-05 15:59:55,485 INFO [train.py:862] Epoch 8, batch 18000, loss[loss=0.2156, simple_loss=0.4601, pruned_loss=0.1093, ctc_loss=0.1625, over 14560.00 frames. ], tot_loss[loss=0.2003, simple_loss=0.447, pruned_loss=0.1008, ctc_loss=0.1471, over 2865553.35 frames. ], batch size: 48, lr: 3.00e-03, 
2023-01-05 15:59:55,485 INFO [train.py:887] Computing validation loss
2023-01-05 16:00:20,959 INFO [train.py:897] Epoch 8, validation: loss=0.2101, simple_loss=0.4601, pruned_loss=0.1063, ctc_loss=0.1559, over 944034.00 frames. 
2023-01-05 16:00:20,959 INFO [train.py:898] Maximum memory allocated so far is 7283MB
2023-01-05 16:00:47,834 INFO [checkpoint.py:75] Saving checkpoint to tiny_transducer_ctc/exp_2m_phone_halfdelay/epoch-8.pt
2023-01-05 16:00:51,110 INFO [train.py:862] Epoch 9, batch 0, loss[loss=0.2675, simple_loss=0.4966, pruned_loss=0.1482, ctc_loss=0.2122, over 14700.00 frames. ], tot_loss[loss=0.2675, simple_loss=0.4966, pruned_loss=0.1482, ctc_loss=0.2122, over 14700.00 frames. ], batch size: 47, lr: 3.00e-03, 
2023-01-05 16:01:56,000 INFO [train.py:862] Epoch 9, batch 500, loss[loss=0.1832, simple_loss=0.4527, pruned_loss=0.08658, ctc_loss=0.1276, over 14645.00 frames. ], tot_loss[loss=0.2001, simple_loss=0.4476, pruned_loss=0.1004, ctc_loss=0.1469, over 2653929.75 frames. ], batch size: 41, lr: 3.00e-03, 
2023-01-05 16:03:00,945 INFO [train.py:862] Epoch 9, batch 1000, loss[loss=0.2074, simple_loss=0.4596, pruned_loss=0.09984, ctc_loss=0.1551, over 13700.00 frames. ], tot_loss[loss=0.1981, simple_loss=0.4454, pruned_loss=0.09922, ctc_loss=0.145, over 2859400.29 frames. ], batch size: 63, lr: 3.00e-03, 
2023-01-05 16:04:05,695 INFO [train.py:862] Epoch 9, batch 1500, loss[loss=0.2193, simple_loss=0.4425, pruned_loss=0.1173, ctc_loss=0.1682, over 14502.00 frames. ], tot_loss[loss=0.2002, simple_loss=0.4468, pruned_loss=0.1006, ctc_loss=0.1471, over 2861103.63 frames. ], batch size: 38, lr: 3.00e-03, 
2023-01-05 16:05:11,147 INFO [train.py:862] Epoch 9, batch 2000, loss[loss=0.1974, simple_loss=0.4643, pruned_loss=0.09384, ctc_loss=0.1422, over 14663.00 frames. ], tot_loss[loss=0.1973, simple_loss=0.4463, pruned_loss=0.09858, ctc_loss=0.144, over 2876006.41 frames. ], batch size: 40, lr: 3.00e-03, 
2023-01-05 16:06:16,087 INFO [train.py:862] Epoch 9, batch 2500, loss[loss=0.2151, simple_loss=0.4671, pruned_loss=0.1117, ctc_loss=0.1593, over 13625.00 frames. ], tot_loss[loss=0.2011, simple_loss=0.4483, pruned_loss=0.1013, ctc_loss=0.1478, over 2861961.73 frames. ], batch size: 63, lr: 3.00e-03, 
2023-01-05 16:07:21,457 INFO [train.py:862] Epoch 9, batch 3000, loss[loss=0.2233, simple_loss=0.4733, pruned_loss=0.1176, ctc_loss=0.1672, over 14853.00 frames. ], tot_loss[loss=0.1988, simple_loss=0.448, pruned_loss=0.09951, ctc_loss=0.1453, over 2866230.23 frames. ], batch size: 40, lr: 3.00e-03, 
2023-01-05 16:08:26,579 INFO [train.py:862] Epoch 9, batch 3500, loss[loss=0.1873, simple_loss=0.4488, pruned_loss=0.09172, ctc_loss=0.132, over 14477.00 frames. ], tot_loss[loss=0.1971, simple_loss=0.4458, pruned_loss=0.0984, ctc_loss=0.1439, over 2883066.17 frames. ], batch size: 50, lr: 3.00e-03, 
2023-01-05 16:09:31,476 INFO [train.py:862] Epoch 9, batch 4000, loss[loss=0.1912, simple_loss=0.4485, pruned_loss=0.09306, ctc_loss=0.1371, over 14581.00 frames. ], tot_loss[loss=0.1997, simple_loss=0.447, pruned_loss=0.1003, ctc_loss=0.1466, over 2861356.72 frames. ], batch size: 43, lr: 3.00e-03, 
2023-01-05 16:10:07,029 INFO [checkpoint.py:75] Saving checkpoint to tiny_transducer_ctc/exp_2m_phone_halfdelay/checkpoint-150000.pt
2023-01-05 16:10:37,116 INFO [train.py:862] Epoch 9, batch 4500, loss[loss=0.2009, simple_loss=0.4515, pruned_loss=0.1017, ctc_loss=0.1467, over 14685.00 frames. ], tot_loss[loss=0.199, simple_loss=0.4458, pruned_loss=0.1001, ctc_loss=0.1458, over 2859383.38 frames. ], batch size: 44, lr: 3.00e-03, 
2023-01-05 16:11:42,149 INFO [train.py:862] Epoch 9, batch 5000, loss[loss=0.2046, simple_loss=0.4549, pruned_loss=0.1037, ctc_loss=0.1503, over 14649.00 frames. ], tot_loss[loss=0.1995, simple_loss=0.4461, pruned_loss=0.1002, ctc_loss=0.1464, over 2850509.30 frames. ], batch size: 40, lr: 3.00e-03, 
2023-01-05 16:12:47,499 INFO [train.py:862] Epoch 9, batch 5500, loss[loss=0.1872, simple_loss=0.4336, pruned_loss=0.08737, ctc_loss=0.1371, over 14650.00 frames. ], tot_loss[loss=0.2015, simple_loss=0.4479, pruned_loss=0.1016, ctc_loss=0.1483, over 2860274.09 frames. ], batch size: 41, lr: 3.00e-03, 
2023-01-05 16:13:52,769 INFO [train.py:862] Epoch 9, batch 6000, loss[loss=0.1973, simple_loss=0.4122, pruned_loss=0.102, ctc_loss=0.1499, over 14536.00 frames. ], tot_loss[loss=0.198, simple_loss=0.4459, pruned_loss=0.09899, ctc_loss=0.1449, over 2874272.08 frames. ], batch size: 34, lr: 3.00e-03, 
2023-01-05 16:14:58,397 INFO [train.py:862] Epoch 9, batch 6500, loss[loss=0.1793, simple_loss=0.4328, pruned_loss=0.08752, ctc_loss=0.1259, over 14512.00 frames. ], tot_loss[loss=0.1979, simple_loss=0.4462, pruned_loss=0.09918, ctc_loss=0.1446, over 2853418.47 frames. ], batch size: 38, lr: 3.00e-03, 
2023-01-05 16:16:03,474 INFO [train.py:862] Epoch 9, batch 7000, loss[loss=0.2115, simple_loss=0.463, pruned_loss=0.113, ctc_loss=0.1546, over 13691.00 frames. ], tot_loss[loss=0.1982, simple_loss=0.4454, pruned_loss=0.09962, ctc_loss=0.145, over 2869626.08 frames. ], batch size: 63, lr: 3.00e-03, 
2023-01-05 16:17:08,668 INFO [train.py:862] Epoch 9, batch 7500, loss[loss=0.1788, simple_loss=0.4295, pruned_loss=0.09002, ctc_loss=0.1248, over 14515.00 frames. ], tot_loss[loss=0.197, simple_loss=0.4454, pruned_loss=0.09842, ctc_loss=0.1438, over 2866444.54 frames. ], batch size: 38, lr: 3.00e-03, 
2023-01-05 16:18:14,325 INFO [train.py:862] Epoch 9, batch 8000, loss[loss=0.1619, simple_loss=0.3929, pruned_loss=0.07861, ctc_loss=0.1134, over 14450.00 frames. ], tot_loss[loss=0.1966, simple_loss=0.4443, pruned_loss=0.09816, ctc_loss=0.1435, over 2866831.75 frames. ], batch size: 32, lr: 3.00e-03, 
2023-01-05 16:19:19,567 INFO [train.py:862] Epoch 9, batch 8500, loss[loss=0.1745, simple_loss=0.4268, pruned_loss=0.0809, ctc_loss=0.1231, over 14518.00 frames. ], tot_loss[loss=0.1978, simple_loss=0.445, pruned_loss=0.09895, ctc_loss=0.1448, over 2864822.65 frames. ], batch size: 37, lr: 3.00e-03, 
2023-01-05 16:20:24,567 INFO [train.py:862] Epoch 9, batch 9000, loss[loss=0.2348, simple_loss=0.4902, pruned_loss=0.1165, ctc_loss=0.1804, over 14702.00 frames. ], tot_loss[loss=0.1965, simple_loss=0.4441, pruned_loss=0.09826, ctc_loss=0.1434, over 2864383.84 frames. ], batch size: 47, lr: 3.00e-03, 
2023-01-05 16:20:24,567 INFO [train.py:887] Computing validation loss
2023-01-05 16:20:50,613 INFO [train.py:897] Epoch 9, validation: loss=0.2074, simple_loss=0.4577, pruned_loss=0.1051, ctc_loss=0.1532, over 944034.00 frames. 
2023-01-05 16:20:50,614 INFO [train.py:898] Maximum memory allocated so far is 7283MB
2023-01-05 16:21:55,859 INFO [train.py:862] Epoch 9, batch 9500, loss[loss=0.1612, simple_loss=0.4083, pruned_loss=0.07338, ctc_loss=0.1114, over 14711.00 frames. ], tot_loss[loss=0.1974, simple_loss=0.4456, pruned_loss=0.09897, ctc_loss=0.1441, over 2862553.35 frames. ], batch size: 38, lr: 3.00e-03, 
2023-01-05 16:23:01,889 INFO [train.py:862] Epoch 9, batch 10000, loss[loss=0.2213, simple_loss=0.4605, pruned_loss=0.1098, ctc_loss=0.1703, over 14549.00 frames. ], tot_loss[loss=0.1976, simple_loss=0.4451, pruned_loss=0.09867, ctc_loss=0.1446, over 2866125.03 frames. ], batch size: 53, lr: 3.00e-03, 
2023-01-05 16:24:07,363 INFO [train.py:862] Epoch 9, batch 10500, loss[loss=0.1791, simple_loss=0.4506, pruned_loss=0.08223, ctc_loss=0.1241, over 14837.00 frames. ], tot_loss[loss=0.1985, simple_loss=0.4466, pruned_loss=0.09926, ctc_loss=0.1453, over 2853294.00 frames. ], batch size: 42, lr: 3.00e-03, 
2023-01-05 16:25:13,148 INFO [train.py:862] Epoch 9, batch 11000, loss[loss=0.1945, simple_loss=0.4461, pruned_loss=0.09692, ctc_loss=0.1407, over 14698.00 frames. ], tot_loss[loss=0.1956, simple_loss=0.4437, pruned_loss=0.09763, ctc_loss=0.1425, over 2863799.40 frames. ], batch size: 44, lr: 3.00e-03, 
2023-01-05 16:26:19,186 INFO [train.py:862] Epoch 9, batch 11500, loss[loss=0.1873, simple_loss=0.4489, pruned_loss=0.09042, ctc_loss=0.1326, over 14722.00 frames. ], tot_loss[loss=0.197, simple_loss=0.4456, pruned_loss=0.09857, ctc_loss=0.1437, over 2864832.98 frames. ], batch size: 46, lr: 3.00e-03, 
2023-01-05 16:27:24,486 INFO [train.py:862] Epoch 9, batch 12000, loss[loss=0.2818, simple_loss=0.5096, pruned_loss=0.1568, ctc_loss=0.2261, over 14466.00 frames. ], tot_loss[loss=0.1964, simple_loss=0.4435, pruned_loss=0.09795, ctc_loss=0.1435, over 2854025.45 frames. ], batch size: 50, lr: 3.00e-03, 
2023-01-05 16:28:30,556 INFO [train.py:862] Epoch 9, batch 12500, loss[loss=0.2115, simple_loss=0.4578, pruned_loss=0.1041, ctc_loss=0.1595, over 14873.00 frames. ], tot_loss[loss=0.1988, simple_loss=0.4467, pruned_loss=0.09941, ctc_loss=0.1456, over 2856906.05 frames. ], batch size: 40, lr: 3.00e-03, 
2023-01-05 16:29:35,959 INFO [train.py:862] Epoch 9, batch 13000, loss[loss=0.2148, simple_loss=0.4618, pruned_loss=0.1072, ctc_loss=0.1619, over 13434.00 frames. ], tot_loss[loss=0.1964, simple_loss=0.445, pruned_loss=0.09789, ctc_loss=0.1433, over 2865790.49 frames. ], batch size: 63, lr: 3.00e-03, 
2023-01-05 16:30:42,215 INFO [train.py:862] Epoch 9, batch 13500, loss[loss=0.2876, simple_loss=0.5135, pruned_loss=0.159, ctc_loss=0.2326, over 14503.00 frames. ], tot_loss[loss=0.195, simple_loss=0.4441, pruned_loss=0.09686, ctc_loss=0.142, over 2879431.65 frames. ], batch size: 48, lr: 3.00e-03, 
2023-01-05 16:31:47,505 INFO [train.py:862] Epoch 9, batch 14000, loss[loss=0.2037, simple_loss=0.4384, pruned_loss=0.1072, ctc_loss=0.1511, over 14696.00 frames. ], tot_loss[loss=0.1981, simple_loss=0.4452, pruned_loss=0.09919, ctc_loss=0.1451, over 2860592.09 frames. ], batch size: 39, lr: 3.00e-03, 
2023-01-05 16:32:22,676 INFO [checkpoint.py:75] Saving checkpoint to tiny_transducer_ctc/exp_2m_phone_halfdelay/checkpoint-160000.pt
2023-01-05 16:32:53,187 INFO [train.py:862] Epoch 9, batch 14500, loss[loss=0.1985, simple_loss=0.4638, pruned_loss=0.09871, ctc_loss=0.1419, over 14529.00 frames. ], tot_loss[loss=0.198, simple_loss=0.445, pruned_loss=0.09933, ctc_loss=0.145, over 2858437.48 frames. ], batch size: 50, lr: 3.00e-03, 
2023-01-05 16:33:58,914 INFO [train.py:862] Epoch 9, batch 15000, loss[loss=0.1705, simple_loss=0.3905, pruned_loss=0.08584, ctc_loss=0.1231, over 14022.00 frames. ], tot_loss[loss=0.1974, simple_loss=0.4451, pruned_loss=0.09853, ctc_loss=0.1444, over 2851897.66 frames. ], batch size: 31, lr: 3.00e-03, 
2023-01-05 16:35:04,355 INFO [train.py:862] Epoch 9, batch 15500, loss[loss=0.2151, simple_loss=0.4651, pruned_loss=0.113, ctc_loss=0.1592, over 14798.00 frames. ], tot_loss[loss=0.1984, simple_loss=0.446, pruned_loss=0.09911, ctc_loss=0.1454, over 2848096.31 frames. ], batch size: 43, lr: 3.00e-03, 
2023-01-05 16:36:10,236 INFO [train.py:862] Epoch 9, batch 16000, loss[loss=0.2287, simple_loss=0.4795, pruned_loss=0.117, ctc_loss=0.1738, over 14229.00 frames. ], tot_loss[loss=0.1968, simple_loss=0.4439, pruned_loss=0.09841, ctc_loss=0.1438, over 2848391.41 frames. ], batch size: 56, lr: 3.00e-03, 
2023-01-05 16:37:15,774 INFO [train.py:862] Epoch 9, batch 16500, loss[loss=0.2021, simple_loss=0.4672, pruned_loss=0.09673, ctc_loss=0.1472, over 14786.00 frames. ], tot_loss[loss=0.1943, simple_loss=0.4426, pruned_loss=0.0967, ctc_loss=0.1413, over 2860564.40 frames. ], batch size: 43, lr: 3.00e-03, 
2023-01-05 16:38:21,470 INFO [train.py:862] Epoch 9, batch 17000, loss[loss=0.1815, simple_loss=0.4228, pruned_loss=0.08842, ctc_loss=0.1308, over 14837.00 frames. ], tot_loss[loss=0.1952, simple_loss=0.4428, pruned_loss=0.09723, ctc_loss=0.1423, over 2857501.82 frames. ], batch size: 42, lr: 3.00e-03, 
2023-01-05 16:39:27,094 INFO [train.py:862] Epoch 9, batch 17500, loss[loss=0.1437, simple_loss=0.4003, pruned_loss=0.06398, ctc_loss=0.09214, over 14515.00 frames. ], tot_loss[loss=0.1968, simple_loss=0.4446, pruned_loss=0.09825, ctc_loss=0.1438, over 2872640.86 frames. ], batch size: 36, lr: 3.00e-03, 
2023-01-05 16:40:30,978 INFO [train.py:862] Epoch 9, batch 18000, loss[loss=0.2194, simple_loss=0.4528, pruned_loss=0.1118, ctc_loss=0.1684, over 10057.00 frames. ], tot_loss[loss=0.1973, simple_loss=0.4451, pruned_loss=0.09851, ctc_loss=0.1443, over 2863406.02 frames. ], batch size: 103, lr: 3.00e-03, 
2023-01-05 16:40:30,979 INFO [train.py:887] Computing validation loss
2023-01-05 16:40:57,006 INFO [train.py:897] Epoch 9, validation: loss=0.2097, simple_loss=0.4589, pruned_loss=0.1066, ctc_loss=0.1556, over 944034.00 frames. 
2023-01-05 16:40:57,007 INFO [train.py:898] Maximum memory allocated so far is 7283MB
2023-01-05 16:41:23,812 INFO [checkpoint.py:75] Saving checkpoint to tiny_transducer_ctc/exp_2m_phone_halfdelay/epoch-9.pt
2023-01-05 16:41:26,547 INFO [train.py:862] Epoch 10, batch 0, loss[loss=0.2402, simple_loss=0.4681, pruned_loss=0.1348, ctc_loss=0.1851, over 14833.00 frames. ], tot_loss[loss=0.2402, simple_loss=0.4681, pruned_loss=0.1348, ctc_loss=0.1851, over 14833.00 frames. ], batch size: 42, lr: 3.00e-03, 
2023-01-05 16:42:31,382 INFO [train.py:862] Epoch 10, batch 500, loss[loss=0.1947, simple_loss=0.4277, pruned_loss=0.08989, ctc_loss=0.148, over 14700.00 frames. ], tot_loss[loss=0.1989, simple_loss=0.447, pruned_loss=0.09969, ctc_loss=0.1456, over 2618744.20 frames. ], batch size: 39, lr: 3.00e-03, 
2023-01-05 16:43:36,424 INFO [train.py:862] Epoch 10, batch 1000, loss[loss=0.1856, simple_loss=0.4471, pruned_loss=0.09112, ctc_loss=0.1303, over 14837.00 frames. ], tot_loss[loss=0.1968, simple_loss=0.4445, pruned_loss=0.09817, ctc_loss=0.1438, over 2854320.74 frames. ], batch size: 42, lr: 3.00e-03, 
2023-01-05 16:44:41,095 INFO [train.py:862] Epoch 10, batch 1500, loss[loss=0.2246, simple_loss=0.4922, pruned_loss=0.1175, ctc_loss=0.1651, over 14686.00 frames. ], tot_loss[loss=0.2015, simple_loss=0.4489, pruned_loss=0.1014, ctc_loss=0.1482, over 2861106.02 frames. ], batch size: 42, lr: 3.00e-03, 
2023-01-05 16:45:46,423 INFO [train.py:862] Epoch 10, batch 2000, loss[loss=0.2479, simple_loss=0.4945, pruned_loss=0.1318, ctc_loss=0.1916, over 14582.00 frames. ], tot_loss[loss=0.1983, simple_loss=0.4455, pruned_loss=0.09934, ctc_loss=0.1452, over 2868919.32 frames. ], batch size: 43, lr: 3.00e-03, 
2023-01-05 16:46:51,690 INFO [train.py:862] Epoch 10, batch 2500, loss[loss=0.2471, simple_loss=0.495, pruned_loss=0.1321, ctc_loss=0.1904, over 14545.00 frames. ], tot_loss[loss=0.197, simple_loss=0.4442, pruned_loss=0.09878, ctc_loss=0.1438, over 2884264.40 frames. ], batch size: 53, lr: 3.00e-03, 
2023-01-05 16:47:56,328 INFO [train.py:862] Epoch 10, batch 3000, loss[loss=0.2193, simple_loss=0.484, pruned_loss=0.1083, ctc_loss=0.1632, over 14669.00 frames. ], tot_loss[loss=0.1938, simple_loss=0.4415, pruned_loss=0.0965, ctc_loss=0.1409, over 2875631.90 frames. ], batch size: 44, lr: 3.00e-03, 
2023-01-05 16:49:01,793 INFO [train.py:862] Epoch 10, batch 3500, loss[loss=0.1783, simple_loss=0.4016, pruned_loss=0.08696, ctc_loss=0.1314, over 14122.00 frames. ], tot_loss[loss=0.1969, simple_loss=0.4452, pruned_loss=0.09815, ctc_loss=0.1438, over 2876250.05 frames. ], batch size: 31, lr: 3.00e-03, 
2023-01-05 16:50:06,698 INFO [train.py:862] Epoch 10, batch 4000, loss[loss=0.2021, simple_loss=0.4697, pruned_loss=0.1005, ctc_loss=0.145, over 14710.00 frames. ], tot_loss[loss=0.1973, simple_loss=0.4452, pruned_loss=0.09829, ctc_loss=0.1443, over 2868352.25 frames. ], batch size: 46, lr: 3.00e-03, 
2023-01-05 16:51:11,509 INFO [train.py:862] Epoch 10, batch 4500, loss[loss=0.1888, simple_loss=0.4402, pruned_loss=0.09437, ctc_loss=0.135, over 14659.00 frames. ], tot_loss[loss=0.1951, simple_loss=0.4422, pruned_loss=0.09744, ctc_loss=0.1422, over 2875774.70 frames. ], batch size: 42, lr: 3.00e-03, 
2023-01-05 16:52:16,813 INFO [train.py:862] Epoch 10, batch 5000, loss[loss=0.2266, simple_loss=0.4654, pruned_loss=0.1182, ctc_loss=0.1733, over 14536.00 frames. ], tot_loss[loss=0.1948, simple_loss=0.4421, pruned_loss=0.09698, ctc_loss=0.142, over 2875735.03 frames. ], batch size: 50, lr: 3.00e-03, 
2023-01-05 16:53:21,900 INFO [train.py:862] Epoch 10, batch 5500, loss[loss=0.1982, simple_loss=0.407, pruned_loss=0.1028, ctc_loss=0.1519, over 14516.00 frames. ], tot_loss[loss=0.1958, simple_loss=0.4427, pruned_loss=0.09784, ctc_loss=0.1429, over 2869685.58 frames. ], batch size: 34, lr: 3.00e-03, 
2023-01-05 16:54:27,131 INFO [train.py:862] Epoch 10, batch 6000, loss[loss=0.2023, simple_loss=0.4437, pruned_loss=0.1066, ctc_loss=0.1482, over 10189.00 frames. ], tot_loss[loss=0.1971, simple_loss=0.4464, pruned_loss=0.09835, ctc_loss=0.1438, over 2860496.49 frames. ], batch size: 104, lr: 3.00e-03, 
2023-01-05 16:54:33,783 INFO [checkpoint.py:75] Saving checkpoint to tiny_transducer_ctc/exp_2m_phone_halfdelay/checkpoint-170000.pt
2023-01-05 16:55:32,301 INFO [train.py:862] Epoch 10, batch 6500, loss[loss=0.2328, simple_loss=0.4793, pruned_loss=0.1175, ctc_loss=0.1795, over 14663.00 frames. ], tot_loss[loss=0.195, simple_loss=0.4441, pruned_loss=0.09702, ctc_loss=0.1418, over 2862918.97 frames. ], batch size: 42, lr: 3.00e-03, 
2023-01-05 16:56:37,182 INFO [train.py:862] Epoch 10, batch 7000, loss[loss=0.1551, simple_loss=0.4076, pruned_loss=0.07147, ctc_loss=0.1036, over 14520.00 frames. ], tot_loss[loss=0.1957, simple_loss=0.443, pruned_loss=0.09772, ctc_loss=0.1428, over 2860810.16 frames. ], batch size: 38, lr: 3.00e-03, 
2023-01-05 16:57:42,519 INFO [train.py:862] Epoch 10, batch 7500, loss[loss=0.1571, simple_loss=0.3937, pruned_loss=0.07597, ctc_loss=0.1076, over 14415.00 frames. ], tot_loss[loss=0.1978, simple_loss=0.446, pruned_loss=0.09896, ctc_loss=0.1445, over 2872175.82 frames. ], batch size: 33, lr: 3.00e-03, 
2023-01-05 16:58:47,277 INFO [train.py:862] Epoch 10, batch 8000, loss[loss=0.2018, simple_loss=0.4451, pruned_loss=0.1009, ctc_loss=0.1497, over 14509.00 frames. ], tot_loss[loss=0.195, simple_loss=0.4434, pruned_loss=0.09726, ctc_loss=0.1419, over 2867365.34 frames. ], batch size: 38, lr: 3.00e-03, 
2023-01-05 16:59:52,356 INFO [train.py:862] Epoch 10, batch 8500, loss[loss=0.1709, simple_loss=0.4024, pruned_loss=0.0862, ctc_loss=0.121, over 14646.00 frames. ], tot_loss[loss=0.1957, simple_loss=0.444, pruned_loss=0.09777, ctc_loss=0.1425, over 2869319.53 frames. ], batch size: 35, lr: 3.00e-03, 
2023-01-05 17:00:57,639 INFO [train.py:862] Epoch 10, batch 9000, loss[loss=0.1874, simple_loss=0.4546, pruned_loss=0.09075, ctc_loss=0.1314, over 14536.00 frames. ], tot_loss[loss=0.1956, simple_loss=0.4438, pruned_loss=0.09746, ctc_loss=0.1426, over 2858450.03 frames. ], batch size: 48, lr: 3.00e-03, 
2023-01-05 17:00:57,639 INFO [train.py:887] Computing validation loss
2023-01-05 17:01:23,164 INFO [train.py:897] Epoch 10, validation: loss=0.2068, simple_loss=0.4568, pruned_loss=0.1051, ctc_loss=0.1525, over 944034.00 frames. 
2023-01-05 17:01:23,165 INFO [train.py:898] Maximum memory allocated so far is 7283MB
2023-01-05 17:02:28,949 INFO [train.py:862] Epoch 10, batch 9500, loss[loss=0.2033, simple_loss=0.4771, pruned_loss=0.09975, ctc_loss=0.1455, over 14571.00 frames. ], tot_loss[loss=0.1943, simple_loss=0.4426, pruned_loss=0.09658, ctc_loss=0.1414, over 2851926.46 frames. ], batch size: 53, lr: 3.00e-03, 
2023-01-05 17:03:34,193 INFO [train.py:862] Epoch 10, batch 10000, loss[loss=0.2032, simple_loss=0.4677, pruned_loss=0.09514, ctc_loss=0.1493, over 14677.00 frames. ], tot_loss[loss=0.1951, simple_loss=0.4431, pruned_loss=0.09723, ctc_loss=0.142, over 2857354.29 frames. ], batch size: 44, lr: 3.00e-03, 
2023-01-05 17:04:39,938 INFO [train.py:862] Epoch 10, batch 10500, loss[loss=0.1751, simple_loss=0.4095, pruned_loss=0.08562, ctc_loss=0.1258, over 14794.00 frames. ], tot_loss[loss=0.196, simple_loss=0.4442, pruned_loss=0.09762, ctc_loss=0.1429, over 2864186.07 frames. ], batch size: 36, lr: 3.00e-03, 
2023-01-05 17:05:45,045 INFO [train.py:862] Epoch 10, batch 11000, loss[loss=0.2123, simple_loss=0.4647, pruned_loss=0.1061, ctc_loss=0.1582, over 14829.00 frames. ], tot_loss[loss=0.1959, simple_loss=0.4443, pruned_loss=0.09758, ctc_loss=0.1429, over 2848321.16 frames. ], batch size: 41, lr: 3.00e-03, 
2023-01-05 17:06:50,552 INFO [train.py:862] Epoch 10, batch 11500, loss[loss=0.2127, simple_loss=0.4786, pruned_loss=0.108, ctc_loss=0.1549, over 14520.00 frames. ], tot_loss[loss=0.193, simple_loss=0.4419, pruned_loss=0.09595, ctc_loss=0.1399, over 2857682.68 frames. ], batch size: 50, lr: 3.00e-03, 
2023-01-05 17:07:56,242 INFO [train.py:862] Epoch 10, batch 12000, loss[loss=0.229, simple_loss=0.4809, pruned_loss=0.1202, ctc_loss=0.1725, over 14679.00 frames. ], tot_loss[loss=0.1945, simple_loss=0.4434, pruned_loss=0.09668, ctc_loss=0.1414, over 2863459.42 frames. ], batch size: 47, lr: 3.00e-03, 
2023-01-05 17:09:01,719 INFO [train.py:862] Epoch 10, batch 12500, loss[loss=0.1648, simple_loss=0.4181, pruned_loss=0.07461, ctc_loss=0.1139, over 14861.00 frames. ], tot_loss[loss=0.1967, simple_loss=0.4443, pruned_loss=0.09808, ctc_loss=0.1438, over 2853756.31 frames. ], batch size: 40, lr: 3.00e-03, 
2023-01-05 17:10:07,921 INFO [train.py:862] Epoch 10, batch 13000, loss[loss=0.224, simple_loss=0.4772, pruned_loss=0.1185, ctc_loss=0.167, over 14541.00 frames. ], tot_loss[loss=0.1938, simple_loss=0.4432, pruned_loss=0.09626, ctc_loss=0.1407, over 2879596.69 frames. ], batch size: 48, lr: 3.00e-03, 
2023-01-05 17:11:13,443 INFO [train.py:862] Epoch 10, batch 13500, loss[loss=0.1683, simple_loss=0.4417, pruned_loss=0.07604, ctc_loss=0.1132, over 14581.00 frames. ], tot_loss[loss=0.1933, simple_loss=0.442, pruned_loss=0.09594, ctc_loss=0.1403, over 2874240.90 frames. ], batch size: 43, lr: 3.00e-03, 
2023-01-05 17:12:19,451 INFO [train.py:862] Epoch 10, batch 14000, loss[loss=0.235, simple_loss=0.4664, pruned_loss=0.1215, ctc_loss=0.1837, over 9717.00 frames. ], tot_loss[loss=0.1931, simple_loss=0.4413, pruned_loss=0.09579, ctc_loss=0.1402, over 2872686.19 frames. ], batch size: 104, lr: 3.00e-03, 
2023-01-05 17:13:24,616 INFO [train.py:862] Epoch 10, batch 14500, loss[loss=0.1568, simple_loss=0.3933, pruned_loss=0.07847, ctc_loss=0.1061, over 14539.00 frames. ], tot_loss[loss=0.1933, simple_loss=0.4409, pruned_loss=0.09591, ctc_loss=0.1405, over 2853254.37 frames. ], batch size: 34, lr: 3.00e-03, 
2023-01-05 17:14:30,108 INFO [train.py:862] Epoch 10, batch 15000, loss[loss=0.1806, simple_loss=0.4026, pruned_loss=0.09006, ctc_loss=0.1332, over 14517.00 frames. ], tot_loss[loss=0.1953, simple_loss=0.4429, pruned_loss=0.09741, ctc_loss=0.1424, over 2855079.32 frames. ], batch size: 36, lr: 3.00e-03, 
2023-01-05 17:15:35,866 INFO [train.py:862] Epoch 10, batch 15500, loss[loss=0.1442, simple_loss=0.3763, pruned_loss=0.06783, ctc_loss=0.09632, over 14404.00 frames. ], tot_loss[loss=0.1953, simple_loss=0.4439, pruned_loss=0.09726, ctc_loss=0.1422, over 2854110.03 frames. ], batch size: 33, lr: 3.00e-03, 
2023-01-05 17:16:41,075 INFO [train.py:862] Epoch 10, batch 16000, loss[loss=0.1761, simple_loss=0.436, pruned_loss=0.0836, ctc_loss=0.1223, over 14520.00 frames. ], tot_loss[loss=0.1937, simple_loss=0.442, pruned_loss=0.09601, ctc_loss=0.1409, over 2855732.95 frames. ], batch size: 38, lr: 3.00e-03, 
2023-01-05 17:16:48,190 INFO [checkpoint.py:75] Saving checkpoint to tiny_transducer_ctc/exp_2m_phone_halfdelay/checkpoint-180000.pt
2023-01-05 17:17:46,776 INFO [train.py:862] Epoch 10, batch 16500, loss[loss=0.1948, simple_loss=0.4674, pruned_loss=0.0928, ctc_loss=0.1384, over 14647.00 frames. ], tot_loss[loss=0.1948, simple_loss=0.4438, pruned_loss=0.09678, ctc_loss=0.1416, over 2845760.25 frames. ], batch size: 41, lr: 3.00e-03, 
2023-01-05 17:18:52,420 INFO [train.py:862] Epoch 10, batch 17000, loss[loss=0.1868, simple_loss=0.4636, pruned_loss=0.08919, ctc_loss=0.1293, over 14451.00 frames. ], tot_loss[loss=0.1948, simple_loss=0.4435, pruned_loss=0.09715, ctc_loss=0.1417, over 2854159.22 frames. ], batch size: 50, lr: 3.00e-03, 
2023-01-05 17:19:58,129 INFO [train.py:862] Epoch 10, batch 17500, loss[loss=0.2464, simple_loss=0.4962, pruned_loss=0.1231, ctc_loss=0.193, over 14124.00 frames. ], tot_loss[loss=0.1943, simple_loss=0.4421, pruned_loss=0.09665, ctc_loss=0.1415, over 2863927.26 frames. ], batch size: 56, lr: 3.00e-03, 
2023-01-05 17:21:02,383 INFO [train.py:862] Epoch 10, batch 18000, loss[loss=0.1754, simple_loss=0.442, pruned_loss=0.08359, ctc_loss=0.12, over 14646.00 frames. ], tot_loss[loss=0.1933, simple_loss=0.4415, pruned_loss=0.09612, ctc_loss=0.1404, over 2877414.69 frames. ], batch size: 41, lr: 3.00e-03, 
2023-01-05 17:21:02,384 INFO [train.py:887] Computing validation loss
2023-01-05 17:21:27,569 INFO [train.py:897] Epoch 10, validation: loss=0.2069, simple_loss=0.4575, pruned_loss=0.1044, ctc_loss=0.1528, over 944034.00 frames. 
2023-01-05 17:21:27,569 INFO [train.py:898] Maximum memory allocated so far is 7283MB
2023-01-05 17:21:54,680 INFO [checkpoint.py:75] Saving checkpoint to tiny_transducer_ctc/exp_2m_phone_halfdelay/epoch-10.pt
2023-01-05 17:21:57,362 INFO [train.py:862] Epoch 11, batch 0, loss[loss=0.2474, simple_loss=0.4709, pruned_loss=0.1384, ctc_loss=0.1932, over 14878.00 frames. ], tot_loss[loss=0.2474, simple_loss=0.4709, pruned_loss=0.1384, ctc_loss=0.1932, over 14878.00 frames. ], batch size: 40, lr: 3.00e-03, 
2023-01-05 17:23:02,706 INFO [train.py:862] Epoch 11, batch 500, loss[loss=0.1688, simple_loss=0.3964, pruned_loss=0.08497, ctc_loss=0.1198, over 14540.00 frames. ], tot_loss[loss=0.1933, simple_loss=0.4426, pruned_loss=0.09584, ctc_loss=0.1402, over 2650831.15 frames. ], batch size: 34, lr: 3.00e-03, 
2023-01-05 17:24:07,694 INFO [train.py:862] Epoch 11, batch 1000, loss[loss=0.1968, simple_loss=0.4454, pruned_loss=0.09448, ctc_loss=0.1452, over 14648.00 frames. ], tot_loss[loss=0.1915, simple_loss=0.4416, pruned_loss=0.09444, ctc_loss=0.1385, over 2840615.55 frames. ], batch size: 40, lr: 3.00e-03, 
2023-01-05 17:25:12,497 INFO [train.py:862] Epoch 11, batch 1500, loss[loss=0.2264, simple_loss=0.488, pruned_loss=0.1164, ctc_loss=0.169, over 14839.00 frames. ], tot_loss[loss=0.1949, simple_loss=0.4431, pruned_loss=0.0969, ctc_loss=0.1419, over 2860474.33 frames. ], batch size: 42, lr: 3.00e-03, 
2023-01-05 17:26:17,871 INFO [train.py:862] Epoch 11, batch 2000, loss[loss=0.1905, simple_loss=0.4626, pruned_loss=0.08763, ctc_loss=0.1354, over 14668.00 frames. ], tot_loss[loss=0.1928, simple_loss=0.441, pruned_loss=0.09558, ctc_loss=0.1399, over 2872711.14 frames. ], batch size: 42, lr: 3.00e-03, 
2023-01-05 17:27:22,936 INFO [train.py:862] Epoch 11, batch 2500, loss[loss=0.1796, simple_loss=0.404, pruned_loss=0.08976, ctc_loss=0.1315, over 14094.00 frames. ], tot_loss[loss=0.1938, simple_loss=0.4422, pruned_loss=0.09617, ctc_loss=0.1408, over 2871101.61 frames. ], batch size: 31, lr: 3.00e-03, 
2023-01-05 17:28:28,041 INFO [train.py:862] Epoch 11, batch 3000, loss[loss=0.2103, simple_loss=0.4715, pruned_loss=0.09993, ctc_loss=0.1565, over 14540.00 frames. ], tot_loss[loss=0.1948, simple_loss=0.4435, pruned_loss=0.09683, ctc_loss=0.1418, over 2870259.41 frames. ], batch size: 48, lr: 3.00e-03, 
2023-01-05 17:29:33,333 INFO [train.py:862] Epoch 11, batch 3500, loss[loss=0.2291, simple_loss=0.4737, pruned_loss=0.1198, ctc_loss=0.1745, over 14848.00 frames. ], tot_loss[loss=0.1939, simple_loss=0.4429, pruned_loss=0.09619, ctc_loss=0.1408, over 2866367.02 frames. ], batch size: 42, lr: 3.00e-03, 
2023-01-05 17:30:38,440 INFO [train.py:862] Epoch 11, batch 4000, loss[loss=0.2316, simple_loss=0.4793, pruned_loss=0.1222, ctc_loss=0.1758, over 14491.00 frames. ], tot_loss[loss=0.1944, simple_loss=0.4429, pruned_loss=0.09653, ctc_loss=0.1414, over 2879221.57 frames. ], batch size: 50, lr: 3.00e-03, 
2023-01-05 17:31:43,798 INFO [train.py:862] Epoch 11, batch 4500, loss[loss=0.1957, simple_loss=0.4263, pruned_loss=0.1004, ctc_loss=0.1452, over 14700.00 frames. ], tot_loss[loss=0.1924, simple_loss=0.4411, pruned_loss=0.09522, ctc_loss=0.1395, over 2860456.68 frames. ], batch size: 38, lr: 3.00e-03, 
2023-01-05 17:32:48,763 INFO [train.py:862] Epoch 11, batch 5000, loss[loss=0.1592, simple_loss=0.3924, pruned_loss=0.07745, ctc_loss=0.1102, over 14534.00 frames. ], tot_loss[loss=0.1945, simple_loss=0.4419, pruned_loss=0.0971, ctc_loss=0.1416, over 2861523.17 frames. ], batch size: 34, lr: 3.00e-03, 
2023-01-05 17:33:53,629 INFO [train.py:862] Epoch 11, batch 5500, loss[loss=0.1768, simple_loss=0.4345, pruned_loss=0.08457, ctc_loss=0.1232, over 14788.00 frames. ], tot_loss[loss=0.194, simple_loss=0.4434, pruned_loss=0.09615, ctc_loss=0.1409, over 2855519.28 frames. ], batch size: 43, lr: 3.00e-03, 
2023-01-05 17:34:59,319 INFO [train.py:862] Epoch 11, batch 6000, loss[loss=0.189, simple_loss=0.4367, pruned_loss=0.09672, ctc_loss=0.135, over 14502.00 frames. ], tot_loss[loss=0.1925, simple_loss=0.442, pruned_loss=0.09535, ctc_loss=0.1394, over 2868516.13 frames. ], batch size: 38, lr: 3.00e-03, 
2023-01-05 17:36:04,213 INFO [train.py:862] Epoch 11, batch 6500, loss[loss=0.2325, simple_loss=0.4341, pruned_loss=0.1289, ctc_loss=0.1839, over 14692.00 frames. ], tot_loss[loss=0.1923, simple_loss=0.4411, pruned_loss=0.09548, ctc_loss=0.1392, over 2866276.09 frames. ], batch size: 35, lr: 3.00e-03, 
2023-01-05 17:37:09,841 INFO [train.py:862] Epoch 11, batch 7000, loss[loss=0.2438, simple_loss=0.4879, pruned_loss=0.1308, ctc_loss=0.1877, over 14844.00 frames. ], tot_loss[loss=0.1942, simple_loss=0.4421, pruned_loss=0.09677, ctc_loss=0.1413, over 2871316.18 frames. ], batch size: 42, lr: 3.00e-03, 
2023-01-05 17:38:14,612 INFO [train.py:862] Epoch 11, batch 7500, loss[loss=0.2263, simple_loss=0.4654, pruned_loss=0.1143, ctc_loss=0.1746, over 14589.00 frames. ], tot_loss[loss=0.1923, simple_loss=0.441, pruned_loss=0.09531, ctc_loss=0.1393, over 2853640.75 frames. ], batch size: 43, lr: 3.00e-03, 
2023-01-05 17:38:58,296 INFO [checkpoint.py:75] Saving checkpoint to tiny_transducer_ctc/exp_2m_phone_halfdelay/checkpoint-190000.pt
2023-01-05 17:39:19,775 INFO [train.py:862] Epoch 11, batch 8000, loss[loss=0.2198, simple_loss=0.465, pruned_loss=0.1095, ctc_loss=0.1675, over 14181.00 frames. ], tot_loss[loss=0.1919, simple_loss=0.4409, pruned_loss=0.09466, ctc_loss=0.139, over 2873243.60 frames. ], batch size: 56, lr: 3.00e-03, 
2023-01-05 17:40:24,980 INFO [train.py:862] Epoch 11, batch 8500, loss[loss=0.1896, simple_loss=0.4174, pruned_loss=0.09846, ctc_loss=0.1392, over 14646.00 frames. ], tot_loss[loss=0.193, simple_loss=0.4414, pruned_loss=0.0957, ctc_loss=0.1401, over 2862797.58 frames. ], batch size: 35, lr: 3.00e-03, 
2023-01-05 17:41:29,989 INFO [train.py:862] Epoch 11, batch 9000, loss[loss=0.1947, simple_loss=0.4583, pruned_loss=0.09325, ctc_loss=0.14, over 14824.00 frames. ], tot_loss[loss=0.1947, simple_loss=0.4434, pruned_loss=0.0969, ctc_loss=0.1417, over 2861980.99 frames. ], batch size: 41, lr: 3.00e-03, 
2023-01-05 17:41:29,990 INFO [train.py:887] Computing validation loss
2023-01-05 17:41:55,791 INFO [train.py:897] Epoch 11, validation: loss=0.2056, simple_loss=0.4557, pruned_loss=0.1041, ctc_loss=0.1515, over 944034.00 frames. 
2023-01-05 17:41:55,791 INFO [train.py:898] Maximum memory allocated so far is 7283MB
2023-01-05 17:43:01,365 INFO [train.py:862] Epoch 11, batch 9500, loss[loss=0.2072, simple_loss=0.4499, pruned_loss=0.1072, ctc_loss=0.1536, over 14880.00 frames. ], tot_loss[loss=0.1927, simple_loss=0.4407, pruned_loss=0.09549, ctc_loss=0.14, over 2863669.40 frames. ], batch size: 40, lr: 3.00e-03, 
2023-01-05 17:44:06,965 INFO [train.py:862] Epoch 11, batch 10000, loss[loss=0.2123, simple_loss=0.4736, pruned_loss=0.105, ctc_loss=0.1568, over 14779.00 frames. ], tot_loss[loss=0.1935, simple_loss=0.4423, pruned_loss=0.09582, ctc_loss=0.1405, over 2851735.50 frames. ], batch size: 46, lr: 3.00e-03, 
2023-01-05 17:45:12,966 INFO [train.py:862] Epoch 11, batch 10500, loss[loss=0.2342, simple_loss=0.4936, pruned_loss=0.1219, ctc_loss=0.1765, over 14747.00 frames. ], tot_loss[loss=0.1945, simple_loss=0.4431, pruned_loss=0.09672, ctc_loss=0.1415, over 2858844.09 frames. ], batch size: 45, lr: 3.00e-03, 
2023-01-05 17:46:18,608 INFO [train.py:862] Epoch 11, batch 11000, loss[loss=0.187, simple_loss=0.4188, pruned_loss=0.09388, ctc_loss=0.1371, over 14867.00 frames. ], tot_loss[loss=0.1909, simple_loss=0.441, pruned_loss=0.09404, ctc_loss=0.138, over 2877260.85 frames. ], batch size: 40, lr: 3.00e-03, 
2023-01-05 17:47:24,750 INFO [train.py:862] Epoch 11, batch 11500, loss[loss=0.1702, simple_loss=0.444, pruned_loss=0.07729, ctc_loss=0.1148, over 14833.00 frames. ], tot_loss[loss=0.1929, simple_loss=0.4416, pruned_loss=0.09587, ctc_loss=0.1399, over 2848012.93 frames. ], batch size: 42, lr: 3.00e-03, 
2023-01-05 17:48:30,341 INFO [train.py:862] Epoch 11, batch 12000, loss[loss=0.211, simple_loss=0.4293, pruned_loss=0.1105, ctc_loss=0.162, over 14671.00 frames. ], tot_loss[loss=0.192, simple_loss=0.4413, pruned_loss=0.09481, ctc_loss=0.139, over 2863501.46 frames. ], batch size: 35, lr: 3.00e-03, 
2023-01-05 17:49:35,631 INFO [train.py:862] Epoch 11, batch 12500, loss[loss=0.2026, simple_loss=0.4638, pruned_loss=0.09988, ctc_loss=0.1472, over 13257.00 frames. ], tot_loss[loss=0.1925, simple_loss=0.4417, pruned_loss=0.09535, ctc_loss=0.1395, over 2859248.34 frames. ], batch size: 76, lr: 3.00e-03, 
2023-01-05 17:50:41,567 INFO [train.py:862] Epoch 11, batch 13000, loss[loss=0.2224, simple_loss=0.439, pruned_loss=0.1174, ctc_loss=0.1734, over 14760.00 frames. ], tot_loss[loss=0.1935, simple_loss=0.4425, pruned_loss=0.0961, ctc_loss=0.1404, over 2848210.70 frames. ], batch size: 36, lr: 3.00e-03, 
2023-01-05 17:51:47,221 INFO [train.py:862] Epoch 11, batch 13500, loss[loss=0.1519, simple_loss=0.3949, pruned_loss=0.07134, ctc_loss=0.1017, over 14664.00 frames. ], tot_loss[loss=0.1917, simple_loss=0.4419, pruned_loss=0.09437, ctc_loss=0.1387, over 2866432.63 frames. ], batch size: 35, lr: 3.00e-03, 
2023-01-05 17:52:53,259 INFO [train.py:862] Epoch 11, batch 14000, loss[loss=0.1858, simple_loss=0.4394, pruned_loss=0.092, ctc_loss=0.1319, over 14504.00 frames. ], tot_loss[loss=0.1911, simple_loss=0.4405, pruned_loss=0.09419, ctc_loss=0.1382, over 2857537.84 frames. ], batch size: 37, lr: 3.00e-03, 
2023-01-05 17:53:58,602 INFO [train.py:862] Epoch 11, batch 14500, loss[loss=0.2168, simple_loss=0.4647, pruned_loss=0.1095, ctc_loss=0.1633, over 14732.00 frames. ], tot_loss[loss=0.1956, simple_loss=0.4432, pruned_loss=0.09714, ctc_loss=0.1428, over 2861568.20 frames. ], batch size: 45, lr: 3.00e-03, 
2023-01-05 17:55:04,998 INFO [train.py:862] Epoch 11, batch 15000, loss[loss=0.1782, simple_loss=0.4532, pruned_loss=0.0824, ctc_loss=0.1222, over 14706.00 frames. ], tot_loss[loss=0.1928, simple_loss=0.4407, pruned_loss=0.0956, ctc_loss=0.14, over 2860513.82 frames. ], batch size: 47, lr: 3.00e-03, 
2023-01-05 17:56:10,274 INFO [train.py:862] Epoch 11, batch 15500, loss[loss=0.1597, simple_loss=0.4294, pruned_loss=0.07091, ctc_loss=0.1058, over 14644.00 frames. ], tot_loss[loss=0.1947, simple_loss=0.4416, pruned_loss=0.09671, ctc_loss=0.1421, over 2857083.33 frames. ], batch size: 40, lr: 3.00e-03, 
2023-01-05 17:57:15,934 INFO [train.py:862] Epoch 11, batch 16000, loss[loss=0.1528, simple_loss=0.379, pruned_loss=0.07686, ctc_loss=0.1042, over 14466.00 frames. ], tot_loss[loss=0.1944, simple_loss=0.4426, pruned_loss=0.09666, ctc_loss=0.1414, over 2854362.60 frames. ], batch size: 32, lr: 3.00e-03, 
2023-01-05 17:58:21,959 INFO [train.py:862] Epoch 11, batch 16500, loss[loss=0.1881, simple_loss=0.4012, pruned_loss=0.09306, ctc_loss=0.1428, over 14687.00 frames. ], tot_loss[loss=0.1916, simple_loss=0.4417, pruned_loss=0.09444, ctc_loss=0.1386, over 2850878.17 frames. ], batch size: 35, lr: 3.00e-03, 
2023-01-05 17:59:27,534 INFO [train.py:862] Epoch 11, batch 17000, loss[loss=0.2033, simple_loss=0.4093, pruned_loss=0.1077, ctc_loss=0.1565, over 14390.00 frames. ], tot_loss[loss=0.1915, simple_loss=0.4398, pruned_loss=0.09489, ctc_loss=0.1386, over 2876461.40 frames. ], batch size: 33, lr: 3.00e-03, 
2023-01-05 18:00:33,555 INFO [train.py:862] Epoch 11, batch 17500, loss[loss=0.2096, simple_loss=0.458, pruned_loss=0.1057, ctc_loss=0.156, over 14704.00 frames. ], tot_loss[loss=0.193, simple_loss=0.4408, pruned_loss=0.09569, ctc_loss=0.1403, over 2865671.20 frames. ], batch size: 44, lr: 3.00e-03, 
2023-01-05 18:01:16,737 INFO [checkpoint.py:75] Saving checkpoint to tiny_transducer_ctc/exp_2m_phone_halfdelay/checkpoint-200000.pt
2023-01-05 18:01:37,839 INFO [train.py:862] Epoch 11, batch 18000, loss[loss=0.2396, simple_loss=0.4854, pruned_loss=0.1257, ctc_loss=0.1844, over 10608.00 frames. ], tot_loss[loss=0.1904, simple_loss=0.4399, pruned_loss=0.09394, ctc_loss=0.1375, over 2880124.81 frames. ], batch size: 104, lr: 3.00e-03, 
2023-01-05 18:01:37,839 INFO [train.py:887] Computing validation loss
2023-01-05 18:02:03,869 INFO [train.py:897] Epoch 11, validation: loss=0.2038, simple_loss=0.4547, pruned_loss=0.1029, ctc_loss=0.1496, over 944034.00 frames. 
2023-01-05 18:02:03,870 INFO [train.py:898] Maximum memory allocated so far is 7283MB
2023-01-05 18:02:30,171 INFO [checkpoint.py:75] Saving checkpoint to tiny_transducer_ctc/exp_2m_phone_halfdelay/epoch-11.pt
2023-01-05 18:02:33,216 INFO [train.py:862] Epoch 12, batch 0, loss[loss=0.2388, simple_loss=0.4849, pruned_loss=0.1253, ctc_loss=0.1836, over 14855.00 frames. ], tot_loss[loss=0.2388, simple_loss=0.4849, pruned_loss=0.1253, ctc_loss=0.1836, over 14855.00 frames. ], batch size: 42, lr: 3.00e-03, 
2023-01-05 18:03:38,196 INFO [train.py:862] Epoch 12, batch 500, loss[loss=0.186, simple_loss=0.4381, pruned_loss=0.08915, ctc_loss=0.1336, over 14871.00 frames. ], tot_loss[loss=0.1928, simple_loss=0.4418, pruned_loss=0.09541, ctc_loss=0.1399, over 2629572.92 frames. ], batch size: 40, lr: 3.00e-03, 
2023-01-05 18:04:43,239 INFO [train.py:862] Epoch 12, batch 1000, loss[loss=0.2012, simple_loss=0.4403, pruned_loss=0.1022, ctc_loss=0.1493, over 14499.00 frames. ], tot_loss[loss=0.1923, simple_loss=0.4411, pruned_loss=0.09482, ctc_loss=0.1395, over 2854223.07 frames. ], batch size: 38, lr: 3.00e-03, 
2023-01-05 18:05:48,192 INFO [train.py:862] Epoch 12, batch 1500, loss[loss=0.2082, simple_loss=0.4258, pruned_loss=0.1094, ctc_loss=0.1593, over 14037.00 frames. ], tot_loss[loss=0.1905, simple_loss=0.4398, pruned_loss=0.09398, ctc_loss=0.1376, over 2872404.72 frames. ], batch size: 31, lr: 3.00e-03, 
2023-01-05 18:06:53,299 INFO [train.py:862] Epoch 12, batch 2000, loss[loss=0.1628, simple_loss=0.4071, pruned_loss=0.07305, ctc_loss=0.114, over 14495.00 frames. ], tot_loss[loss=0.191, simple_loss=0.4403, pruned_loss=0.09432, ctc_loss=0.1381, over 2873711.64 frames. ], batch size: 38, lr: 3.00e-03, 
2023-01-05 18:07:59,106 INFO [train.py:862] Epoch 12, batch 2500, loss[loss=0.17, simple_loss=0.4068, pruned_loss=0.0839, ctc_loss=0.1197, over 14088.00 frames. ], tot_loss[loss=0.191, simple_loss=0.4403, pruned_loss=0.09427, ctc_loss=0.138, over 2878055.52 frames. ], batch size: 31, lr: 3.00e-03, 
2023-01-05 18:09:03,874 INFO [train.py:862] Epoch 12, batch 3000, loss[loss=0.2024, simple_loss=0.4561, pruned_loss=0.1008, ctc_loss=0.1483, over 10254.00 frames. ], tot_loss[loss=0.1921, simple_loss=0.4419, pruned_loss=0.09479, ctc_loss=0.1391, over 2875957.21 frames. ], batch size: 105, lr: 3.00e-03, 
2023-01-05 18:10:09,065 INFO [train.py:862] Epoch 12, batch 3500, loss[loss=0.2102, simple_loss=0.4402, pruned_loss=0.1142, ctc_loss=0.157, over 14738.00 frames. ], tot_loss[loss=0.1921, simple_loss=0.4415, pruned_loss=0.09512, ctc_loss=0.1391, over 2881989.77 frames. ], batch size: 37, lr: 3.00e-03, 
2023-01-05 18:11:14,462 INFO [train.py:862] Epoch 12, batch 4000, loss[loss=0.1502, simple_loss=0.4042, pruned_loss=0.06506, ctc_loss=0.1001, over 14805.00 frames. ], tot_loss[loss=0.1906, simple_loss=0.4392, pruned_loss=0.09412, ctc_loss=0.1378, over 2871868.77 frames. ], batch size: 36, lr: 3.00e-03, 
2023-01-05 18:12:19,291 INFO [train.py:862] Epoch 12, batch 4500, loss[loss=0.2133, simple_loss=0.4461, pruned_loss=0.1093, ctc_loss=0.1623, over 14643.00 frames. ], tot_loss[loss=0.191, simple_loss=0.4397, pruned_loss=0.09441, ctc_loss=0.1381, over 2847673.33 frames. ], batch size: 41, lr: 3.00e-03, 
2023-01-05 18:13:24,323 INFO [train.py:862] Epoch 12, batch 5000, loss[loss=0.2081, simple_loss=0.4762, pruned_loss=0.1044, ctc_loss=0.1505, over 14522.00 frames. ], tot_loss[loss=0.1908, simple_loss=0.4395, pruned_loss=0.0943, ctc_loss=0.1381, over 2861188.83 frames. ], batch size: 50, lr: 3.00e-03, 
2023-01-05 18:14:29,539 INFO [train.py:862] Epoch 12, batch 5500, loss[loss=0.1612, simple_loss=0.3935, pruned_loss=0.07524, ctc_loss=0.1137, over 14565.00 frames. ], tot_loss[loss=0.1921, simple_loss=0.4422, pruned_loss=0.09488, ctc_loss=0.139, over 2875567.62 frames. ], batch size: 34, lr: 3.00e-03, 
2023-01-05 18:15:34,522 INFO [train.py:862] Epoch 12, batch 6000, loss[loss=0.2267, simple_loss=0.4648, pruned_loss=0.1153, ctc_loss=0.1749, over 14699.00 frames. ], tot_loss[loss=0.191, simple_loss=0.4401, pruned_loss=0.09441, ctc_loss=0.1381, over 2860335.45 frames. ], batch size: 39, lr: 3.00e-03, 
2023-01-05 18:16:39,558 INFO [train.py:862] Epoch 12, batch 6500, loss[loss=0.1835, simple_loss=0.4183, pruned_loss=0.09162, ctc_loss=0.1332, over 14525.00 frames. ], tot_loss[loss=0.1915, simple_loss=0.4389, pruned_loss=0.09476, ctc_loss=0.1389, over 2875878.04 frames. ], batch size: 36, lr: 3.00e-03, 
2023-01-05 18:17:45,180 INFO [train.py:862] Epoch 12, batch 7000, loss[loss=0.1895, simple_loss=0.4643, pruned_loss=0.08954, ctc_loss=0.1328, over 14579.00 frames. ], tot_loss[loss=0.1918, simple_loss=0.4421, pruned_loss=0.09485, ctc_loss=0.1386, over 2869713.52 frames. ], batch size: 43, lr: 3.00e-03, 
2023-01-05 18:18:50,276 INFO [train.py:862] Epoch 12, batch 7500, loss[loss=0.1936, simple_loss=0.4572, pruned_loss=0.0932, ctc_loss=0.1386, over 14840.00 frames. ], tot_loss[loss=0.1892, simple_loss=0.4385, pruned_loss=0.09304, ctc_loss=0.1364, over 2859841.15 frames. ], batch size: 41, lr: 3.00e-03, 
2023-01-05 18:19:55,829 INFO [train.py:862] Epoch 12, batch 8000, loss[loss=0.1825, simple_loss=0.4338, pruned_loss=0.08965, ctc_loss=0.1293, over 14523.00 frames. ], tot_loss[loss=0.1904, simple_loss=0.4391, pruned_loss=0.0941, ctc_loss=0.1376, over 2863959.76 frames. ], batch size: 37, lr: 3.00e-03, 
2023-01-05 18:21:00,731 INFO [train.py:862] Epoch 12, batch 8500, loss[loss=0.1805, simple_loss=0.4349, pruned_loss=0.0863, ctc_loss=0.1277, over 14098.00 frames. ], tot_loss[loss=0.1918, simple_loss=0.4413, pruned_loss=0.09458, ctc_loss=0.1389, over 2866065.06 frames. ], batch size: 56, lr: 3.00e-03, 
2023-01-05 18:22:06,218 INFO [train.py:862] Epoch 12, batch 9000, loss[loss=0.1776, simple_loss=0.401, pruned_loss=0.08847, ctc_loss=0.1298, over 14410.00 frames. ], tot_loss[loss=0.1927, simple_loss=0.4416, pruned_loss=0.09531, ctc_loss=0.1398, over 2860526.61 frames. ], batch size: 33, lr: 3.00e-03, 
2023-01-05 18:22:06,219 INFO [train.py:887] Computing validation loss
2023-01-05 18:22:31,505 INFO [train.py:897] Epoch 12, validation: loss=0.2048, simple_loss=0.4551, pruned_loss=0.1045, ctc_loss=0.1503, over 944034.00 frames. 
2023-01-05 18:22:31,505 INFO [train.py:898] Maximum memory allocated so far is 7283MB
2023-01-05 18:23:36,505 INFO [train.py:862] Epoch 12, batch 9500, loss[loss=0.154, simple_loss=0.4238, pruned_loss=0.07032, ctc_loss=0.09906, over 14752.00 frames. ], tot_loss[loss=0.1898, simple_loss=0.4395, pruned_loss=0.0934, ctc_loss=0.1369, over 2842990.61 frames. ], batch size: 45, lr: 3.00e-03, 
2023-01-05 18:23:52,359 INFO [checkpoint.py:75] Saving checkpoint to tiny_transducer_ctc/exp_2m_phone_halfdelay/checkpoint-210000.pt
2023-01-05 18:24:42,139 INFO [train.py:862] Epoch 12, batch 10000, loss[loss=0.2464, simple_loss=0.485, pruned_loss=0.1313, ctc_loss=0.1917, over 14588.00 frames. ], tot_loss[loss=0.1905, simple_loss=0.4384, pruned_loss=0.09422, ctc_loss=0.1379, over 2860003.02 frames. ], batch size: 43, lr: 3.00e-03, 
2023-01-05 18:25:47,599 INFO [train.py:862] Epoch 12, batch 10500, loss[loss=0.1879, simple_loss=0.4549, pruned_loss=0.08918, ctc_loss=0.1326, over 14836.00 frames. ], tot_loss[loss=0.1897, simple_loss=0.4401, pruned_loss=0.09303, ctc_loss=0.1369, over 2864431.69 frames. ], batch size: 41, lr: 3.00e-03, 
2023-01-05 18:26:53,021 INFO [train.py:862] Epoch 12, batch 11000, loss[loss=0.2491, simple_loss=0.4883, pruned_loss=0.1265, ctc_loss=0.197, over 14665.00 frames. ], tot_loss[loss=0.1918, simple_loss=0.4416, pruned_loss=0.09463, ctc_loss=0.1388, over 2855101.93 frames. ], batch size: 42, lr: 3.00e-03, 
2023-01-05 18:27:58,331 INFO [train.py:862] Epoch 12, batch 11500, loss[loss=0.1891, simple_loss=0.4523, pruned_loss=0.08977, ctc_loss=0.1348, over 14658.00 frames. ], tot_loss[loss=0.1914, simple_loss=0.4403, pruned_loss=0.09493, ctc_loss=0.1384, over 2862304.41 frames. ], batch size: 40, lr: 3.00e-03, 
2023-01-05 18:29:03,384 INFO [train.py:862] Epoch 12, batch 12000, loss[loss=0.2692, simple_loss=0.4945, pruned_loss=0.1401, ctc_loss=0.2185, over 14653.00 frames. ], tot_loss[loss=0.1905, simple_loss=0.4409, pruned_loss=0.09372, ctc_loss=0.1375, over 2875368.15 frames. ], batch size: 41, lr: 3.00e-03, 
2023-01-05 18:30:08,743 INFO [train.py:862] Epoch 12, batch 12500, loss[loss=0.1438, simple_loss=0.3738, pruned_loss=0.06425, ctc_loss=0.09777, over 14511.00 frames. ], tot_loss[loss=0.1916, simple_loss=0.442, pruned_loss=0.09468, ctc_loss=0.1384, over 2846816.90 frames. ], batch size: 36, lr: 3.00e-03, 
2023-01-05 18:31:14,181 INFO [train.py:862] Epoch 12, batch 13000, loss[loss=0.1894, simple_loss=0.4235, pruned_loss=0.09376, ctc_loss=0.1397, over 14790.00 frames. ], tot_loss[loss=0.1883, simple_loss=0.4388, pruned_loss=0.09228, ctc_loss=0.1354, over 2861553.10 frames. ], batch size: 36, lr: 3.00e-03, 
2023-01-05 18:32:19,618 INFO [train.py:862] Epoch 12, batch 13500, loss[loss=0.1472, simple_loss=0.3759, pruned_loss=0.07076, ctc_loss=0.09941, over 14402.00 frames. ], tot_loss[loss=0.191, simple_loss=0.4413, pruned_loss=0.09413, ctc_loss=0.1379, over 2861724.05 frames. ], batch size: 33, lr: 3.00e-03, 
2023-01-05 18:33:24,754 INFO [train.py:862] Epoch 12, batch 14000, loss[loss=0.1738, simple_loss=0.408, pruned_loss=0.08416, ctc_loss=0.1249, over 14518.00 frames. ], tot_loss[loss=0.1906, simple_loss=0.4398, pruned_loss=0.0937, ctc_loss=0.138, over 2848845.69 frames. ], batch size: 34, lr: 3.00e-03, 
2023-01-05 18:34:30,200 INFO [train.py:862] Epoch 12, batch 14500, loss[loss=0.22, simple_loss=0.4759, pruned_loss=0.1118, ctc_loss=0.1644, over 14807.00 frames. ], tot_loss[loss=0.1923, simple_loss=0.4407, pruned_loss=0.09532, ctc_loss=0.1394, over 2864009.25 frames. ], batch size: 43, lr: 3.00e-03, 
2023-01-05 18:35:35,716 INFO [train.py:862] Epoch 12, batch 15000, loss[loss=0.2011, simple_loss=0.4613, pruned_loss=0.1021, ctc_loss=0.1447, over 14736.00 frames. ], tot_loss[loss=0.1923, simple_loss=0.4419, pruned_loss=0.09491, ctc_loss=0.1394, over 2858807.98 frames. ], batch size: 46, lr: 3.00e-03, 
2023-01-05 18:36:40,994 INFO [train.py:862] Epoch 12, batch 15500, loss[loss=0.1836, simple_loss=0.4497, pruned_loss=0.08444, ctc_loss=0.1298, over 14650.00 frames. ], tot_loss[loss=0.19, simple_loss=0.4381, pruned_loss=0.09377, ctc_loss=0.1374, over 2866621.10 frames. ], batch size: 41, lr: 3.00e-03, 
2023-01-05 18:37:46,676 INFO [train.py:862] Epoch 12, batch 16000, loss[loss=0.2333, simple_loss=0.4741, pruned_loss=0.1135, ctc_loss=0.183, over 14668.00 frames. ], tot_loss[loss=0.1874, simple_loss=0.4373, pruned_loss=0.0917, ctc_loss=0.1347, over 2869735.30 frames. ], batch size: 42, lr: 3.00e-03, 
2023-01-05 18:38:51,965 INFO [train.py:862] Epoch 12, batch 16500, loss[loss=0.1685, simple_loss=0.4004, pruned_loss=0.08208, ctc_loss=0.1197, over 14413.00 frames. ], tot_loss[loss=0.1886, simple_loss=0.4392, pruned_loss=0.09253, ctc_loss=0.1357, over 2861565.69 frames. ], batch size: 33, lr: 3.00e-03, 
2023-01-05 18:39:57,174 INFO [train.py:862] Epoch 12, batch 17000, loss[loss=0.1642, simple_loss=0.3949, pruned_loss=0.07795, ctc_loss=0.1166, over 14552.00 frames. ], tot_loss[loss=0.1889, simple_loss=0.4385, pruned_loss=0.09285, ctc_loss=0.1361, over 2871013.06 frames. ], batch size: 34, lr: 3.00e-03, 
2023-01-05 18:41:02,525 INFO [train.py:862] Epoch 12, batch 17500, loss[loss=0.1895, simple_loss=0.4468, pruned_loss=0.09065, ctc_loss=0.1361, over 14835.00 frames. ], tot_loss[loss=0.1925, simple_loss=0.4415, pruned_loss=0.09551, ctc_loss=0.1395, over 2852168.77 frames. ], batch size: 42, lr: 3.00e-03, 
2023-01-05 18:42:06,112 INFO [train.py:862] Epoch 12, batch 18000, loss[loss=0.21, simple_loss=0.4616, pruned_loss=0.1035, ctc_loss=0.1567, over 14664.00 frames. ], tot_loss[loss=0.1907, simple_loss=0.4401, pruned_loss=0.09409, ctc_loss=0.1378, over 2849307.32 frames. ], batch size: 40, lr: 3.00e-03, 
2023-01-05 18:42:06,112 INFO [train.py:887] Computing validation loss
2023-01-05 18:42:32,180 INFO [train.py:897] Epoch 12, validation: loss=0.2042, simple_loss=0.4539, pruned_loss=0.1029, ctc_loss=0.1503, over 944034.00 frames. 
2023-01-05 18:42:32,181 INFO [train.py:898] Maximum memory allocated so far is 7283MB
2023-01-05 18:42:58,787 INFO [checkpoint.py:75] Saving checkpoint to tiny_transducer_ctc/exp_2m_phone_halfdelay/epoch-12.pt
2023-01-05 18:43:01,520 INFO [train.py:862] Epoch 13, batch 0, loss[loss=0.2186, simple_loss=0.4623, pruned_loss=0.1208, ctc_loss=0.1615, over 14656.00 frames. ], tot_loss[loss=0.2186, simple_loss=0.4623, pruned_loss=0.1208, ctc_loss=0.1615, over 14656.00 frames. ], batch size: 41, lr: 3.00e-03, 
2023-01-05 18:44:06,531 INFO [train.py:862] Epoch 13, batch 500, loss[loss=0.1762, simple_loss=0.441, pruned_loss=0.08289, ctc_loss=0.1218, over 14576.00 frames. ], tot_loss[loss=0.1916, simple_loss=0.4399, pruned_loss=0.09471, ctc_loss=0.1389, over 2643421.82 frames. ], batch size: 50, lr: 3.00e-03, 
2023-01-05 18:45:11,571 INFO [train.py:862] Epoch 13, batch 1000, loss[loss=0.1786, simple_loss=0.4525, pruned_loss=0.08405, ctc_loss=0.1221, over 14699.00 frames. ], tot_loss[loss=0.1904, simple_loss=0.4404, pruned_loss=0.09363, ctc_loss=0.1375, over 2857709.54 frames. ], batch size: 44, lr: 3.00e-03, 
2023-01-05 18:46:04,484 INFO [checkpoint.py:75] Saving checkpoint to tiny_transducer_ctc/exp_2m_phone_halfdelay/checkpoint-220000.pt
2023-01-05 18:46:16,774 INFO [train.py:862] Epoch 13, batch 1500, loss[loss=0.1787, simple_loss=0.4428, pruned_loss=0.08385, ctc_loss=0.1245, over 14516.00 frames. ], tot_loss[loss=0.1892, simple_loss=0.4397, pruned_loss=0.09291, ctc_loss=0.1363, over 2857449.75 frames. ], batch size: 50, lr: 3.00e-03, 
2023-01-05 18:47:21,596 INFO [train.py:862] Epoch 13, batch 2000, loss[loss=0.1545, simple_loss=0.3932, pruned_loss=0.07157, ctc_loss=0.1058, over 14690.00 frames. ], tot_loss[loss=0.1882, simple_loss=0.4379, pruned_loss=0.09224, ctc_loss=0.1355, over 2862727.15 frames. ], batch size: 35, lr: 3.00e-03, 
2023-01-05 18:48:27,018 INFO [train.py:862] Epoch 13, batch 2500, loss[loss=0.2539, simple_loss=0.4851, pruned_loss=0.1309, ctc_loss=0.2027, over 14834.00 frames. ], tot_loss[loss=0.1899, simple_loss=0.4409, pruned_loss=0.0933, ctc_loss=0.1369, over 2874066.57 frames. ], batch size: 42, lr: 3.00e-03, 
2023-01-05 18:49:32,117 INFO [train.py:862] Epoch 13, batch 3000, loss[loss=0.1696, simple_loss=0.4308, pruned_loss=0.07985, ctc_loss=0.1158, over 14590.00 frames. ], tot_loss[loss=0.1904, simple_loss=0.4404, pruned_loss=0.09389, ctc_loss=0.1375, over 2873268.48 frames. ], batch size: 43, lr: 3.00e-03, 
2023-01-05 18:50:37,512 INFO [train.py:862] Epoch 13, batch 3500, loss[loss=0.2074, simple_loss=0.4279, pruned_loss=0.1038, ctc_loss=0.1601, over 14718.00 frames. ], tot_loss[loss=0.191, simple_loss=0.4412, pruned_loss=0.094, ctc_loss=0.1381, over 2871655.69 frames. ], batch size: 37, lr: 3.00e-03, 
2023-01-05 18:51:42,362 INFO [train.py:862] Epoch 13, batch 4000, loss[loss=0.1593, simple_loss=0.4185, pruned_loss=0.07399, ctc_loss=0.1062, over 14696.00 frames. ], tot_loss[loss=0.1883, simple_loss=0.4375, pruned_loss=0.09245, ctc_loss=0.1356, over 2869786.62 frames. ], batch size: 38, lr: 3.00e-03, 
2023-01-05 18:52:47,526 INFO [train.py:862] Epoch 13, batch 4500, loss[loss=0.1876, simple_loss=0.4523, pruned_loss=0.09026, ctc_loss=0.1324, over 14637.00 frames. ], tot_loss[loss=0.1883, simple_loss=0.4397, pruned_loss=0.09235, ctc_loss=0.1351, over 2855051.74 frames. ], batch size: 41, lr: 3.00e-03, 
2023-01-05 18:53:52,344 INFO [train.py:862] Epoch 13, batch 5000, loss[loss=0.1929, simple_loss=0.4611, pruned_loss=0.0921, ctc_loss=0.1373, over 14741.00 frames. ], tot_loss[loss=0.1888, simple_loss=0.4397, pruned_loss=0.09279, ctc_loss=0.1358, over 2861078.19 frames. ], batch size: 46, lr: 3.00e-03, 
2023-01-05 18:54:57,569 INFO [train.py:862] Epoch 13, batch 5500, loss[loss=0.2217, simple_loss=0.4439, pruned_loss=0.1204, ctc_loss=0.17, over 14694.00 frames. ], tot_loss[loss=0.1916, simple_loss=0.4405, pruned_loss=0.09482, ctc_loss=0.1386, over 2859216.85 frames. ], batch size: 39, lr: 3.00e-03, 
2023-01-05 18:56:02,897 INFO [train.py:862] Epoch 13, batch 6000, loss[loss=0.1509, simple_loss=0.3904, pruned_loss=0.06826, ctc_loss=0.1027, over 14526.00 frames. ], tot_loss[loss=0.1904, simple_loss=0.4392, pruned_loss=0.09412, ctc_loss=0.1375, over 2872542.48 frames. ], batch size: 36, lr: 3.00e-03, 
2023-01-05 18:57:07,762 INFO [train.py:862] Epoch 13, batch 6500, loss[loss=0.1772, simple_loss=0.4462, pruned_loss=0.08556, ctc_loss=0.1209, over 14488.00 frames. ], tot_loss[loss=0.1907, simple_loss=0.4395, pruned_loss=0.09416, ctc_loss=0.1378, over 2860602.46 frames. ], batch size: 50, lr: 3.00e-03, 
2023-01-05 18:58:13,367 INFO [train.py:862] Epoch 13, batch 7000, loss[loss=0.1674, simple_loss=0.4277, pruned_loss=0.07796, ctc_loss=0.114, over 12932.00 frames. ], tot_loss[loss=0.1888, simple_loss=0.4374, pruned_loss=0.09296, ctc_loss=0.1362, over 2886566.70 frames. ], batch size: 76, lr: 3.00e-03, 
2023-01-05 18:59:18,530 INFO [train.py:862] Epoch 13, batch 7500, loss[loss=0.2099, simple_loss=0.4328, pruned_loss=0.1118, ctc_loss=0.1592, over 14683.00 frames. ], tot_loss[loss=0.1898, simple_loss=0.4391, pruned_loss=0.09358, ctc_loss=0.137, over 2870131.01 frames. ], batch size: 35, lr: 3.00e-03, 
2023-01-05 19:00:23,750 INFO [train.py:862] Epoch 13, batch 8000, loss[loss=0.1992, simple_loss=0.4593, pruned_loss=0.09648, ctc_loss=0.1448, over 14806.00 frames. ], tot_loss[loss=0.1919, simple_loss=0.441, pruned_loss=0.09506, ctc_loss=0.1389, over 2862866.98 frames. ], batch size: 43, lr: 3.00e-03, 
2023-01-05 19:01:29,079 INFO [train.py:862] Epoch 13, batch 8500, loss[loss=0.1929, simple_loss=0.459, pruned_loss=0.09392, ctc_loss=0.137, over 14659.00 frames. ], tot_loss[loss=0.1898, simple_loss=0.4398, pruned_loss=0.09343, ctc_loss=0.1368, over 2873292.44 frames. ], batch size: 42, lr: 3.00e-03, 
2023-01-05 19:02:34,071 INFO [train.py:862] Epoch 13, batch 9000, loss[loss=0.1861, simple_loss=0.4347, pruned_loss=0.0911, ctc_loss=0.1336, over 14899.00 frames. ], tot_loss[loss=0.1895, simple_loss=0.4391, pruned_loss=0.09325, ctc_loss=0.1367, over 2867973.53 frames. ], batch size: 39, lr: 3.00e-03, 
2023-01-05 19:02:34,072 INFO [train.py:887] Computing validation loss
2023-01-05 19:02:59,555 INFO [train.py:897] Epoch 13, validation: loss=0.2031, simple_loss=0.4534, pruned_loss=0.1015, ctc_loss=0.1494, over 944034.00 frames. 
2023-01-05 19:02:59,556 INFO [train.py:898] Maximum memory allocated so far is 7283MB
2023-01-05 19:04:05,570 INFO [train.py:862] Epoch 13, batch 9500, loss[loss=0.1958, simple_loss=0.4306, pruned_loss=0.104, ctc_loss=0.1429, over 14728.00 frames. ], tot_loss[loss=0.1892, simple_loss=0.4388, pruned_loss=0.0932, ctc_loss=0.1363, over 2881352.11 frames. ], batch size: 37, lr: 3.00e-03, 
2023-01-05 19:05:10,866 INFO [train.py:862] Epoch 13, batch 10000, loss[loss=0.2228, simple_loss=0.4456, pruned_loss=0.1149, ctc_loss=0.1736, over 9810.00 frames. ], tot_loss[loss=0.1901, simple_loss=0.4395, pruned_loss=0.09357, ctc_loss=0.1373, over 2852645.59 frames. ], batch size: 103, lr: 3.00e-03, 
2023-01-05 19:06:16,894 INFO [train.py:862] Epoch 13, batch 10500, loss[loss=0.1697, simple_loss=0.4447, pruned_loss=0.07526, ctc_loss=0.1149, over 14732.00 frames. ], tot_loss[loss=0.1892, simple_loss=0.4387, pruned_loss=0.09272, ctc_loss=0.1365, over 2878645.36 frames. ], batch size: 45, lr: 3.00e-03, 
2023-01-05 19:07:21,961 INFO [train.py:862] Epoch 13, batch 11000, loss[loss=0.1676, simple_loss=0.4209, pruned_loss=0.08256, ctc_loss=0.1139, over 14505.00 frames. ], tot_loss[loss=0.1892, simple_loss=0.439, pruned_loss=0.09294, ctc_loss=0.1364, over 2859131.66 frames. ], batch size: 37, lr: 3.00e-03, 
2023-01-05 19:08:15,148 INFO [checkpoint.py:75] Saving checkpoint to tiny_transducer_ctc/exp_2m_phone_halfdelay/checkpoint-230000.pt
2023-01-05 19:08:27,934 INFO [train.py:862] Epoch 13, batch 11500, loss[loss=0.1845, simple_loss=0.4473, pruned_loss=0.08582, ctc_loss=0.131, over 14515.00 frames. ], tot_loss[loss=0.188, simple_loss=0.4392, pruned_loss=0.0921, ctc_loss=0.135, over 2864515.85 frames. ], batch size: 53, lr: 3.00e-03, 
2023-01-05 19:09:33,398 INFO [train.py:862] Epoch 13, batch 12000, loss[loss=0.1992, simple_loss=0.4453, pruned_loss=0.1009, ctc_loss=0.146, over 9794.00 frames. ], tot_loss[loss=0.1878, simple_loss=0.4383, pruned_loss=0.09173, ctc_loss=0.135, over 2859455.58 frames. ], batch size: 104, lr: 3.00e-03, 
2023-01-05 19:10:38,879 INFO [train.py:862] Epoch 13, batch 12500, loss[loss=0.1559, simple_loss=0.384, pruned_loss=0.07706, ctc_loss=0.1074, over 14536.00 frames. ], tot_loss[loss=0.1853, simple_loss=0.4371, pruned_loss=0.09034, ctc_loss=0.1324, over 2868807.65 frames. ], batch size: 34, lr: 3.00e-03, 
2023-01-05 19:11:44,485 INFO [train.py:862] Epoch 13, batch 13000, loss[loss=0.233, simple_loss=0.4902, pruned_loss=0.121, ctc_loss=0.176, over 14231.00 frames. ], tot_loss[loss=0.1919, simple_loss=0.442, pruned_loss=0.09482, ctc_loss=0.1387, over 2848453.50 frames. ], batch size: 56, lr: 3.00e-03, 
2023-01-05 19:12:50,051 INFO [train.py:862] Epoch 13, batch 13500, loss[loss=0.2227, simple_loss=0.4768, pruned_loss=0.1171, ctc_loss=0.1658, over 14798.00 frames. ], tot_loss[loss=0.1895, simple_loss=0.4392, pruned_loss=0.09333, ctc_loss=0.1366, over 2841755.94 frames. ], batch size: 43, lr: 3.00e-03, 
2023-01-05 19:13:55,755 INFO [train.py:862] Epoch 13, batch 14000, loss[loss=0.2089, simple_loss=0.4602, pruned_loss=0.105, ctc_loss=0.1547, over 14682.00 frames. ], tot_loss[loss=0.1871, simple_loss=0.4373, pruned_loss=0.0914, ctc_loss=0.1345, over 2863411.17 frames. ], batch size: 44, lr: 3.00e-03, 
2023-01-05 19:15:01,288 INFO [train.py:862] Epoch 13, batch 14500, loss[loss=0.1983, simple_loss=0.4412, pruned_loss=0.1023, ctc_loss=0.1449, over 14519.00 frames. ], tot_loss[loss=0.1881, simple_loss=0.4378, pruned_loss=0.0922, ctc_loss=0.1354, over 2868127.13 frames. ], batch size: 36, lr: 3.00e-03, 
2023-01-05 19:16:06,630 INFO [train.py:862] Epoch 13, batch 15000, loss[loss=0.1839, simple_loss=0.4143, pruned_loss=0.09486, ctc_loss=0.1332, over 14526.00 frames. ], tot_loss[loss=0.1894, simple_loss=0.4396, pruned_loss=0.09322, ctc_loss=0.1365, over 2869401.48 frames. ], batch size: 34, lr: 3.00e-03, 
2023-01-05 19:17:12,619 INFO [train.py:862] Epoch 13, batch 15500, loss[loss=0.1664, simple_loss=0.4052, pruned_loss=0.08207, ctc_loss=0.1158, over 14688.00 frames. ], tot_loss[loss=0.1896, simple_loss=0.4393, pruned_loss=0.09316, ctc_loss=0.1368, over 2864996.67 frames. ], batch size: 35, lr: 3.00e-03, 
2023-01-05 19:18:17,909 INFO [train.py:862] Epoch 13, batch 16000, loss[loss=0.1373, simple_loss=0.3865, pruned_loss=0.0619, ctc_loss=0.08674, over 14440.00 frames. ], tot_loss[loss=0.1873, simple_loss=0.4368, pruned_loss=0.09183, ctc_loss=0.1346, over 2848593.71 frames. ], batch size: 33, lr: 3.00e-03, 
2023-01-05 19:19:23,688 INFO [train.py:862] Epoch 13, batch 16500, loss[loss=0.1719, simple_loss=0.4114, pruned_loss=0.08462, ctc_loss=0.1212, over 14700.00 frames. ], tot_loss[loss=0.1868, simple_loss=0.4383, pruned_loss=0.09119, ctc_loss=0.1339, over 2863386.71 frames. ], batch size: 38, lr: 3.00e-03, 
2023-01-05 19:20:29,086 INFO [train.py:862] Epoch 13, batch 17000, loss[loss=0.1949, simple_loss=0.4538, pruned_loss=0.09606, ctc_loss=0.14, over 14237.00 frames. ], tot_loss[loss=0.1882, simple_loss=0.4378, pruned_loss=0.09238, ctc_loss=0.1354, over 2868243.91 frames. ], batch size: 52, lr: 3.00e-03, 
2023-01-05 19:21:35,094 INFO [train.py:862] Epoch 13, batch 17500, loss[loss=0.2293, simple_loss=0.4676, pruned_loss=0.1195, ctc_loss=0.1761, over 13681.00 frames. ], tot_loss[loss=0.1874, simple_loss=0.4377, pruned_loss=0.09197, ctc_loss=0.1346, over 2866268.57 frames. ], batch size: 63, lr: 3.00e-03, 
2023-01-05 19:22:38,939 INFO [train.py:862] Epoch 13, batch 18000, loss[loss=0.1425, simple_loss=0.3859, pruned_loss=0.06249, ctc_loss=0.09412, over 14780.00 frames. ], tot_loss[loss=0.1877, simple_loss=0.437, pruned_loss=0.09182, ctc_loss=0.1351, over 2863428.31 frames. ], batch size: 36, lr: 3.00e-03, 
2023-01-05 19:22:38,939 INFO [train.py:887] Computing validation loss
2023-01-05 19:23:04,339 INFO [train.py:897] Epoch 13, validation: loss=0.2065, simple_loss=0.4554, pruned_loss=0.1036, ctc_loss=0.153, over 944034.00 frames. 
2023-01-05 19:23:04,339 INFO [train.py:898] Maximum memory allocated so far is 7283MB
2023-01-05 19:23:31,402 INFO [checkpoint.py:75] Saving checkpoint to tiny_transducer_ctc/exp_2m_phone_halfdelay/epoch-13.pt
2023-01-05 19:23:34,096 INFO [train.py:862] Epoch 14, batch 0, loss[loss=0.2468, simple_loss=0.4972, pruned_loss=0.1366, ctc_loss=0.1875, over 14704.00 frames. ], tot_loss[loss=0.2468, simple_loss=0.4972, pruned_loss=0.1366, ctc_loss=0.1875, over 14704.00 frames. ], batch size: 46, lr: 3.00e-03, 
2023-01-05 19:24:39,165 INFO [train.py:862] Epoch 14, batch 500, loss[loss=0.2333, simple_loss=0.468, pruned_loss=0.1228, ctc_loss=0.1803, over 10069.00 frames. ], tot_loss[loss=0.1907, simple_loss=0.44, pruned_loss=0.09387, ctc_loss=0.1379, over 2624651.04 frames. ], batch size: 104, lr: 3.00e-03, 
2023-01-05 19:25:44,048 INFO [train.py:862] Epoch 14, batch 1000, loss[loss=0.1876, simple_loss=0.4188, pruned_loss=0.09407, ctc_loss=0.1379, over 14401.00 frames. ], tot_loss[loss=0.1881, simple_loss=0.4373, pruned_loss=0.09239, ctc_loss=0.1355, over 2842522.83 frames. ], batch size: 33, lr: 3.00e-03, 
2023-01-05 19:26:49,228 INFO [train.py:862] Epoch 14, batch 1500, loss[loss=0.1737, simple_loss=0.4467, pruned_loss=0.07612, ctc_loss=0.1198, over 14595.00 frames. ], tot_loss[loss=0.1887, simple_loss=0.4391, pruned_loss=0.09243, ctc_loss=0.1359, over 2851438.38 frames. ], batch size: 43, lr: 3.00e-03, 
2023-01-05 19:27:54,063 INFO [train.py:862] Epoch 14, batch 2000, loss[loss=0.1741, simple_loss=0.3939, pruned_loss=0.08502, ctc_loss=0.1278, over 14536.00 frames. ], tot_loss[loss=0.1881, simple_loss=0.4388, pruned_loss=0.09241, ctc_loss=0.1351, over 2876615.76 frames. ], batch size: 34, lr: 3.00e-03, 
2023-01-05 19:28:58,469 INFO [train.py:862] Epoch 14, batch 2500, loss[loss=0.1527, simple_loss=0.3809, pruned_loss=0.07109, ctc_loss=0.1061, over 14007.00 frames. ], tot_loss[loss=0.1876, simple_loss=0.4389, pruned_loss=0.09163, ctc_loss=0.1347, over 2861472.21 frames. ], batch size: 31, lr: 3.00e-03, 
2023-01-05 19:30:03,896 INFO [train.py:862] Epoch 14, batch 3000, loss[loss=0.1682, simple_loss=0.4349, pruned_loss=0.07643, ctc_loss=0.1144, over 14714.00 frames. ], tot_loss[loss=0.1884, simple_loss=0.4378, pruned_loss=0.09254, ctc_loss=0.1357, over 2876143.56 frames. ], batch size: 47, lr: 3.00e-03, 
2023-01-05 19:30:28,591 INFO [checkpoint.py:75] Saving checkpoint to tiny_transducer_ctc/exp_2m_phone_halfdelay/checkpoint-240000.pt
2023-01-05 19:31:08,976 INFO [train.py:862] Epoch 14, batch 3500, loss[loss=0.1857, simple_loss=0.4346, pruned_loss=0.09297, ctc_loss=0.1323, over 14522.00 frames. ], tot_loss[loss=0.1884, simple_loss=0.4379, pruned_loss=0.09256, ctc_loss=0.1356, over 2867246.11 frames. ], batch size: 37, lr: 3.00e-03, 
2023-01-05 19:32:13,933 INFO [train.py:862] Epoch 14, batch 4000, loss[loss=0.1925, simple_loss=0.4543, pruned_loss=0.09029, ctc_loss=0.1389, over 14652.00 frames. ], tot_loss[loss=0.188, simple_loss=0.4369, pruned_loss=0.09233, ctc_loss=0.1354, over 2856328.29 frames. ], batch size: 42, lr: 3.00e-03, 
2023-01-05 19:33:19,453 INFO [train.py:862] Epoch 14, batch 4500, loss[loss=0.1742, simple_loss=0.4342, pruned_loss=0.08164, ctc_loss=0.1209, over 14742.00 frames. ], tot_loss[loss=0.1892, simple_loss=0.4396, pruned_loss=0.0931, ctc_loss=0.1362, over 2864425.01 frames. ], batch size: 45, lr: 3.00e-03, 
2023-01-05 19:34:24,348 INFO [train.py:862] Epoch 14, batch 5000, loss[loss=0.1809, simple_loss=0.4551, pruned_loss=0.08152, ctc_loss=0.126, over 14684.00 frames. ], tot_loss[loss=0.189, simple_loss=0.4388, pruned_loss=0.0929, ctc_loss=0.1362, over 2881681.35 frames. ], batch size: 47, lr: 3.00e-03, 
2023-01-05 19:35:29,511 INFO [train.py:862] Epoch 14, batch 5500, loss[loss=0.1594, simple_loss=0.4176, pruned_loss=0.07472, ctc_loss=0.1062, over 14667.00 frames. ], tot_loss[loss=0.19, simple_loss=0.4408, pruned_loss=0.09354, ctc_loss=0.1369, over 2854269.99 frames. ], batch size: 40, lr: 3.00e-03, 
2023-01-05 19:36:34,513 INFO [train.py:862] Epoch 14, batch 6000, loss[loss=0.1985, simple_loss=0.4452, pruned_loss=0.09718, ctc_loss=0.1465, over 14640.00 frames. ], tot_loss[loss=0.1911, simple_loss=0.4413, pruned_loss=0.0941, ctc_loss=0.1381, over 2853377.65 frames. ], batch size: 41, lr: 3.00e-03, 
2023-01-05 19:37:39,382 INFO [train.py:862] Epoch 14, batch 6500, loss[loss=0.2179, simple_loss=0.4395, pruned_loss=0.1193, ctc_loss=0.166, over 14723.00 frames. ], tot_loss[loss=0.1856, simple_loss=0.4357, pruned_loss=0.09071, ctc_loss=0.1329, over 2869207.85 frames. ], batch size: 38, lr: 3.00e-03, 
2023-01-05 19:38:44,719 INFO [train.py:862] Epoch 14, batch 7000, loss[loss=0.1626, simple_loss=0.3931, pruned_loss=0.07909, ctc_loss=0.1142, over 14812.00 frames. ], tot_loss[loss=0.1884, simple_loss=0.438, pruned_loss=0.09251, ctc_loss=0.1356, over 2856131.62 frames. ], batch size: 36, lr: 3.00e-03, 
2023-01-05 19:39:49,469 INFO [train.py:862] Epoch 14, batch 7500, loss[loss=0.1439, simple_loss=0.3878, pruned_loss=0.06686, ctc_loss=0.09387, over 14550.00 frames. ], tot_loss[loss=0.187, simple_loss=0.4367, pruned_loss=0.09167, ctc_loss=0.1343, over 2874588.35 frames. ], batch size: 36, lr: 3.00e-03, 
2023-01-05 19:40:55,035 INFO [train.py:862] Epoch 14, batch 8000, loss[loss=0.1671, simple_loss=0.4434, pruned_loss=0.07753, ctc_loss=0.1104, over 14540.00 frames. ], tot_loss[loss=0.1867, simple_loss=0.4365, pruned_loss=0.09136, ctc_loss=0.1341, over 2855925.99 frames. ], batch size: 50, lr: 3.00e-03, 
2023-01-05 19:41:59,697 INFO [train.py:862] Epoch 14, batch 8500, loss[loss=0.213, simple_loss=0.442, pruned_loss=0.099, ctc_loss=0.1672, over 14498.00 frames. ], tot_loss[loss=0.1857, simple_loss=0.4354, pruned_loss=0.09082, ctc_loss=0.133, over 2869510.80 frames. ], batch size: 37, lr: 3.00e-03, 
2023-01-05 19:43:04,723 INFO [train.py:862] Epoch 14, batch 9000, loss[loss=0.1667, simple_loss=0.4173, pruned_loss=0.0796, ctc_loss=0.1146, over 14694.00 frames. ], tot_loss[loss=0.1878, simple_loss=0.4376, pruned_loss=0.09201, ctc_loss=0.1351, over 2872994.88 frames. ], batch size: 39, lr: 3.00e-03, 
2023-01-05 19:43:04,723 INFO [train.py:887] Computing validation loss
2023-01-05 19:43:30,333 INFO [train.py:897] Epoch 14, validation: loss=0.2029, simple_loss=0.4529, pruned_loss=0.102, ctc_loss=0.1491, over 944034.00 frames. 
2023-01-05 19:43:30,333 INFO [train.py:898] Maximum memory allocated so far is 7283MB
2023-01-05 19:44:35,955 INFO [train.py:862] Epoch 14, batch 9500, loss[loss=0.1659, simple_loss=0.3998, pruned_loss=0.07917, ctc_loss=0.1174, over 14695.00 frames. ], tot_loss[loss=0.1872, simple_loss=0.4379, pruned_loss=0.09168, ctc_loss=0.1343, over 2878811.77 frames. ], batch size: 38, lr: 3.00e-03, 
2023-01-05 19:45:41,951 INFO [train.py:862] Epoch 14, batch 10000, loss[loss=0.1928, simple_loss=0.4412, pruned_loss=0.09544, ctc_loss=0.1399, over 14656.00 frames. ], tot_loss[loss=0.189, simple_loss=0.4385, pruned_loss=0.09273, ctc_loss=0.1363, over 2838221.92 frames. ], batch size: 41, lr: 3.00e-03, 
2023-01-05 19:46:47,137 INFO [train.py:862] Epoch 14, batch 10500, loss[loss=0.1627, simple_loss=0.394, pruned_loss=0.07877, ctc_loss=0.1142, over 14513.00 frames. ], tot_loss[loss=0.1913, simple_loss=0.4409, pruned_loss=0.09407, ctc_loss=0.1384, over 2869579.14 frames. ], batch size: 36, lr: 3.00e-03, 
2023-01-05 19:47:52,575 INFO [train.py:862] Epoch 14, batch 11000, loss[loss=0.2037, simple_loss=0.4689, pruned_loss=0.09956, ctc_loss=0.1479, over 13519.00 frames. ], tot_loss[loss=0.1865, simple_loss=0.4369, pruned_loss=0.09125, ctc_loss=0.1336, over 2855546.64 frames. ], batch size: 63, lr: 3.00e-03, 
2023-01-05 19:48:58,564 INFO [train.py:862] Epoch 14, batch 11500, loss[loss=0.213, simple_loss=0.4659, pruned_loss=0.1086, ctc_loss=0.1579, over 14647.00 frames. ], tot_loss[loss=0.1854, simple_loss=0.4359, pruned_loss=0.0906, ctc_loss=0.1327, over 2867514.67 frames. ], batch size: 42, lr: 3.00e-03, 
2023-01-05 19:50:04,148 INFO [train.py:862] Epoch 14, batch 12000, loss[loss=0.1635, simple_loss=0.3969, pruned_loss=0.07717, ctc_loss=0.1155, over 14781.00 frames. ], tot_loss[loss=0.1883, simple_loss=0.4393, pruned_loss=0.09236, ctc_loss=0.1353, over 2867777.27 frames. ], batch size: 36, lr: 3.00e-03, 
2023-01-05 19:51:10,012 INFO [train.py:862] Epoch 14, batch 12500, loss[loss=0.1909, simple_loss=0.4123, pruned_loss=0.09907, ctc_loss=0.1419, over 14539.00 frames. ], tot_loss[loss=0.1887, simple_loss=0.4381, pruned_loss=0.09272, ctc_loss=0.136, over 2880059.05 frames. ], batch size: 34, lr: 3.00e-03, 
2023-01-05 19:52:15,578 INFO [train.py:862] Epoch 14, batch 13000, loss[loss=0.2251, simple_loss=0.4795, pruned_loss=0.1148, ctc_loss=0.1695, over 14520.00 frames. ], tot_loss[loss=0.1861, simple_loss=0.4375, pruned_loss=0.09067, ctc_loss=0.1332, over 2858207.32 frames. ], batch size: 48, lr: 3.00e-03, 
2023-01-05 19:52:40,680 INFO [checkpoint.py:75] Saving checkpoint to tiny_transducer_ctc/exp_2m_phone_halfdelay/checkpoint-250000.pt
2023-01-05 19:53:21,422 INFO [train.py:862] Epoch 14, batch 13500, loss[loss=0.2162, simple_loss=0.4548, pruned_loss=0.1145, ctc_loss=0.1623, over 14695.00 frames. ], tot_loss[loss=0.1882, simple_loss=0.4387, pruned_loss=0.09236, ctc_loss=0.1352, over 2868272.14 frames. ], batch size: 39, lr: 3.00e-03, 
2023-01-05 19:54:27,560 INFO [train.py:862] Epoch 14, batch 14000, loss[loss=0.1696, simple_loss=0.4189, pruned_loss=0.07614, ctc_loss=0.1199, over 14886.00 frames. ], tot_loss[loss=0.1879, simple_loss=0.4371, pruned_loss=0.09205, ctc_loss=0.1353, over 2841687.26 frames. ], batch size: 39, lr: 3.00e-03, 
2023-01-05 19:55:32,926 INFO [train.py:862] Epoch 14, batch 14500, loss[loss=0.1926, simple_loss=0.4444, pruned_loss=0.09459, ctc_loss=0.1393, over 14648.00 frames. ], tot_loss[loss=0.1876, simple_loss=0.4374, pruned_loss=0.09181, ctc_loss=0.1349, over 2875847.77 frames. ], batch size: 40, lr: 3.00e-03, 
2023-01-05 19:56:39,005 INFO [train.py:862] Epoch 14, batch 15000, loss[loss=0.1989, simple_loss=0.4111, pruned_loss=0.1016, ctc_loss=0.1525, over 14030.00 frames. ], tot_loss[loss=0.1881, simple_loss=0.4379, pruned_loss=0.0921, ctc_loss=0.1354, over 2857868.73 frames. ], batch size: 31, lr: 3.00e-03, 
2023-01-05 19:57:44,389 INFO [train.py:862] Epoch 14, batch 15500, loss[loss=0.1484, simple_loss=0.4182, pruned_loss=0.06733, ctc_loss=0.09358, over 14651.00 frames. ], tot_loss[loss=0.187, simple_loss=0.4369, pruned_loss=0.09172, ctc_loss=0.1342, over 2859502.42 frames. ], batch size: 40, lr: 3.00e-03, 
2023-01-05 19:58:50,399 INFO [train.py:862] Epoch 14, batch 16000, loss[loss=0.2384, simple_loss=0.485, pruned_loss=0.1233, ctc_loss=0.1838, over 14658.00 frames. ], tot_loss[loss=0.1875, simple_loss=0.4373, pruned_loss=0.09196, ctc_loss=0.1348, over 2846781.65 frames. ], batch size: 42, lr: 3.00e-03, 
2023-01-05 19:59:56,258 INFO [train.py:862] Epoch 14, batch 16500, loss[loss=0.1785, simple_loss=0.4387, pruned_loss=0.08423, ctc_loss=0.1249, over 14689.00 frames. ], tot_loss[loss=0.1868, simple_loss=0.4369, pruned_loss=0.09128, ctc_loss=0.1342, over 2866940.47 frames. ], batch size: 39, lr: 3.00e-03, 
2023-01-05 20:01:01,567 INFO [train.py:862] Epoch 14, batch 17000, loss[loss=0.2092, simple_loss=0.4613, pruned_loss=0.1057, ctc_loss=0.1547, over 14826.00 frames. ], tot_loss[loss=0.1868, simple_loss=0.4371, pruned_loss=0.09138, ctc_loss=0.134, over 2856959.78 frames. ], batch size: 41, lr: 3.00e-03, 
2023-01-05 20:02:07,408 INFO [train.py:862] Epoch 14, batch 17500, loss[loss=0.2273, simple_loss=0.4649, pruned_loss=0.1196, ctc_loss=0.1738, over 13219.00 frames. ], tot_loss[loss=0.1859, simple_loss=0.4357, pruned_loss=0.09088, ctc_loss=0.1333, over 2867676.70 frames. ], batch size: 76, lr: 3.00e-03, 
2023-01-05 20:03:11,493 INFO [train.py:862] Epoch 14, batch 18000, loss[loss=0.1773, simple_loss=0.4342, pruned_loss=0.08019, ctc_loss=0.1259, over 14004.00 frames. ], tot_loss[loss=0.1849, simple_loss=0.4357, pruned_loss=0.09007, ctc_loss=0.1322, over 2855264.80 frames. ], batch size: 56, lr: 3.00e-03, 
2023-01-05 20:03:11,494 INFO [train.py:887] Computing validation loss
2023-01-05 20:03:37,404 INFO [train.py:897] Epoch 14, validation: loss=0.2018, simple_loss=0.4524, pruned_loss=0.1009, ctc_loss=0.1482, over 944034.00 frames. 
2023-01-05 20:03:37,405 INFO [train.py:898] Maximum memory allocated so far is 7283MB
2023-01-05 20:04:04,072 INFO [checkpoint.py:75] Saving checkpoint to tiny_transducer_ctc/exp_2m_phone_halfdelay/epoch-14.pt
2023-01-05 20:04:07,138 INFO [train.py:862] Epoch 15, batch 0, loss[loss=0.218, simple_loss=0.4592, pruned_loss=0.1086, ctc_loss=0.1665, over 14094.00 frames. ], tot_loss[loss=0.218, simple_loss=0.4592, pruned_loss=0.1086, ctc_loss=0.1665, over 14094.00 frames. ], batch size: 56, lr: 3.00e-03, 
2023-01-05 20:05:11,887 INFO [train.py:862] Epoch 15, batch 500, loss[loss=0.1583, simple_loss=0.3916, pruned_loss=0.07593, ctc_loss=0.1097, over 14431.00 frames. ], tot_loss[loss=0.1893, simple_loss=0.4397, pruned_loss=0.09292, ctc_loss=0.1364, over 2636379.45 frames. ], batch size: 33, lr: 3.00e-03, 
2023-01-05 20:06:16,674 INFO [train.py:862] Epoch 15, batch 1000, loss[loss=0.1934, simple_loss=0.4378, pruned_loss=0.0931, ctc_loss=0.1426, over 14831.00 frames. ], tot_loss[loss=0.187, simple_loss=0.4385, pruned_loss=0.09125, ctc_loss=0.134, over 2846112.67 frames. ], batch size: 41, lr: 3.00e-03, 
2023-01-05 20:07:21,926 INFO [train.py:862] Epoch 15, batch 1500, loss[loss=0.1647, simple_loss=0.4393, pruned_loss=0.07208, ctc_loss=0.1102, over 14669.00 frames. ], tot_loss[loss=0.1892, simple_loss=0.4377, pruned_loss=0.0928, ctc_loss=0.1367, over 2860946.59 frames. ], batch size: 42, lr: 3.00e-03, 
2023-01-05 20:08:26,606 INFO [train.py:862] Epoch 15, batch 2000, loss[loss=0.1656, simple_loss=0.4221, pruned_loss=0.07817, ctc_loss=0.1126, over 14832.00 frames. ], tot_loss[loss=0.1874, simple_loss=0.4374, pruned_loss=0.09173, ctc_loss=0.1347, over 2861097.96 frames. ], batch size: 41, lr: 3.00e-03, 
2023-01-05 20:09:32,283 INFO [train.py:862] Epoch 15, batch 2500, loss[loss=0.2332, simple_loss=0.4879, pruned_loss=0.1217, ctc_loss=0.1764, over 14545.00 frames. ], tot_loss[loss=0.1861, simple_loss=0.4389, pruned_loss=0.09072, ctc_loss=0.133, over 2870803.40 frames. ], batch size: 53, lr: 3.00e-03, 
2023-01-05 20:10:37,194 INFO [train.py:862] Epoch 15, batch 3000, loss[loss=0.1955, simple_loss=0.4507, pruned_loss=0.09499, ctc_loss=0.1421, over 14750.00 frames. ], tot_loss[loss=0.1863, simple_loss=0.4363, pruned_loss=0.09099, ctc_loss=0.1336, over 2874498.54 frames. ], batch size: 46, lr: 3.00e-03, 
2023-01-05 20:11:42,417 INFO [train.py:862] Epoch 15, batch 3500, loss[loss=0.1818, simple_loss=0.4002, pruned_loss=0.09324, ctc_loss=0.134, over 14553.00 frames. ], tot_loss[loss=0.1879, simple_loss=0.4373, pruned_loss=0.09218, ctc_loss=0.1352, over 2871992.82 frames. ], batch size: 34, lr: 3.00e-03, 
2023-01-05 20:12:47,892 INFO [train.py:862] Epoch 15, batch 4000, loss[loss=0.1933, simple_loss=0.4623, pruned_loss=0.09553, ctc_loss=0.1361, over 14641.00 frames. ], tot_loss[loss=0.1871, simple_loss=0.4367, pruned_loss=0.09165, ctc_loss=0.1344, over 2869268.90 frames. ], batch size: 41, lr: 3.00e-03, 
2023-01-05 20:13:52,873 INFO [train.py:862] Epoch 15, batch 4500, loss[loss=0.1821, simple_loss=0.4367, pruned_loss=0.08658, ctc_loss=0.1294, over 14635.00 frames. ], tot_loss[loss=0.1865, simple_loss=0.437, pruned_loss=0.09114, ctc_loss=0.1338, over 2869254.54 frames. ], batch size: 40, lr: 3.00e-03, 
2023-01-05 20:14:54,178 INFO [checkpoint.py:75] Saving checkpoint to tiny_transducer_ctc/exp_2m_phone_halfdelay/checkpoint-260000.pt
2023-01-05 20:14:57,932 INFO [train.py:862] Epoch 15, batch 5000, loss[loss=0.2127, simple_loss=0.4583, pruned_loss=0.1074, ctc_loss=0.1597, over 9679.00 frames. ], tot_loss[loss=0.1867, simple_loss=0.4362, pruned_loss=0.0913, ctc_loss=0.1341, over 2866690.19 frames. ], batch size: 104, lr: 3.00e-03, 
2023-01-05 20:16:03,349 INFO [train.py:862] Epoch 15, batch 5500, loss[loss=0.2424, simple_loss=0.4629, pruned_loss=0.1216, ctc_loss=0.195, over 10087.00 frames. ], tot_loss[loss=0.1877, simple_loss=0.4372, pruned_loss=0.09194, ctc_loss=0.135, over 2859620.56 frames. ], batch size: 104, lr: 3.00e-03, 
2023-01-05 20:17:08,291 INFO [train.py:862] Epoch 15, batch 6000, loss[loss=0.1983, simple_loss=0.443, pruned_loss=0.09971, ctc_loss=0.1457, over 14659.00 frames. ], tot_loss[loss=0.1869, simple_loss=0.4375, pruned_loss=0.09142, ctc_loss=0.134, over 2879444.28 frames. ], batch size: 41, lr: 3.00e-03, 
2023-01-05 20:18:13,155 INFO [train.py:862] Epoch 15, batch 6500, loss[loss=0.246, simple_loss=0.48, pruned_loss=0.1384, ctc_loss=0.1893, over 10266.00 frames. ], tot_loss[loss=0.187, simple_loss=0.4374, pruned_loss=0.09146, ctc_loss=0.1342, over 2863676.89 frames. ], batch size: 104, lr: 3.00e-03, 
2023-01-05 20:19:18,547 INFO [train.py:862] Epoch 15, batch 7000, loss[loss=0.1656, simple_loss=0.3966, pruned_loss=0.08075, ctc_loss=0.117, over 14382.00 frames. ], tot_loss[loss=0.1881, simple_loss=0.438, pruned_loss=0.09247, ctc_loss=0.1352, over 2864977.22 frames. ], batch size: 32, lr: 3.00e-03, 
2023-01-05 20:20:23,590 INFO [train.py:862] Epoch 15, batch 7500, loss[loss=0.1942, simple_loss=0.4233, pruned_loss=0.09269, ctc_loss=0.147, over 14711.00 frames. ], tot_loss[loss=0.1857, simple_loss=0.4362, pruned_loss=0.0904, ctc_loss=0.1331, over 2854263.15 frames. ], batch size: 38, lr: 3.00e-03, 
2023-01-05 20:21:29,097 INFO [train.py:862] Epoch 15, batch 8000, loss[loss=0.2026, simple_loss=0.4678, pruned_loss=0.1011, ctc_loss=0.1458, over 14571.00 frames. ], tot_loss[loss=0.188, simple_loss=0.4377, pruned_loss=0.0924, ctc_loss=0.1352, over 2854403.46 frames. ], batch size: 43, lr: 3.00e-03, 
2023-01-05 20:22:33,652 INFO [train.py:862] Epoch 15, batch 8500, loss[loss=0.1862, simple_loss=0.4337, pruned_loss=0.09276, ctc_loss=0.1333, over 14589.00 frames. ], tot_loss[loss=0.1865, simple_loss=0.4363, pruned_loss=0.09123, ctc_loss=0.1338, over 2869484.50 frames. ], batch size: 43, lr: 3.00e-03, 
2023-01-05 20:23:38,408 INFO [train.py:862] Epoch 15, batch 9000, loss[loss=0.1661, simple_loss=0.4471, pruned_loss=0.07354, ctc_loss=0.11, over 14542.00 frames. ], tot_loss[loss=0.1855, simple_loss=0.4372, pruned_loss=0.09022, ctc_loss=0.1327, over 2852815.15 frames. ], batch size: 48, lr: 3.00e-03, 
2023-01-05 20:23:38,409 INFO [train.py:887] Computing validation loss
2023-01-05 20:24:04,560 INFO [train.py:897] Epoch 15, validation: loss=0.2009, simple_loss=0.4517, pruned_loss=0.1009, ctc_loss=0.1469, over 944034.00 frames. 
2023-01-05 20:24:04,561 INFO [train.py:898] Maximum memory allocated so far is 7283MB
2023-01-05 20:25:10,584 INFO [train.py:862] Epoch 15, batch 9500, loss[loss=0.2077, simple_loss=0.4748, pruned_loss=0.1044, ctc_loss=0.1501, over 14837.00 frames. ], tot_loss[loss=0.1853, simple_loss=0.4362, pruned_loss=0.09039, ctc_loss=0.1326, over 2872449.65 frames. ], batch size: 42, lr: 3.00e-03, 
2023-01-05 20:26:17,044 INFO [train.py:862] Epoch 15, batch 10000, loss[loss=0.1985, simple_loss=0.4384, pruned_loss=0.09816, ctc_loss=0.1476, over 14514.00 frames. ], tot_loss[loss=0.1877, simple_loss=0.4379, pruned_loss=0.09179, ctc_loss=0.135, over 2849183.01 frames. ], batch size: 38, lr: 3.00e-03, 
2023-01-05 20:27:23,190 INFO [train.py:862] Epoch 15, batch 10500, loss[loss=0.1888, simple_loss=0.4575, pruned_loss=0.08881, ctc_loss=0.1337, over 14654.00 frames. ], tot_loss[loss=0.1862, simple_loss=0.4372, pruned_loss=0.09101, ctc_loss=0.1334, over 2866593.81 frames. ], batch size: 42, lr: 3.00e-03, 
2023-01-05 20:28:29,595 INFO [train.py:862] Epoch 15, batch 11000, loss[loss=0.1628, simple_loss=0.4141, pruned_loss=0.07606, ctc_loss=0.1113, over 14700.00 frames. ], tot_loss[loss=0.1851, simple_loss=0.4362, pruned_loss=0.0901, ctc_loss=0.1324, over 2856614.78 frames. ], batch size: 39, lr: 3.00e-03, 
2023-01-05 20:29:35,806 INFO [train.py:862] Epoch 15, batch 11500, loss[loss=0.2069, simple_loss=0.4691, pruned_loss=0.1029, ctc_loss=0.151, over 14761.00 frames. ], tot_loss[loss=0.1864, simple_loss=0.4374, pruned_loss=0.09108, ctc_loss=0.1336, over 2871593.00 frames. ], batch size: 46, lr: 3.00e-03, 
2023-01-05 20:30:41,556 INFO [train.py:862] Epoch 15, batch 12000, loss[loss=0.1548, simple_loss=0.4168, pruned_loss=0.07102, ctc_loss=0.1015, over 14694.00 frames. ], tot_loss[loss=0.186, simple_loss=0.4374, pruned_loss=0.09072, ctc_loss=0.1331, over 2866878.31 frames. ], batch size: 37, lr: 3.00e-03, 
2023-01-05 20:31:48,286 INFO [train.py:862] Epoch 15, batch 12500, loss[loss=0.2218, simple_loss=0.4628, pruned_loss=0.114, ctc_loss=0.1688, over 14536.00 frames. ], tot_loss[loss=0.1876, simple_loss=0.4373, pruned_loss=0.09201, ctc_loss=0.1348, over 2871769.25 frames. ], batch size: 53, lr: 3.00e-03, 
2023-01-05 20:32:54,191 INFO [train.py:862] Epoch 15, batch 13000, loss[loss=0.1626, simple_loss=0.3976, pruned_loss=0.07693, ctc_loss=0.1141, over 14683.00 frames. ], tot_loss[loss=0.1856, simple_loss=0.436, pruned_loss=0.09071, ctc_loss=0.1328, over 2857212.14 frames. ], batch size: 35, lr: 3.00e-03, 
2023-01-05 20:34:00,532 INFO [train.py:862] Epoch 15, batch 13500, loss[loss=0.2066, simple_loss=0.4581, pruned_loss=0.1035, ctc_loss=0.1527, over 14128.00 frames. ], tot_loss[loss=0.1877, simple_loss=0.4384, pruned_loss=0.09173, ctc_loss=0.1349, over 2874723.84 frames. ], batch size: 56, lr: 3.00e-03, 
2023-01-05 20:35:07,291 INFO [train.py:862] Epoch 15, batch 14000, loss[loss=0.1833, simple_loss=0.3981, pruned_loss=0.09374, ctc_loss=0.1363, over 14388.00 frames. ], tot_loss[loss=0.1854, simple_loss=0.4358, pruned_loss=0.09036, ctc_loss=0.1328, over 2847423.25 frames. ], batch size: 33, lr: 3.00e-03, 
2023-01-05 20:36:13,551 INFO [train.py:862] Epoch 15, batch 14500, loss[loss=0.1719, simple_loss=0.4478, pruned_loss=0.08338, ctc_loss=0.1139, over 14492.00 frames. ], tot_loss[loss=0.1853, simple_loss=0.4362, pruned_loss=0.09036, ctc_loss=0.1325, over 2866908.42 frames. ], batch size: 48, lr: 3.00e-03, 
2023-01-05 20:37:16,743 INFO [checkpoint.py:75] Saving checkpoint to tiny_transducer_ctc/exp_2m_phone_halfdelay/checkpoint-270000.pt
2023-01-05 20:37:20,296 INFO [train.py:862] Epoch 15, batch 15000, loss[loss=0.2195, simple_loss=0.4657, pruned_loss=0.1122, ctc_loss=0.1657, over 12827.00 frames. ], tot_loss[loss=0.1875, simple_loss=0.4383, pruned_loss=0.09158, ctc_loss=0.1347, over 2860504.32 frames. ], batch size: 76, lr: 3.00e-03, 
2023-01-05 20:38:26,223 INFO [train.py:862] Epoch 15, batch 15500, loss[loss=0.2088, simple_loss=0.4624, pruned_loss=0.1066, ctc_loss=0.1536, over 14668.00 frames. ], tot_loss[loss=0.1855, simple_loss=0.4364, pruned_loss=0.09052, ctc_loss=0.1327, over 2868725.50 frames. ], batch size: 40, lr: 3.00e-03, 
2023-01-05 20:39:32,870 INFO [train.py:862] Epoch 15, batch 16000, loss[loss=0.155, simple_loss=0.4059, pruned_loss=0.07125, ctc_loss=0.104, over 14702.00 frames. ], tot_loss[loss=0.1848, simple_loss=0.4351, pruned_loss=0.09007, ctc_loss=0.1322, over 2866392.58 frames. ], batch size: 38, lr: 3.00e-03, 
2023-01-05 20:40:39,028 INFO [train.py:862] Epoch 15, batch 16500, loss[loss=0.192, simple_loss=0.4557, pruned_loss=0.09255, ctc_loss=0.137, over 13639.00 frames. ], tot_loss[loss=0.1858, simple_loss=0.4362, pruned_loss=0.09088, ctc_loss=0.133, over 2853185.06 frames. ], batch size: 63, lr: 3.00e-03, 
2023-01-05 20:41:45,174 INFO [train.py:862] Epoch 15, batch 17000, loss[loss=0.173, simple_loss=0.3969, pruned_loss=0.08738, ctc_loss=0.1246, over 14522.00 frames. ], tot_loss[loss=0.1859, simple_loss=0.4364, pruned_loss=0.09091, ctc_loss=0.1331, over 2854687.62 frames. ], batch size: 34, lr: 3.00e-03, 
2023-01-05 20:42:51,821 INFO [train.py:862] Epoch 15, batch 17500, loss[loss=0.17, simple_loss=0.4369, pruned_loss=0.08112, ctc_loss=0.1144, over 14651.00 frames. ], tot_loss[loss=0.1859, simple_loss=0.4377, pruned_loss=0.0906, ctc_loss=0.133, over 2852936.23 frames. ], batch size: 40, lr: 3.00e-03, 
2023-01-05 20:43:56,657 INFO [train.py:862] Epoch 15, batch 18000, loss[loss=0.2137, simple_loss=0.4661, pruned_loss=0.1073, ctc_loss=0.1594, over 14552.00 frames. ], tot_loss[loss=0.1862, simple_loss=0.4357, pruned_loss=0.09118, ctc_loss=0.1335, over 2863459.80 frames. ], batch size: 48, lr: 3.00e-03, 
2023-01-05 20:43:56,657 INFO [train.py:887] Computing validation loss
2023-01-05 20:44:22,630 INFO [train.py:897] Epoch 15, validation: loss=0.2019, simple_loss=0.4522, pruned_loss=0.1015, ctc_loss=0.1481, over 944034.00 frames. 
2023-01-05 20:44:22,630 INFO [train.py:898] Maximum memory allocated so far is 7283MB
2023-01-05 20:44:49,576 INFO [checkpoint.py:75] Saving checkpoint to tiny_transducer_ctc/exp_2m_phone_halfdelay/epoch-15.pt
2023-01-05 20:44:52,398 INFO [train.py:862] Epoch 16, batch 0, loss[loss=0.2185, simple_loss=0.4725, pruned_loss=0.1137, ctc_loss=0.1622, over 14532.00 frames. ], tot_loss[loss=0.2185, simple_loss=0.4725, pruned_loss=0.1137, ctc_loss=0.1622, over 14532.00 frames. ], batch size: 48, lr: 3.00e-03, 
2023-01-05 20:45:57,854 INFO [train.py:862] Epoch 16, batch 500, loss[loss=0.181, simple_loss=0.4659, pruned_loss=0.08588, ctc_loss=0.1219, over 14722.00 frames. ], tot_loss[loss=0.1863, simple_loss=0.4375, pruned_loss=0.09096, ctc_loss=0.1334, over 2658899.21 frames. ], batch size: 45, lr: 3.00e-03, 
2023-01-05 20:47:03,402 INFO [train.py:862] Epoch 16, batch 1000, loss[loss=0.2026, simple_loss=0.473, pruned_loss=0.09828, ctc_loss=0.146, over 14732.00 frames. ], tot_loss[loss=0.1854, simple_loss=0.4363, pruned_loss=0.0903, ctc_loss=0.1327, over 2861684.24 frames. ], batch size: 45, lr: 3.00e-03, 
2023-01-05 20:48:08,451 INFO [train.py:862] Epoch 16, batch 1500, loss[loss=0.2461, simple_loss=0.483, pruned_loss=0.1251, ctc_loss=0.1945, over 12736.00 frames. ], tot_loss[loss=0.1859, simple_loss=0.4358, pruned_loss=0.09056, ctc_loss=0.1333, over 2864357.25 frames. ], batch size: 76, lr: 3.00e-03, 
2023-01-05 20:49:13,471 INFO [train.py:862] Epoch 16, batch 2000, loss[loss=0.19, simple_loss=0.4386, pruned_loss=0.09503, ctc_loss=0.1367, over 13970.00 frames. ], tot_loss[loss=0.1871, simple_loss=0.4391, pruned_loss=0.09134, ctc_loss=0.134, over 2850928.57 frames. ], batch size: 63, lr: 3.00e-03, 
2023-01-05 20:50:19,209 INFO [train.py:862] Epoch 16, batch 2500, loss[loss=0.1607, simple_loss=0.3964, pruned_loss=0.07794, ctc_loss=0.1112, over 14679.00 frames. ], tot_loss[loss=0.1855, simple_loss=0.4374, pruned_loss=0.09046, ctc_loss=0.1325, over 2866112.67 frames. ], batch size: 35, lr: 3.00e-03, 
2023-01-05 20:51:24,463 INFO [train.py:862] Epoch 16, batch 3000, loss[loss=0.1522, simple_loss=0.4017, pruned_loss=0.06654, ctc_loss=0.1028, over 14706.00 frames. ], tot_loss[loss=0.1846, simple_loss=0.4355, pruned_loss=0.08976, ctc_loss=0.132, over 2881973.43 frames. ], batch size: 38, lr: 3.00e-03, 
2023-01-05 20:52:29,810 INFO [train.py:862] Epoch 16, batch 3500, loss[loss=0.1997, simple_loss=0.471, pruned_loss=0.0959, ctc_loss=0.1432, over 14545.00 frames. ], tot_loss[loss=0.1862, simple_loss=0.4368, pruned_loss=0.09079, ctc_loss=0.1335, over 2862907.49 frames. ], batch size: 48, lr: 3.00e-03, 
2023-01-05 20:53:35,806 INFO [train.py:862] Epoch 16, batch 4000, loss[loss=0.2068, simple_loss=0.469, pruned_loss=0.1022, ctc_loss=0.1512, over 14136.00 frames. ], tot_loss[loss=0.1845, simple_loss=0.4364, pruned_loss=0.08946, ctc_loss=0.1318, over 2877565.81 frames. ], batch size: 56, lr: 3.00e-03, 
2023-01-05 20:54:40,742 INFO [train.py:862] Epoch 16, batch 4500, loss[loss=0.1749, simple_loss=0.4517, pruned_loss=0.07775, ctc_loss=0.1197, over 14582.00 frames. ], tot_loss[loss=0.1873, simple_loss=0.4364, pruned_loss=0.09153, ctc_loss=0.1348, over 2872814.05 frames. ], batch size: 43, lr: 3.00e-03, 
2023-01-05 20:55:46,326 INFO [train.py:862] Epoch 16, batch 5000, loss[loss=0.1922, simple_loss=0.443, pruned_loss=0.09445, ctc_loss=0.1391, over 14659.00 frames. ], tot_loss[loss=0.1854, simple_loss=0.4359, pruned_loss=0.0903, ctc_loss=0.1327, over 2854044.12 frames. ], batch size: 42, lr: 3.00e-03, 
2023-01-05 20:56:51,419 INFO [train.py:862] Epoch 16, batch 5500, loss[loss=0.1922, simple_loss=0.4376, pruned_loss=0.09522, ctc_loss=0.1399, over 14705.00 frames. ], tot_loss[loss=0.1842, simple_loss=0.4361, pruned_loss=0.08936, ctc_loss=0.1315, over 2869541.34 frames. ], batch size: 39, lr: 3.00e-03, 
2023-01-05 20:57:56,969 INFO [train.py:862] Epoch 16, batch 6000, loss[loss=0.1613, simple_loss=0.4278, pruned_loss=0.07087, ctc_loss=0.1085, over 14878.00 frames. ], tot_loss[loss=0.1845, simple_loss=0.4351, pruned_loss=0.08994, ctc_loss=0.1319, over 2871846.80 frames. ], batch size: 40, lr: 3.00e-03, 
2023-01-05 20:59:02,661 INFO [train.py:862] Epoch 16, batch 6500, loss[loss=0.1498, simple_loss=0.3987, pruned_loss=0.06989, ctc_loss=0.09858, over 14866.00 frames. ], tot_loss[loss=0.1847, simple_loss=0.4353, pruned_loss=0.09017, ctc_loss=0.1319, over 2873773.99 frames. ], batch size: 40, lr: 3.00e-03, 
2023-01-05 20:59:36,328 INFO [checkpoint.py:75] Saving checkpoint to tiny_transducer_ctc/exp_2m_phone_halfdelay/checkpoint-280000.pt
2023-01-05 21:00:08,381 INFO [train.py:862] Epoch 16, batch 7000, loss[loss=0.1934, simple_loss=0.4659, pruned_loss=0.09008, ctc_loss=0.1378, over 14503.00 frames. ], tot_loss[loss=0.1857, simple_loss=0.4356, pruned_loss=0.09076, ctc_loss=0.133, over 2869592.86 frames. ], batch size: 50, lr: 3.00e-03, 
2023-01-05 21:01:14,526 INFO [train.py:862] Epoch 16, batch 7500, loss[loss=0.1744, simple_loss=0.3987, pruned_loss=0.08822, ctc_loss=0.1259, over 14402.00 frames. ], tot_loss[loss=0.1848, simple_loss=0.4356, pruned_loss=0.09004, ctc_loss=0.1321, over 2867347.77 frames. ], batch size: 33, lr: 3.00e-03, 
2023-01-05 21:02:19,865 INFO [train.py:862] Epoch 16, batch 8000, loss[loss=0.2047, simple_loss=0.4543, pruned_loss=0.1036, ctc_loss=0.1507, over 13057.00 frames. ], tot_loss[loss=0.1853, simple_loss=0.4365, pruned_loss=0.09023, ctc_loss=0.1325, over 2865480.61 frames. ], batch size: 76, lr: 3.00e-03, 
2023-01-05 21:03:25,298 INFO [train.py:862] Epoch 16, batch 8500, loss[loss=0.1945, simple_loss=0.4518, pruned_loss=0.09832, ctc_loss=0.1389, over 14664.00 frames. ], tot_loss[loss=0.1843, simple_loss=0.4359, pruned_loss=0.08973, ctc_loss=0.1314, over 2877695.60 frames. ], batch size: 40, lr: 3.00e-03, 
2023-01-05 21:04:31,127 INFO [train.py:862] Epoch 16, batch 9000, loss[loss=0.1884, simple_loss=0.475, pruned_loss=0.08812, ctc_loss=0.1295, over 14529.00 frames. ], tot_loss[loss=0.1832, simple_loss=0.4349, pruned_loss=0.08869, ctc_loss=0.1306, over 2865617.98 frames. ], batch size: 48, lr: 3.00e-03, 
2023-01-05 21:04:31,127 INFO [train.py:887] Computing validation loss
2023-01-05 21:04:56,605 INFO [train.py:897] Epoch 16, validation: loss=0.2007, simple_loss=0.4514, pruned_loss=0.09956, ctc_loss=0.1473, over 944034.00 frames. 
2023-01-05 21:04:56,605 INFO [train.py:898] Maximum memory allocated so far is 7283MB
2023-01-05 21:06:02,332 INFO [train.py:862] Epoch 16, batch 9500, loss[loss=0.1771, simple_loss=0.3965, pruned_loss=0.08763, ctc_loss=0.1304, over 14019.00 frames. ], tot_loss[loss=0.1836, simple_loss=0.4346, pruned_loss=0.08911, ctc_loss=0.131, over 2858202.12 frames. ], batch size: 31, lr: 3.00e-03, 
2023-01-05 21:07:07,717 INFO [train.py:862] Epoch 16, batch 10000, loss[loss=0.1425, simple_loss=0.4006, pruned_loss=0.06327, ctc_loss=0.09058, over 14515.00 frames. ], tot_loss[loss=0.186, simple_loss=0.4361, pruned_loss=0.09076, ctc_loss=0.1333, over 2833086.76 frames. ], batch size: 37, lr: 3.00e-03, 
2023-01-05 21:08:13,648 INFO [train.py:862] Epoch 16, batch 10500, loss[loss=0.1818, simple_loss=0.4556, pruned_loss=0.08869, ctc_loss=0.1241, over 14749.00 frames. ], tot_loss[loss=0.1869, simple_loss=0.4374, pruned_loss=0.09132, ctc_loss=0.1341, over 2846989.29 frames. ], batch size: 46, lr: 3.00e-03, 
2023-01-05 21:09:19,294 INFO [train.py:862] Epoch 16, batch 11000, loss[loss=0.238, simple_loss=0.4765, pruned_loss=0.1265, ctc_loss=0.1836, over 14846.00 frames. ], tot_loss[loss=0.1844, simple_loss=0.4352, pruned_loss=0.08965, ctc_loss=0.1317, over 2869841.63 frames. ], batch size: 42, lr: 3.00e-03, 
2023-01-05 21:10:24,461 INFO [train.py:862] Epoch 16, batch 11500, loss[loss=0.1832, simple_loss=0.4537, pruned_loss=0.08492, ctc_loss=0.1281, over 14738.00 frames. ], tot_loss[loss=0.1866, simple_loss=0.4374, pruned_loss=0.09128, ctc_loss=0.1337, over 2862070.58 frames. ], batch size: 45, lr: 3.00e-03, 
2023-01-05 21:11:30,422 INFO [train.py:862] Epoch 16, batch 12000, loss[loss=0.1732, simple_loss=0.3983, pruned_loss=0.08897, ctc_loss=0.1239, over 14659.00 frames. ], tot_loss[loss=0.1856, simple_loss=0.4367, pruned_loss=0.09032, ctc_loss=0.1329, over 2871265.83 frames. ], batch size: 35, lr: 3.00e-03, 
2023-01-05 21:12:35,489 INFO [train.py:862] Epoch 16, batch 12500, loss[loss=0.2275, simple_loss=0.452, pruned_loss=0.1166, ctc_loss=0.1781, over 14727.00 frames. ], tot_loss[loss=0.187, simple_loss=0.4369, pruned_loss=0.09141, ctc_loss=0.1344, over 2856726.61 frames. ], batch size: 38, lr: 3.00e-03, 
2023-01-05 21:13:41,511 INFO [train.py:862] Epoch 16, batch 13000, loss[loss=0.1594, simple_loss=0.417, pruned_loss=0.0759, ctc_loss=0.1058, over 14718.00 frames. ], tot_loss[loss=0.1848, simple_loss=0.4353, pruned_loss=0.09002, ctc_loss=0.1322, over 2854616.07 frames. ], batch size: 37, lr: 3.00e-03, 
2023-01-05 21:14:47,016 INFO [train.py:862] Epoch 16, batch 13500, loss[loss=0.2154, simple_loss=0.4525, pruned_loss=0.1103, ctc_loss=0.1634, over 13645.00 frames. ], tot_loss[loss=0.1862, simple_loss=0.4372, pruned_loss=0.09077, ctc_loss=0.1334, over 2841149.20 frames. ], batch size: 63, lr: 3.00e-03, 
2023-01-05 21:15:53,074 INFO [train.py:862] Epoch 16, batch 14000, loss[loss=0.1638, simple_loss=0.408, pruned_loss=0.08032, ctc_loss=0.1121, over 14779.00 frames. ], tot_loss[loss=0.1841, simple_loss=0.435, pruned_loss=0.08932, ctc_loss=0.1315, over 2857207.62 frames. ], batch size: 36, lr: 3.00e-03, 
2023-01-05 21:16:58,440 INFO [train.py:862] Epoch 16, batch 14500, loss[loss=0.1629, simple_loss=0.4294, pruned_loss=0.07395, ctc_loss=0.109, over 14807.00 frames. ], tot_loss[loss=0.1848, simple_loss=0.4355, pruned_loss=0.09009, ctc_loss=0.1321, over 2857398.75 frames. ], batch size: 43, lr: 3.00e-03, 
2023-01-05 21:18:03,934 INFO [train.py:862] Epoch 16, batch 15000, loss[loss=0.1705, simple_loss=0.4539, pruned_loss=0.07508, ctc_loss=0.1141, over 14505.00 frames. ], tot_loss[loss=0.1834, simple_loss=0.4349, pruned_loss=0.08895, ctc_loss=0.1306, over 2857244.09 frames. ], batch size: 50, lr: 3.00e-03, 
2023-01-05 21:19:09,872 INFO [train.py:862] Epoch 16, batch 15500, loss[loss=0.1641, simple_loss=0.4015, pruned_loss=0.08166, ctc_loss=0.1133, over 14422.00 frames. ], tot_loss[loss=0.1862, simple_loss=0.4369, pruned_loss=0.09084, ctc_loss=0.1334, over 2861213.79 frames. ], batch size: 32, lr: 3.00e-03, 
2023-01-05 21:20:15,548 INFO [train.py:862] Epoch 16, batch 16000, loss[loss=0.1721, simple_loss=0.4577, pruned_loss=0.08037, ctc_loss=0.1133, over 14665.00 frames. ], tot_loss[loss=0.1844, simple_loss=0.4351, pruned_loss=0.08974, ctc_loss=0.1317, over 2858296.92 frames. ], batch size: 42, lr: 3.00e-03, 
2023-01-05 21:21:21,492 INFO [train.py:862] Epoch 16, batch 16500, loss[loss=0.2082, simple_loss=0.4518, pruned_loss=0.1041, ctc_loss=0.156, over 14835.00 frames. ], tot_loss[loss=0.1831, simple_loss=0.4346, pruned_loss=0.08895, ctc_loss=0.1303, over 2865989.29 frames. ], batch size: 42, lr: 3.00e-03, 
2023-01-05 21:21:54,846 INFO [checkpoint.py:75] Saving checkpoint to tiny_transducer_ctc/exp_2m_phone_halfdelay/checkpoint-290000.pt
2023-01-05 21:22:26,768 INFO [train.py:862] Epoch 16, batch 17000, loss[loss=0.1559, simple_loss=0.3944, pruned_loss=0.0746, ctc_loss=0.1063, over 14435.00 frames. ], tot_loss[loss=0.1854, simple_loss=0.4362, pruned_loss=0.09042, ctc_loss=0.1327, over 2873734.08 frames. ], batch size: 33, lr: 3.00e-03, 
2023-01-05 21:23:32,395 INFO [train.py:862] Epoch 16, batch 17500, loss[loss=0.1769, simple_loss=0.4436, pruned_loss=0.08516, ctc_loss=0.1212, over 14750.00 frames. ], tot_loss[loss=0.1831, simple_loss=0.4333, pruned_loss=0.08894, ctc_loss=0.1306, over 2860291.74 frames. ], batch size: 45, lr: 3.00e-03, 
2023-01-05 21:24:36,682 INFO [train.py:862] Epoch 16, batch 18000, loss[loss=0.2194, simple_loss=0.4474, pruned_loss=0.1222, ctc_loss=0.1652, over 14514.00 frames. ], tot_loss[loss=0.1838, simple_loss=0.4358, pruned_loss=0.0893, ctc_loss=0.1309, over 2849111.26 frames. ], batch size: 37, lr: 3.00e-03, 
2023-01-05 21:24:36,683 INFO [train.py:887] Computing validation loss
2023-01-05 21:25:02,230 INFO [train.py:897] Epoch 16, validation: loss=0.2001, simple_loss=0.4512, pruned_loss=0.1, ctc_loss=0.1464, over 944034.00 frames. 
2023-01-05 21:25:02,230 INFO [train.py:898] Maximum memory allocated so far is 7283MB
2023-01-05 21:25:29,248 INFO [checkpoint.py:75] Saving checkpoint to tiny_transducer_ctc/exp_2m_phone_halfdelay/epoch-16.pt
2023-01-05 21:25:32,022 INFO [train.py:862] Epoch 17, batch 0, loss[loss=0.2352, simple_loss=0.4901, pruned_loss=0.1213, ctc_loss=0.1789, over 14583.00 frames. ], tot_loss[loss=0.2352, simple_loss=0.4901, pruned_loss=0.1213, ctc_loss=0.1789, over 14583.00 frames. ], batch size: 43, lr: 3.00e-03, 
2023-01-05 21:26:37,752 INFO [train.py:862] Epoch 17, batch 500, loss[loss=0.1846, simple_loss=0.4152, pruned_loss=0.0917, ctc_loss=0.1355, over 14715.00 frames. ], tot_loss[loss=0.1848, simple_loss=0.4369, pruned_loss=0.08982, ctc_loss=0.1319, over 2630983.84 frames. ], batch size: 37, lr: 3.00e-03, 
2023-01-05 21:27:42,978 INFO [train.py:862] Epoch 17, batch 1000, loss[loss=0.1835, simple_loss=0.4083, pruned_loss=0.09109, ctc_loss=0.1355, over 14706.00 frames. ], tot_loss[loss=0.1865, simple_loss=0.4375, pruned_loss=0.0909, ctc_loss=0.1337, over 2860693.95 frames. ], batch size: 38, lr: 3.00e-03, 
2023-01-05 21:28:48,985 INFO [train.py:862] Epoch 17, batch 1500, loss[loss=0.2236, simple_loss=0.487, pruned_loss=0.1133, ctc_loss=0.1665, over 14667.00 frames. ], tot_loss[loss=0.1847, simple_loss=0.4352, pruned_loss=0.08974, ctc_loss=0.1322, over 2870111.58 frames. ], batch size: 42, lr: 3.00e-03, 
2023-01-05 21:29:54,150 INFO [train.py:862] Epoch 17, batch 2000, loss[loss=0.173, simple_loss=0.419, pruned_loss=0.08256, ctc_loss=0.122, over 14504.00 frames. ], tot_loss[loss=0.1826, simple_loss=0.435, pruned_loss=0.08835, ctc_loss=0.1298, over 2882520.50 frames. ], batch size: 37, lr: 3.00e-03, 
2023-01-05 21:30:59,244 INFO [train.py:862] Epoch 17, batch 2500, loss[loss=0.1625, simple_loss=0.3834, pruned_loss=0.08063, ctc_loss=0.1154, over 14703.00 frames. ], tot_loss[loss=0.1825, simple_loss=0.4352, pruned_loss=0.08827, ctc_loss=0.1296, over 2865345.49 frames. ], batch size: 35, lr: 3.00e-03, 
2023-01-05 21:32:04,943 INFO [train.py:862] Epoch 17, batch 3000, loss[loss=0.1825, simple_loss=0.4245, pruned_loss=0.08851, ctc_loss=0.1318, over 14508.00 frames. ], tot_loss[loss=0.1838, simple_loss=0.4348, pruned_loss=0.08916, ctc_loss=0.1312, over 2865693.38 frames. ], batch size: 37, lr: 3.00e-03, 
2023-01-05 21:33:10,250 INFO [train.py:862] Epoch 17, batch 3500, loss[loss=0.2163, simple_loss=0.4488, pruned_loss=0.1153, ctc_loss=0.1635, over 14880.00 frames. ], tot_loss[loss=0.1841, simple_loss=0.4341, pruned_loss=0.08944, ctc_loss=0.1317, over 2883213.62 frames. ], batch size: 39, lr: 3.00e-03, 
2023-01-05 21:34:15,493 INFO [train.py:862] Epoch 17, batch 4000, loss[loss=0.2142, simple_loss=0.4598, pruned_loss=0.1092, ctc_loss=0.1606, over 13682.00 frames. ], tot_loss[loss=0.1838, simple_loss=0.4356, pruned_loss=0.08943, ctc_loss=0.1309, over 2874131.27 frames. ], batch size: 63, lr: 3.00e-03, 
2023-01-05 21:35:21,154 INFO [train.py:862] Epoch 17, batch 4500, loss[loss=0.1591, simple_loss=0.398, pruned_loss=0.07162, ctc_loss=0.1114, over 13981.00 frames. ], tot_loss[loss=0.1815, simple_loss=0.4322, pruned_loss=0.0881, ctc_loss=0.1289, over 2881665.88 frames. ], batch size: 31, lr: 3.00e-03, 
2023-01-05 21:36:26,285 INFO [train.py:862] Epoch 17, batch 5000, loss[loss=0.1987, simple_loss=0.4644, pruned_loss=0.09888, ctc_loss=0.142, over 14518.00 frames. ], tot_loss[loss=0.1841, simple_loss=0.4357, pruned_loss=0.08952, ctc_loss=0.1313, over 2886318.04 frames. ], batch size: 50, lr: 3.00e-03, 
2023-01-05 21:37:31,343 INFO [train.py:862] Epoch 17, batch 5500, loss[loss=0.1613, simple_loss=0.4275, pruned_loss=0.07815, ctc_loss=0.1054, over 14670.00 frames. ], tot_loss[loss=0.1836, simple_loss=0.4364, pruned_loss=0.08895, ctc_loss=0.1306, over 2867888.46 frames. ], batch size: 40, lr: 3.00e-03, 
2023-01-05 21:38:36,980 INFO [train.py:862] Epoch 17, batch 6000, loss[loss=0.1877, simple_loss=0.4488, pruned_loss=0.09301, ctc_loss=0.1321, over 14824.00 frames. ], tot_loss[loss=0.1835, simple_loss=0.4342, pruned_loss=0.08935, ctc_loss=0.1308, over 2858804.48 frames. ], batch size: 41, lr: 3.00e-03, 
2023-01-05 21:39:41,891 INFO [train.py:862] Epoch 17, batch 6500, loss[loss=0.234, simple_loss=0.483, pruned_loss=0.1179, ctc_loss=0.1802, over 14814.00 frames. ], tot_loss[loss=0.1833, simple_loss=0.4355, pruned_loss=0.08894, ctc_loss=0.1305, over 2871222.19 frames. ], batch size: 43, lr: 3.00e-03, 
2023-01-05 21:40:47,134 INFO [train.py:862] Epoch 17, batch 7000, loss[loss=0.2079, simple_loss=0.4366, pruned_loss=0.1032, ctc_loss=0.1592, over 14655.00 frames. ], tot_loss[loss=0.1847, simple_loss=0.4349, pruned_loss=0.09034, ctc_loss=0.132, over 2845072.37 frames. ], batch size: 42, lr: 3.00e-03, 
2023-01-05 21:41:52,281 INFO [train.py:862] Epoch 17, batch 7500, loss[loss=0.2013, simple_loss=0.4305, pruned_loss=0.1049, ctc_loss=0.1504, over 14713.00 frames. ], tot_loss[loss=0.1836, simple_loss=0.4351, pruned_loss=0.08903, ctc_loss=0.131, over 2878112.61 frames. ], batch size: 37, lr: 3.00e-03, 
2023-01-05 21:42:57,299 INFO [train.py:862] Epoch 17, batch 8000, loss[loss=0.1786, simple_loss=0.436, pruned_loss=0.0791, ctc_loss=0.1278, over 14719.00 frames. ], tot_loss[loss=0.1816, simple_loss=0.4328, pruned_loss=0.08789, ctc_loss=0.1291, over 2872552.54 frames. ], batch size: 46, lr: 3.00e-03, 
2023-01-05 21:44:02,817 INFO [train.py:862] Epoch 17, batch 8500, loss[loss=0.187, simple_loss=0.3957, pruned_loss=0.0959, ctc_loss=0.1412, over 14521.00 frames. ], tot_loss[loss=0.1851, simple_loss=0.4359, pruned_loss=0.09018, ctc_loss=0.1324, over 2874678.46 frames. ], batch size: 34, lr: 3.00e-03, 
2023-01-05 21:44:08,103 INFO [checkpoint.py:75] Saving checkpoint to tiny_transducer_ctc/exp_2m_phone_halfdelay/checkpoint-300000.pt
2023-01-05 21:45:07,951 INFO [train.py:862] Epoch 17, batch 9000, loss[loss=0.1543, simple_loss=0.3913, pruned_loss=0.0765, ctc_loss=0.1037, over 14426.00 frames. ], tot_loss[loss=0.1817, simple_loss=0.4335, pruned_loss=0.08795, ctc_loss=0.129, over 2862748.17 frames. ], batch size: 33, lr: 3.00e-03, 
2023-01-05 21:45:07,952 INFO [train.py:887] Computing validation loss
2023-01-05 21:45:33,339 INFO [train.py:897] Epoch 17, validation: loss=0.1982, simple_loss=0.4498, pruned_loss=0.09859, ctc_loss=0.1445, over 944034.00 frames. 
2023-01-05 21:45:33,339 INFO [train.py:898] Maximum memory allocated so far is 7283MB
2023-01-05 21:46:38,903 INFO [train.py:862] Epoch 17, batch 9500, loss[loss=0.2016, simple_loss=0.4424, pruned_loss=0.09776, ctc_loss=0.1512, over 14662.00 frames. ], tot_loss[loss=0.184, simple_loss=0.4343, pruned_loss=0.08958, ctc_loss=0.1314, over 2864378.56 frames. ], batch size: 40, lr: 3.00e-03, 
2023-01-05 21:47:44,480 INFO [train.py:862] Epoch 17, batch 10000, loss[loss=0.2133, simple_loss=0.4687, pruned_loss=0.1063, ctc_loss=0.1588, over 14536.00 frames. ], tot_loss[loss=0.1868, simple_loss=0.4373, pruned_loss=0.09119, ctc_loss=0.1341, over 2831779.91 frames. ], batch size: 48, lr: 3.00e-03, 
2023-01-05 21:48:50,075 INFO [train.py:862] Epoch 17, batch 10500, loss[loss=0.2206, simple_loss=0.4778, pruned_loss=0.1132, ctc_loss=0.1643, over 14841.00 frames. ], tot_loss[loss=0.1858, simple_loss=0.4365, pruned_loss=0.09047, ctc_loss=0.133, over 2867514.40 frames. ], batch size: 42, lr: 3.00e-03, 
2023-01-05 21:49:55,059 INFO [train.py:862] Epoch 17, batch 11000, loss[loss=0.2254, simple_loss=0.4406, pruned_loss=0.1264, ctc_loss=0.1734, over 13993.00 frames. ], tot_loss[loss=0.1843, simple_loss=0.4353, pruned_loss=0.08971, ctc_loss=0.1316, over 2866065.74 frames. ], batch size: 31, lr: 3.00e-03, 
2023-01-05 21:51:00,492 INFO [train.py:862] Epoch 17, batch 11500, loss[loss=0.1777, simple_loss=0.4375, pruned_loss=0.08271, ctc_loss=0.1246, over 14674.00 frames. ], tot_loss[loss=0.185, simple_loss=0.4363, pruned_loss=0.09016, ctc_loss=0.1322, over 2850198.78 frames. ], batch size: 40, lr: 3.00e-03, 
2023-01-05 21:52:05,591 INFO [train.py:862] Epoch 17, batch 12000, loss[loss=0.1913, simple_loss=0.437, pruned_loss=0.09708, ctc_loss=0.138, over 13770.00 frames. ], tot_loss[loss=0.1849, simple_loss=0.4357, pruned_loss=0.09006, ctc_loss=0.1321, over 2841765.77 frames. ], batch size: 63, lr: 3.00e-03, 
2023-01-05 21:53:10,585 INFO [train.py:862] Epoch 17, batch 12500, loss[loss=0.2391, simple_loss=0.4914, pruned_loss=0.124, ctc_loss=0.1832, over 14547.00 frames. ], tot_loss[loss=0.1845, simple_loss=0.4363, pruned_loss=0.0896, ctc_loss=0.1317, over 2866084.13 frames. ], batch size: 50, lr: 3.00e-03, 
2023-01-05 21:54:16,379 INFO [train.py:862] Epoch 17, batch 13000, loss[loss=0.1728, simple_loss=0.4275, pruned_loss=0.07692, ctc_loss=0.1222, over 14669.00 frames. ], tot_loss[loss=0.1823, simple_loss=0.4333, pruned_loss=0.08829, ctc_loss=0.1297, over 2875739.36 frames. ], batch size: 40, lr: 3.00e-03, 
2023-01-05 21:55:21,420 INFO [train.py:862] Epoch 17, batch 13500, loss[loss=0.1863, simple_loss=0.4215, pruned_loss=0.0959, ctc_loss=0.1348, over 14654.00 frames. ], tot_loss[loss=0.1839, simple_loss=0.4356, pruned_loss=0.08916, ctc_loss=0.1311, over 2858005.16 frames. ], batch size: 40, lr: 3.00e-03, 
2023-01-05 21:56:26,987 INFO [train.py:862] Epoch 17, batch 14000, loss[loss=0.1817, simple_loss=0.4225, pruned_loss=0.08979, ctc_loss=0.1305, over 14831.00 frames. ], tot_loss[loss=0.1836, simple_loss=0.4352, pruned_loss=0.08906, ctc_loss=0.1309, over 2849640.83 frames. ], batch size: 41, lr: 3.00e-03, 
2023-01-05 21:57:31,951 INFO [train.py:862] Epoch 17, batch 14500, loss[loss=0.1794, simple_loss=0.4333, pruned_loss=0.0854, ctc_loss=0.1268, over 12807.00 frames. ], tot_loss[loss=0.1856, simple_loss=0.4359, pruned_loss=0.09063, ctc_loss=0.1329, over 2851474.27 frames. ], batch size: 76, lr: 3.00e-03, 
2023-01-05 21:58:37,392 INFO [train.py:862] Epoch 17, batch 15000, loss[loss=0.1569, simple_loss=0.4258, pruned_loss=0.06862, ctc_loss=0.1035, over 14843.00 frames. ], tot_loss[loss=0.1836, simple_loss=0.4342, pruned_loss=0.08905, ctc_loss=0.1311, over 2866264.15 frames. ], batch size: 41, lr: 3.00e-03, 
2023-01-05 21:59:42,918 INFO [train.py:862] Epoch 17, batch 15500, loss[loss=0.1449, simple_loss=0.4054, pruned_loss=0.06464, ctc_loss=0.09249, over 14701.00 frames. ], tot_loss[loss=0.1832, simple_loss=0.4348, pruned_loss=0.08876, ctc_loss=0.1305, over 2872994.81 frames. ], batch size: 39, lr: 3.00e-03, 
2023-01-05 22:00:47,957 INFO [train.py:862] Epoch 17, batch 16000, loss[loss=0.1732, simple_loss=0.3948, pruned_loss=0.08612, ctc_loss=0.1259, over 14673.00 frames. ], tot_loss[loss=0.1816, simple_loss=0.4333, pruned_loss=0.08787, ctc_loss=0.1289, over 2863604.31 frames. ], batch size: 35, lr: 3.00e-03, 
2023-01-05 22:01:53,323 INFO [train.py:862] Epoch 17, batch 16500, loss[loss=0.1532, simple_loss=0.4206, pruned_loss=0.06639, ctc_loss=0.1002, over 14524.00 frames. ], tot_loss[loss=0.1823, simple_loss=0.4344, pruned_loss=0.08816, ctc_loss=0.1295, over 2853825.05 frames. ], batch size: 36, lr: 3.00e-03, 
2023-01-05 22:02:58,667 INFO [train.py:862] Epoch 17, batch 17000, loss[loss=0.1802, simple_loss=0.41, pruned_loss=0.09003, ctc_loss=0.1309, over 14110.00 frames. ], tot_loss[loss=0.1833, simple_loss=0.4337, pruned_loss=0.08893, ctc_loss=0.1308, over 2860681.98 frames. ], batch size: 31, lr: 3.00e-03, 
2023-01-05 22:04:04,080 INFO [train.py:862] Epoch 17, batch 17500, loss[loss=0.1814, simple_loss=0.4361, pruned_loss=0.08605, ctc_loss=0.1288, over 14686.00 frames. ], tot_loss[loss=0.1837, simple_loss=0.4354, pruned_loss=0.0892, ctc_loss=0.1309, over 2848385.53 frames. ], batch size: 44, lr: 3.00e-03, 
2023-01-05 22:05:07,967 INFO [train.py:862] Epoch 17, batch 18000, loss[loss=0.1783, simple_loss=0.4335, pruned_loss=0.08862, ctc_loss=0.1238, over 14681.00 frames. ], tot_loss[loss=0.1821, simple_loss=0.4331, pruned_loss=0.08811, ctc_loss=0.1296, over 2860670.16 frames. ], batch size: 39, lr: 3.00e-03, 
2023-01-05 22:05:07,968 INFO [train.py:887] Computing validation loss
2023-01-05 22:05:33,298 INFO [train.py:897] Epoch 17, validation: loss=0.2005, simple_loss=0.4509, pruned_loss=0.1006, ctc_loss=0.1467, over 944034.00 frames. 
2023-01-05 22:05:33,298 INFO [train.py:898] Maximum memory allocated so far is 7283MB
2023-01-05 22:06:00,739 INFO [checkpoint.py:75] Saving checkpoint to tiny_transducer_ctc/exp_2m_phone_halfdelay/epoch-17.pt
2023-01-05 22:06:03,450 INFO [train.py:862] Epoch 18, batch 0, loss[loss=0.2436, simple_loss=0.4674, pruned_loss=0.1306, ctc_loss=0.1919, over 14817.00 frames. ], tot_loss[loss=0.2436, simple_loss=0.4674, pruned_loss=0.1306, ctc_loss=0.1919, over 14817.00 frames. ], batch size: 41, lr: 3.00e-03, 
2023-01-05 22:06:45,864 INFO [checkpoint.py:75] Saving checkpoint to tiny_transducer_ctc/exp_2m_phone_halfdelay/checkpoint-310000.pt
2023-01-05 22:07:08,796 INFO [train.py:862] Epoch 18, batch 500, loss[loss=0.1591, simple_loss=0.4105, pruned_loss=0.07532, ctc_loss=0.107, over 14714.00 frames. ], tot_loss[loss=0.1855, simple_loss=0.4373, pruned_loss=0.09036, ctc_loss=0.1326, over 2646740.10 frames. ], batch size: 38, lr: 3.00e-03, 
2023-01-05 22:08:13,768 INFO [train.py:862] Epoch 18, batch 1000, loss[loss=0.1428, simple_loss=0.3819, pruned_loss=0.06509, ctc_loss=0.09428, over 14765.00 frames. ], tot_loss[loss=0.1845, simple_loss=0.4352, pruned_loss=0.0897, ctc_loss=0.1318, over 2850992.32 frames. ], batch size: 36, lr: 3.00e-03, 
2023-01-05 22:09:19,600 INFO [train.py:862] Epoch 18, batch 1500, loss[loss=0.1596, simple_loss=0.4076, pruned_loss=0.07372, ctc_loss=0.109, over 14700.00 frames. ], tot_loss[loss=0.1843, simple_loss=0.4355, pruned_loss=0.08939, ctc_loss=0.1316, over 2866217.75 frames. ], batch size: 39, lr: 3.00e-03, 
2023-01-05 22:10:24,601 INFO [train.py:862] Epoch 18, batch 2000, loss[loss=0.1298, simple_loss=0.372, pruned_loss=0.05412, ctc_loss=0.08251, over 14655.00 frames. ], tot_loss[loss=0.1804, simple_loss=0.4326, pruned_loss=0.08691, ctc_loss=0.1278, over 2859006.39 frames. ], batch size: 35, lr: 3.00e-03, 
2023-01-05 22:11:29,948 INFO [train.py:862] Epoch 18, batch 2500, loss[loss=0.1393, simple_loss=0.3838, pruned_loss=0.05992, ctc_loss=0.09101, over 14524.00 frames. ], tot_loss[loss=0.182, simple_loss=0.4335, pruned_loss=0.08802, ctc_loss=0.1293, over 2864432.40 frames. ], batch size: 38, lr: 3.00e-03, 
2023-01-05 22:12:35,646 INFO [train.py:862] Epoch 18, batch 3000, loss[loss=0.1758, simple_loss=0.4505, pruned_loss=0.07671, ctc_loss=0.1218, over 14787.00 frames. ], tot_loss[loss=0.1836, simple_loss=0.4343, pruned_loss=0.08922, ctc_loss=0.131, over 2869465.43 frames. ], batch size: 43, lr: 3.00e-03, 
2023-01-05 22:13:41,096 INFO [train.py:862] Epoch 18, batch 3500, loss[loss=0.1856, simple_loss=0.4593, pruned_loss=0.08867, ctc_loss=0.1287, over 14741.00 frames. ], tot_loss[loss=0.1829, simple_loss=0.4357, pruned_loss=0.08863, ctc_loss=0.1299, over 2874529.50 frames. ], batch size: 45, lr: 3.00e-03, 
2023-01-05 22:14:46,490 INFO [train.py:862] Epoch 18, batch 4000, loss[loss=0.1652, simple_loss=0.3949, pruned_loss=0.07858, ctc_loss=0.1178, over 14689.00 frames. ], tot_loss[loss=0.1839, simple_loss=0.4331, pruned_loss=0.08943, ctc_loss=0.1316, over 2871006.19 frames. ], batch size: 35, lr: 3.00e-03, 
2023-01-05 22:15:52,616 INFO [train.py:862] Epoch 18, batch 4500, loss[loss=0.2215, simple_loss=0.4749, pruned_loss=0.1082, ctc_loss=0.1683, over 14516.00 frames. ], tot_loss[loss=0.1831, simple_loss=0.4339, pruned_loss=0.08887, ctc_loss=0.1305, over 2873390.38 frames. ], batch size: 50, lr: 3.00e-03, 
2023-01-05 22:16:58,132 INFO [train.py:862] Epoch 18, batch 5000, loss[loss=0.1824, simple_loss=0.4376, pruned_loss=0.08789, ctc_loss=0.1291, over 14671.00 frames. ], tot_loss[loss=0.1812, simple_loss=0.4338, pruned_loss=0.08755, ctc_loss=0.1284, over 2881720.90 frames. ], batch size: 42, lr: 3.00e-03, 
2023-01-05 22:18:03,374 INFO [train.py:862] Epoch 18, batch 5500, loss[loss=0.1867, simple_loss=0.407, pruned_loss=0.09799, ctc_loss=0.1376, over 14662.00 frames. ], tot_loss[loss=0.1834, simple_loss=0.435, pruned_loss=0.08907, ctc_loss=0.1307, over 2869388.77 frames. ], batch size: 35, lr: 3.00e-03, 
2023-01-05 22:19:09,214 INFO [train.py:862] Epoch 18, batch 6000, loss[loss=0.1495, simple_loss=0.3879, pruned_loss=0.06948, ctc_loss=0.1007, over 14026.00 frames. ], tot_loss[loss=0.1827, simple_loss=0.4346, pruned_loss=0.08853, ctc_loss=0.1299, over 2867375.88 frames. ], batch size: 31, lr: 3.00e-03, 
2023-01-05 22:20:14,807 INFO [train.py:862] Epoch 18, batch 6500, loss[loss=0.2124, simple_loss=0.4551, pruned_loss=0.1004, ctc_loss=0.1628, over 14011.00 frames. ], tot_loss[loss=0.1814, simple_loss=0.433, pruned_loss=0.0877, ctc_loss=0.1288, over 2867884.03 frames. ], batch size: 56, lr: 3.00e-03, 
2023-01-05 22:21:20,456 INFO [train.py:862] Epoch 18, batch 7000, loss[loss=0.2128, simple_loss=0.4767, pruned_loss=0.1039, ctc_loss=0.1573, over 14806.00 frames. ], tot_loss[loss=0.1852, simple_loss=0.4359, pruned_loss=0.09017, ctc_loss=0.1325, over 2869073.50 frames. ], batch size: 43, lr: 3.00e-03, 
2023-01-05 22:22:25,663 INFO [train.py:862] Epoch 18, batch 7500, loss[loss=0.1774, simple_loss=0.4235, pruned_loss=0.08482, ctc_loss=0.1263, over 14835.00 frames. ], tot_loss[loss=0.1834, simple_loss=0.4346, pruned_loss=0.0891, ctc_loss=0.1306, over 2854737.73 frames. ], batch size: 41, lr: 3.00e-03, 
2023-01-05 22:23:31,071 INFO [train.py:862] Epoch 18, batch 8000, loss[loss=0.1939, simple_loss=0.4519, pruned_loss=0.0953, ctc_loss=0.1393, over 13535.00 frames. ], tot_loss[loss=0.1844, simple_loss=0.4362, pruned_loss=0.08952, ctc_loss=0.1316, over 2866030.28 frames. ], batch size: 63, lr: 3.00e-03, 
2023-01-05 22:24:36,918 INFO [train.py:862] Epoch 18, batch 8500, loss[loss=0.1654, simple_loss=0.4159, pruned_loss=0.08154, ctc_loss=0.1123, over 14710.00 frames. ], tot_loss[loss=0.1826, simple_loss=0.4337, pruned_loss=0.0884, ctc_loss=0.1301, over 2852469.45 frames. ], batch size: 38, lr: 3.00e-03, 
2023-01-05 22:25:41,899 INFO [train.py:862] Epoch 18, batch 9000, loss[loss=0.1517, simple_loss=0.3865, pruned_loss=0.07283, ctc_loss=0.1026, over 14519.00 frames. ], tot_loss[loss=0.1808, simple_loss=0.4322, pruned_loss=0.08713, ctc_loss=0.1284, over 2859376.69 frames. ], batch size: 34, lr: 3.00e-03, 
2023-01-05 22:25:41,899 INFO [train.py:887] Computing validation loss
2023-01-05 22:26:08,163 INFO [train.py:897] Epoch 18, validation: loss=0.1985, simple_loss=0.45, pruned_loss=0.09916, ctc_loss=0.1446, over 944034.00 frames. 
2023-01-05 22:26:08,164 INFO [train.py:898] Maximum memory allocated so far is 7283MB
2023-01-05 22:27:14,795 INFO [train.py:862] Epoch 18, batch 9500, loss[loss=0.1734, simple_loss=0.432, pruned_loss=0.07948, ctc_loss=0.1211, over 14819.00 frames. ], tot_loss[loss=0.1809, simple_loss=0.4329, pruned_loss=0.08734, ctc_loss=0.1282, over 2865033.28 frames. ], batch size: 42, lr: 3.00e-03, 
2023-01-05 22:28:21,601 INFO [train.py:862] Epoch 18, batch 10000, loss[loss=0.1947, simple_loss=0.4509, pruned_loss=0.09413, ctc_loss=0.1412, over 13652.00 frames. ], tot_loss[loss=0.1839, simple_loss=0.435, pruned_loss=0.08944, ctc_loss=0.1312, over 2864459.71 frames. ], batch size: 63, lr: 3.00e-03, 
2023-01-05 22:29:05,239 INFO [checkpoint.py:75] Saving checkpoint to tiny_transducer_ctc/exp_2m_phone_halfdelay/checkpoint-320000.pt
2023-01-05 22:29:28,821 INFO [train.py:862] Epoch 18, batch 10500, loss[loss=0.2053, simple_loss=0.4685, pruned_loss=0.1011, ctc_loss=0.1496, over 13674.00 frames. ], tot_loss[loss=0.181, simple_loss=0.4339, pruned_loss=0.08728, ctc_loss=0.1282, over 2848010.89 frames. ], batch size: 63, lr: 3.00e-03, 
2023-01-05 22:30:35,771 INFO [train.py:862] Epoch 18, batch 11000, loss[loss=0.1507, simple_loss=0.3835, pruned_loss=0.07208, ctc_loss=0.1022, over 14534.00 frames. ], tot_loss[loss=0.1828, simple_loss=0.4331, pruned_loss=0.08865, ctc_loss=0.1303, over 2861464.78 frames. ], batch size: 34, lr: 3.00e-03, 
2023-01-05 22:31:43,053 INFO [train.py:862] Epoch 18, batch 11500, loss[loss=0.1855, simple_loss=0.4494, pruned_loss=0.0915, ctc_loss=0.1294, over 14488.00 frames. ], tot_loss[loss=0.182, simple_loss=0.434, pruned_loss=0.08805, ctc_loss=0.1293, over 2860928.86 frames. ], batch size: 50, lr: 3.00e-03, 
2023-01-05 22:32:50,244 INFO [train.py:862] Epoch 18, batch 12000, loss[loss=0.156, simple_loss=0.4285, pruned_loss=0.07231, ctc_loss=0.1001, over 14820.00 frames. ], tot_loss[loss=0.182, simple_loss=0.4331, pruned_loss=0.088, ctc_loss=0.1295, over 2872255.94 frames. ], batch size: 41, lr: 3.00e-03, 
2023-01-05 22:33:57,524 INFO [train.py:862] Epoch 18, batch 12500, loss[loss=0.1863, simple_loss=0.4562, pruned_loss=0.08951, ctc_loss=0.1301, over 14683.00 frames. ], tot_loss[loss=0.1838, simple_loss=0.4358, pruned_loss=0.08912, ctc_loss=0.131, over 2865193.78 frames. ], batch size: 44, lr: 3.00e-03, 
2023-01-05 22:35:04,179 INFO [train.py:862] Epoch 18, batch 13000, loss[loss=0.182, simple_loss=0.4448, pruned_loss=0.08546, ctc_loss=0.1281, over 14563.00 frames. ], tot_loss[loss=0.1837, simple_loss=0.4358, pruned_loss=0.08891, ctc_loss=0.1309, over 2868252.11 frames. ], batch size: 48, lr: 3.00e-03, 
2023-01-05 22:36:10,761 INFO [train.py:862] Epoch 18, batch 13500, loss[loss=0.2157, simple_loss=0.4643, pruned_loss=0.1113, ctc_loss=0.1609, over 14845.00 frames. ], tot_loss[loss=0.1839, simple_loss=0.4352, pruned_loss=0.08918, ctc_loss=0.1312, over 2854582.99 frames. ], batch size: 42, lr: 3.00e-03, 
2023-01-05 22:37:17,939 INFO [train.py:862] Epoch 18, batch 14000, loss[loss=0.1406, simple_loss=0.375, pruned_loss=0.06193, ctc_loss=0.09403, over 14797.00 frames. ], tot_loss[loss=0.1832, simple_loss=0.4352, pruned_loss=0.08836, ctc_loss=0.1306, over 2853012.69 frames. ], batch size: 36, lr: 3.00e-03, 
2023-01-05 22:38:24,957 INFO [train.py:862] Epoch 18, batch 14500, loss[loss=0.2006, simple_loss=0.463, pruned_loss=0.09604, ctc_loss=0.1462, over 14721.00 frames. ], tot_loss[loss=0.1805, simple_loss=0.4325, pruned_loss=0.08696, ctc_loss=0.1279, over 2855644.76 frames. ], batch size: 46, lr: 3.00e-03, 
2023-01-05 22:39:31,894 INFO [train.py:862] Epoch 18, batch 15000, loss[loss=0.2399, simple_loss=0.4576, pruned_loss=0.1327, ctc_loss=0.1878, over 14825.00 frames. ], tot_loss[loss=0.1839, simple_loss=0.4354, pruned_loss=0.08924, ctc_loss=0.1312, over 2860382.61 frames. ], batch size: 41, lr: 3.00e-03, 
2023-01-05 22:40:38,657 INFO [train.py:862] Epoch 18, batch 15500, loss[loss=0.1735, simple_loss=0.4015, pruned_loss=0.08455, ctc_loss=0.1256, over 14536.00 frames. ], tot_loss[loss=0.1823, simple_loss=0.434, pruned_loss=0.08848, ctc_loss=0.1295, over 2875113.19 frames. ], batch size: 36, lr: 3.00e-03, 
2023-01-05 22:41:45,297 INFO [train.py:862] Epoch 18, batch 16000, loss[loss=0.2319, simple_loss=0.4288, pruned_loss=0.1261, ctc_loss=0.1854, over 14536.00 frames. ], tot_loss[loss=0.1827, simple_loss=0.4342, pruned_loss=0.08866, ctc_loss=0.13, over 2872672.34 frames. ], batch size: 34, lr: 3.00e-03, 
2023-01-05 22:42:52,890 INFO [train.py:862] Epoch 18, batch 16500, loss[loss=0.1757, simple_loss=0.421, pruned_loss=0.08532, ctc_loss=0.1242, over 14796.00 frames. ], tot_loss[loss=0.1816, simple_loss=0.4342, pruned_loss=0.08761, ctc_loss=0.1289, over 2859103.66 frames. ], batch size: 36, lr: 3.00e-03, 
2023-01-05 22:43:59,508 INFO [train.py:862] Epoch 18, batch 17000, loss[loss=0.1878, simple_loss=0.4278, pruned_loss=0.08723, ctc_loss=0.1392, over 12889.00 frames. ], tot_loss[loss=0.1811, simple_loss=0.4326, pruned_loss=0.08751, ctc_loss=0.1285, over 2868083.68 frames. ], batch size: 76, lr: 3.00e-03, 
2023-01-05 22:45:06,780 INFO [train.py:862] Epoch 18, batch 17500, loss[loss=0.2068, simple_loss=0.4185, pruned_loss=0.1119, ctc_loss=0.1578, over 14033.00 frames. ], tot_loss[loss=0.1804, simple_loss=0.4328, pruned_loss=0.08724, ctc_loss=0.1276, over 2862820.18 frames. ], batch size: 31, lr: 3.00e-03, 
2023-01-05 22:46:11,669 INFO [train.py:862] Epoch 18, batch 18000, loss[loss=0.1565, simple_loss=0.3947, pruned_loss=0.07504, ctc_loss=0.1069, over 14416.00 frames. ], tot_loss[loss=0.1816, simple_loss=0.4333, pruned_loss=0.08787, ctc_loss=0.1289, over 2858793.86 frames. ], batch size: 33, lr: 3.00e-03, 
2023-01-05 22:46:11,670 INFO [train.py:887] Computing validation loss
2023-01-05 22:46:37,290 INFO [train.py:897] Epoch 18, validation: loss=0.1967, simple_loss=0.4487, pruned_loss=0.09831, ctc_loss=0.1427, over 944034.00 frames. 
2023-01-05 22:46:37,291 INFO [train.py:898] Maximum memory allocated so far is 7283MB
2023-01-05 22:47:04,822 INFO [checkpoint.py:75] Saving checkpoint to tiny_transducer_ctc/exp_2m_phone_halfdelay/epoch-18.pt
2023-01-05 22:47:07,921 INFO [train.py:862] Epoch 19, batch 0, loss[loss=0.2204, simple_loss=0.4838, pruned_loss=0.1082, ctc_loss=0.1649, over 14467.00 frames. ], tot_loss[loss=0.2204, simple_loss=0.4838, pruned_loss=0.1082, ctc_loss=0.1649, over 14467.00 frames. ], batch size: 50, lr: 3.00e-03, 
2023-01-05 22:48:12,711 INFO [train.py:862] Epoch 19, batch 500, loss[loss=0.1368, simple_loss=0.3893, pruned_loss=0.06189, ctc_loss=0.08544, over 14510.00 frames. ], tot_loss[loss=0.1861, simple_loss=0.4378, pruned_loss=0.09058, ctc_loss=0.1332, over 2637512.37 frames. ], batch size: 37, lr: 3.00e-03, 
2023-01-05 22:49:17,955 INFO [train.py:862] Epoch 19, batch 1000, loss[loss=0.1828, simple_loss=0.4412, pruned_loss=0.08777, ctc_loss=0.1289, over 14674.00 frames. ], tot_loss[loss=0.1821, simple_loss=0.4345, pruned_loss=0.08789, ctc_loss=0.1293, over 2846658.49 frames. ], batch size: 42, lr: 3.00e-03, 
2023-01-05 22:50:23,751 INFO [train.py:862] Epoch 19, batch 1500, loss[loss=0.1732, simple_loss=0.4471, pruned_loss=0.08472, ctc_loss=0.1153, over 14792.00 frames. ], tot_loss[loss=0.1837, simple_loss=0.4346, pruned_loss=0.08947, ctc_loss=0.131, over 2871609.86 frames. ], batch size: 43, lr: 3.00e-03, 
2023-01-05 22:51:28,885 INFO [train.py:862] Epoch 19, batch 2000, loss[loss=0.173, simple_loss=0.4277, pruned_loss=0.08125, ctc_loss=0.1207, over 14870.00 frames. ], tot_loss[loss=0.181, simple_loss=0.432, pruned_loss=0.08743, ctc_loss=0.1285, over 2865639.34 frames. ], batch size: 40, lr: 3.00e-03, 
2023-01-05 22:51:43,018 INFO [checkpoint.py:75] Saving checkpoint to tiny_transducer_ctc/exp_2m_phone_halfdelay/checkpoint-330000.pt
2023-01-05 22:52:34,306 INFO [train.py:862] Epoch 19, batch 2500, loss[loss=0.2022, simple_loss=0.4685, pruned_loss=0.1017, ctc_loss=0.1449, over 14790.00 frames. ], tot_loss[loss=0.1808, simple_loss=0.4344, pruned_loss=0.08684, ctc_loss=0.1279, over 2882610.97 frames. ], batch size: 43, lr: 3.00e-03, 
2023-01-05 22:53:40,130 INFO [train.py:862] Epoch 19, batch 3000, loss[loss=0.1712, simple_loss=0.4451, pruned_loss=0.08279, ctc_loss=0.1137, over 14830.00 frames. ], tot_loss[loss=0.181, simple_loss=0.4344, pruned_loss=0.08728, ctc_loss=0.1281, over 2878465.36 frames. ], batch size: 42, lr: 3.00e-03, 
2023-01-05 22:54:45,070 INFO [train.py:862] Epoch 19, batch 3500, loss[loss=0.1412, simple_loss=0.3833, pruned_loss=0.06015, ctc_loss=0.09385, over 14521.00 frames. ], tot_loss[loss=0.183, simple_loss=0.4355, pruned_loss=0.08843, ctc_loss=0.1302, over 2846492.40 frames. ], batch size: 37, lr: 3.00e-03, 
2023-01-05 22:55:50,226 INFO [train.py:862] Epoch 19, batch 4000, loss[loss=0.1608, simple_loss=0.427, pruned_loss=0.07427, ctc_loss=0.1064, over 14693.00 frames. ], tot_loss[loss=0.1823, simple_loss=0.434, pruned_loss=0.08814, ctc_loss=0.1297, over 2865737.76 frames. ], batch size: 44, lr: 3.00e-03, 
2023-01-05 22:56:55,576 INFO [train.py:862] Epoch 19, batch 4500, loss[loss=0.1414, simple_loss=0.3817, pruned_loss=0.06631, ctc_loss=0.09182, over 14679.00 frames. ], tot_loss[loss=0.1828, simple_loss=0.4348, pruned_loss=0.08843, ctc_loss=0.1301, over 2868577.80 frames. ], batch size: 35, lr: 3.00e-03, 
2023-01-05 22:58:00,766 INFO [train.py:862] Epoch 19, batch 5000, loss[loss=0.1819, simple_loss=0.4522, pruned_loss=0.08578, ctc_loss=0.1262, over 14727.00 frames. ], tot_loss[loss=0.1821, simple_loss=0.4329, pruned_loss=0.08839, ctc_loss=0.1295, over 2871627.76 frames. ], batch size: 46, lr: 3.00e-03, 
2023-01-05 22:59:06,475 INFO [train.py:862] Epoch 19, batch 5500, loss[loss=0.1856, simple_loss=0.438, pruned_loss=0.08648, ctc_loss=0.1342, over 13802.00 frames. ], tot_loss[loss=0.1826, simple_loss=0.4346, pruned_loss=0.08843, ctc_loss=0.1298, over 2871147.41 frames. ], batch size: 63, lr: 3.00e-03, 
2023-01-05 23:00:11,772 INFO [train.py:862] Epoch 19, batch 6000, loss[loss=0.1824, simple_loss=0.4235, pruned_loss=0.08789, ctc_loss=0.1322, over 14832.00 frames. ], tot_loss[loss=0.1815, simple_loss=0.4336, pruned_loss=0.0877, ctc_loss=0.1287, over 2862779.00 frames. ], batch size: 41, lr: 3.00e-03, 
2023-01-05 23:01:16,789 INFO [train.py:862] Epoch 19, batch 6500, loss[loss=0.2158, simple_loss=0.4797, pruned_loss=0.1093, ctc_loss=0.1586, over 14659.00 frames. ], tot_loss[loss=0.1821, simple_loss=0.4333, pruned_loss=0.08807, ctc_loss=0.1296, over 2865877.00 frames. ], batch size: 41, lr: 3.00e-03, 
2023-01-05 23:02:22,481 INFO [train.py:862] Epoch 19, batch 7000, loss[loss=0.192, simple_loss=0.4518, pruned_loss=0.09273, ctc_loss=0.1378, over 13968.00 frames. ], tot_loss[loss=0.1807, simple_loss=0.4325, pruned_loss=0.08719, ctc_loss=0.1281, over 2864654.83 frames. ], batch size: 56, lr: 3.00e-03, 
2023-01-05 23:03:27,658 INFO [train.py:862] Epoch 19, batch 7500, loss[loss=0.1666, simple_loss=0.413, pruned_loss=0.07973, ctc_loss=0.1153, over 14516.00 frames. ], tot_loss[loss=0.1817, simple_loss=0.4345, pruned_loss=0.08787, ctc_loss=0.1288, over 2869786.15 frames. ], batch size: 37, lr: 3.00e-03, 
2023-01-05 23:04:33,000 INFO [train.py:862] Epoch 19, batch 8000, loss[loss=0.1821, simple_loss=0.4224, pruned_loss=0.08815, ctc_loss=0.1319, over 14797.00 frames. ], tot_loss[loss=0.1823, simple_loss=0.4342, pruned_loss=0.08803, ctc_loss=0.1296, over 2843157.71 frames. ], batch size: 36, lr: 3.00e-03, 
2023-01-05 23:05:38,053 INFO [train.py:862] Epoch 19, batch 8500, loss[loss=0.1707, simple_loss=0.4254, pruned_loss=0.07543, ctc_loss=0.1203, over 14692.00 frames. ], tot_loss[loss=0.1821, simple_loss=0.4348, pruned_loss=0.08785, ctc_loss=0.1293, over 2875124.99 frames. ], batch size: 39, lr: 3.00e-03, 
2023-01-05 23:06:43,113 INFO [train.py:862] Epoch 19, batch 9000, loss[loss=0.19, simple_loss=0.4529, pruned_loss=0.09082, ctc_loss=0.1354, over 14184.00 frames. ], tot_loss[loss=0.1809, simple_loss=0.4329, pruned_loss=0.08721, ctc_loss=0.1283, over 2864399.34 frames. ], batch size: 56, lr: 3.00e-03, 
2023-01-05 23:06:43,114 INFO [train.py:887] Computing validation loss
2023-01-05 23:07:09,330 INFO [train.py:897] Epoch 19, validation: loss=0.1967, simple_loss=0.4485, pruned_loss=0.09732, ctc_loss=0.1432, over 944034.00 frames. 
2023-01-05 23:07:09,331 INFO [train.py:898] Maximum memory allocated so far is 7283MB
2023-01-05 23:08:16,224 INFO [train.py:862] Epoch 19, batch 9500, loss[loss=0.1882, simple_loss=0.4349, pruned_loss=0.09252, ctc_loss=0.136, over 12946.00 frames. ], tot_loss[loss=0.1811, simple_loss=0.433, pruned_loss=0.08757, ctc_loss=0.1283, over 2862467.77 frames. ], batch size: 75, lr: 3.00e-03, 
2023-01-05 23:09:23,407 INFO [train.py:862] Epoch 19, batch 10000, loss[loss=0.1743, simple_loss=0.4094, pruned_loss=0.08805, ctc_loss=0.1235, over 14711.00 frames. ], tot_loss[loss=0.181, simple_loss=0.4334, pruned_loss=0.0872, ctc_loss=0.1284, over 2861935.08 frames. ], batch size: 38, lr: 3.00e-03, 
2023-01-05 23:10:30,444 INFO [train.py:862] Epoch 19, batch 10500, loss[loss=0.1551, simple_loss=0.4229, pruned_loss=0.07057, ctc_loss=0.1008, over 14661.00 frames. ], tot_loss[loss=0.1813, simple_loss=0.4332, pruned_loss=0.08783, ctc_loss=0.1286, over 2849610.64 frames. ], batch size: 42, lr: 3.00e-03, 
2023-01-05 23:11:37,444 INFO [train.py:862] Epoch 19, batch 11000, loss[loss=0.1667, simple_loss=0.4258, pruned_loss=0.0705, ctc_loss=0.1167, over 14477.00 frames. ], tot_loss[loss=0.1827, simple_loss=0.4348, pruned_loss=0.08838, ctc_loss=0.1299, over 2859707.99 frames. ], batch size: 48, lr: 3.00e-03, 
2023-01-05 23:12:44,847 INFO [train.py:862] Epoch 19, batch 11500, loss[loss=0.1694, simple_loss=0.4418, pruned_loss=0.07549, ctc_loss=0.115, over 14086.00 frames. ], tot_loss[loss=0.1791, simple_loss=0.4326, pruned_loss=0.08613, ctc_loss=0.1263, over 2869715.04 frames. ], batch size: 56, lr: 3.00e-03, 
2023-01-05 23:13:51,437 INFO [train.py:862] Epoch 19, batch 12000, loss[loss=0.2086, simple_loss=0.4664, pruned_loss=0.105, ctc_loss=0.153, over 13221.00 frames. ], tot_loss[loss=0.1811, simple_loss=0.4332, pruned_loss=0.08737, ctc_loss=0.1284, over 2862311.89 frames. ], batch size: 76, lr: 3.00e-03, 
2023-01-05 23:14:06,501 INFO [checkpoint.py:75] Saving checkpoint to tiny_transducer_ctc/exp_2m_phone_halfdelay/checkpoint-340000.pt
2023-01-05 23:14:59,132 INFO [train.py:862] Epoch 19, batch 12500, loss[loss=0.2346, simple_loss=0.4783, pruned_loss=0.1187, ctc_loss=0.1818, over 13676.00 frames. ], tot_loss[loss=0.1839, simple_loss=0.437, pruned_loss=0.08902, ctc_loss=0.1309, over 2855182.82 frames. ], batch size: 63, lr: 3.00e-03, 
2023-01-05 23:16:05,870 INFO [train.py:862] Epoch 19, batch 13000, loss[loss=0.207, simple_loss=0.4669, pruned_loss=0.1012, ctc_loss=0.1524, over 14497.00 frames. ], tot_loss[loss=0.1816, simple_loss=0.4339, pruned_loss=0.08778, ctc_loss=0.1289, over 2854176.60 frames. ], batch size: 50, lr: 3.00e-03, 
2023-01-05 23:17:12,526 INFO [train.py:862] Epoch 19, batch 13500, loss[loss=0.2098, simple_loss=0.475, pruned_loss=0.1039, ctc_loss=0.1534, over 14753.00 frames. ], tot_loss[loss=0.1812, simple_loss=0.4339, pruned_loss=0.0874, ctc_loss=0.1285, over 2871625.75 frames. ], batch size: 46, lr: 3.00e-03, 
2023-01-05 23:18:20,186 INFO [train.py:862] Epoch 19, batch 14000, loss[loss=0.1471, simple_loss=0.3702, pruned_loss=0.07055, ctc_loss=0.1006, over 14410.00 frames. ], tot_loss[loss=0.1785, simple_loss=0.4305, pruned_loss=0.08566, ctc_loss=0.1261, over 2862750.89 frames. ], batch size: 33, lr: 3.00e-03, 
2023-01-05 23:19:27,123 INFO [train.py:862] Epoch 19, batch 14500, loss[loss=0.1849, simple_loss=0.4489, pruned_loss=0.08953, ctc_loss=0.1295, over 14652.00 frames. ], tot_loss[loss=0.1829, simple_loss=0.4344, pruned_loss=0.08843, ctc_loss=0.1303, over 2840837.11 frames. ], batch size: 41, lr: 3.00e-03, 
2023-01-05 23:20:34,790 INFO [train.py:862] Epoch 19, batch 15000, loss[loss=0.1494, simple_loss=0.4133, pruned_loss=0.06223, ctc_loss=0.09816, over 14831.00 frames. ], tot_loss[loss=0.1827, simple_loss=0.4342, pruned_loss=0.08846, ctc_loss=0.1301, over 2881295.58 frames. ], batch size: 41, lr: 3.00e-03, 
2023-01-05 23:21:41,132 INFO [train.py:862] Epoch 19, batch 15500, loss[loss=0.1745, simple_loss=0.4513, pruned_loss=0.08116, ctc_loss=0.1178, over 14571.00 frames. ], tot_loss[loss=0.1795, simple_loss=0.4323, pruned_loss=0.08636, ctc_loss=0.1267, over 2867205.40 frames. ], batch size: 48, lr: 3.00e-03, 
2023-01-05 23:22:47,673 INFO [train.py:862] Epoch 19, batch 16000, loss[loss=0.1926, simple_loss=0.4227, pruned_loss=0.09815, ctc_loss=0.1425, over 14717.00 frames. ], tot_loss[loss=0.1812, simple_loss=0.4336, pruned_loss=0.08746, ctc_loss=0.1285, over 2869740.11 frames. ], batch size: 37, lr: 3.00e-03, 
2023-01-05 23:23:55,449 INFO [train.py:862] Epoch 19, batch 16500, loss[loss=0.189, simple_loss=0.441, pruned_loss=0.09237, ctc_loss=0.1358, over 14058.00 frames. ], tot_loss[loss=0.1784, simple_loss=0.4309, pruned_loss=0.08561, ctc_loss=0.1259, over 2863519.69 frames. ], batch size: 56, lr: 3.00e-03, 
2023-01-05 23:25:02,244 INFO [train.py:862] Epoch 19, batch 17000, loss[loss=0.2073, simple_loss=0.4517, pruned_loss=0.1039, ctc_loss=0.1548, over 14579.00 frames. ], tot_loss[loss=0.1807, simple_loss=0.4327, pruned_loss=0.08746, ctc_loss=0.1279, over 2860218.89 frames. ], batch size: 43, lr: 3.00e-03, 
2023-01-05 23:26:09,257 INFO [train.py:862] Epoch 19, batch 17500, loss[loss=0.1504, simple_loss=0.4054, pruned_loss=0.06718, ctc_loss=0.09918, over 14716.00 frames. ], tot_loss[loss=0.1805, simple_loss=0.4327, pruned_loss=0.08702, ctc_loss=0.1279, over 2871534.70 frames. ], batch size: 37, lr: 3.00e-03, 
2023-01-05 23:27:13,943 INFO [train.py:862] Epoch 19, batch 18000, loss[loss=0.1964, simple_loss=0.4485, pruned_loss=0.09989, ctc_loss=0.1417, over 13647.00 frames. ], tot_loss[loss=0.1816, simple_loss=0.4333, pruned_loss=0.08799, ctc_loss=0.1289, over 2855451.12 frames. ], batch size: 63, lr: 3.00e-03, 
2023-01-05 23:27:13,944 INFO [train.py:887] Computing validation loss
2023-01-05 23:27:40,114 INFO [train.py:897] Epoch 19, validation: loss=0.1977, simple_loss=0.4489, pruned_loss=0.09858, ctc_loss=0.144, over 944034.00 frames. 
2023-01-05 23:27:40,114 INFO [train.py:898] Maximum memory allocated so far is 7283MB
2023-01-05 23:28:06,883 INFO [checkpoint.py:75] Saving checkpoint to tiny_transducer_ctc/exp_2m_phone_halfdelay/epoch-19.pt
2023-01-05 23:28:10,052 INFO [train.py:862] Epoch 20, batch 0, loss[loss=0.2056, simple_loss=0.4317, pruned_loss=0.1056, ctc_loss=0.156, over 14702.00 frames. ], tot_loss[loss=0.2056, simple_loss=0.4317, pruned_loss=0.1056, ctc_loss=0.156, over 14702.00 frames. ], batch size: 38, lr: 3.00e-03, 
2023-01-05 23:29:15,482 INFO [train.py:862] Epoch 20, batch 500, loss[loss=0.1656, simple_loss=0.4511, pruned_loss=0.07516, ctc_loss=0.1076, over 14695.00 frames. ], tot_loss[loss=0.1833, simple_loss=0.4368, pruned_loss=0.08852, ctc_loss=0.1303, over 2637907.03 frames. ], batch size: 47, lr: 3.00e-03, 
2023-01-05 23:30:21,105 INFO [train.py:862] Epoch 20, batch 1000, loss[loss=0.197, simple_loss=0.4626, pruned_loss=0.09376, ctc_loss=0.1422, over 14704.00 frames. ], tot_loss[loss=0.1795, simple_loss=0.4325, pruned_loss=0.08629, ctc_loss=0.1267, over 2858856.24 frames. ], batch size: 47, lr: 3.00e-03, 
2023-01-05 23:31:27,396 INFO [train.py:862] Epoch 20, batch 1500, loss[loss=0.1981, simple_loss=0.4377, pruned_loss=0.102, ctc_loss=0.1455, over 14647.00 frames. ], tot_loss[loss=0.1808, simple_loss=0.4331, pruned_loss=0.08691, ctc_loss=0.1282, over 2862277.10 frames. ], batch size: 40, lr: 3.00e-03, 
2023-01-05 23:32:33,118 INFO [train.py:862] Epoch 20, batch 2000, loss[loss=0.1495, simple_loss=0.3964, pruned_loss=0.07014, ctc_loss=0.09862, over 14793.00 frames. ], tot_loss[loss=0.1793, simple_loss=0.4309, pruned_loss=0.08629, ctc_loss=0.1268, over 2859868.85 frames. ], batch size: 36, lr: 3.00e-03, 
2023-01-05 23:33:38,867 INFO [train.py:862] Epoch 20, batch 2500, loss[loss=0.1657, simple_loss=0.4276, pruned_loss=0.07615, ctc_loss=0.1125, over 14817.00 frames. ], tot_loss[loss=0.181, simple_loss=0.4339, pruned_loss=0.08715, ctc_loss=0.1283, over 2883472.46 frames. ], batch size: 41, lr: 3.00e-03, 
2023-01-05 23:34:44,980 INFO [train.py:862] Epoch 20, batch 3000, loss[loss=0.1539, simple_loss=0.397, pruned_loss=0.07135, ctc_loss=0.1042, over 14670.00 frames. ], tot_loss[loss=0.18, simple_loss=0.4325, pruned_loss=0.0865, ctc_loss=0.1273, over 2870937.71 frames. ], batch size: 35, lr: 3.00e-03, 
2023-01-05 23:35:50,803 INFO [train.py:862] Epoch 20, batch 3500, loss[loss=0.1468, simple_loss=0.4172, pruned_loss=0.06133, ctc_loss=0.09398, over 14732.00 frames. ], tot_loss[loss=0.1801, simple_loss=0.4313, pruned_loss=0.08671, ctc_loss=0.1277, over 2852324.04 frames. ], batch size: 45, lr: 3.00e-03, 
2023-01-05 23:36:42,815 INFO [checkpoint.py:75] Saving checkpoint to tiny_transducer_ctc/exp_2m_phone_halfdelay/checkpoint-350000.pt
2023-01-05 23:36:56,418 INFO [train.py:862] Epoch 20, batch 4000, loss[loss=0.1904, simple_loss=0.4547, pruned_loss=0.09042, ctc_loss=0.1358, over 14651.00 frames. ], tot_loss[loss=0.1802, simple_loss=0.4322, pruned_loss=0.08684, ctc_loss=0.1276, over 2862801.58 frames. ], batch size: 42, lr: 3.00e-03, 
2023-01-05 23:38:02,680 INFO [train.py:862] Epoch 20, batch 4500, loss[loss=0.1646, simple_loss=0.4084, pruned_loss=0.07819, ctc_loss=0.1142, over 14725.00 frames. ], tot_loss[loss=0.1797, simple_loss=0.4323, pruned_loss=0.0864, ctc_loss=0.127, over 2859244.16 frames. ], batch size: 37, lr: 3.00e-03, 
2023-01-05 23:39:08,223 INFO [train.py:862] Epoch 20, batch 5000, loss[loss=0.1685, simple_loss=0.4149, pruned_loss=0.07943, ctc_loss=0.1177, over 14701.00 frames. ], tot_loss[loss=0.1795, simple_loss=0.4309, pruned_loss=0.08634, ctc_loss=0.127, over 2856377.49 frames. ], batch size: 38, lr: 3.00e-03, 
2023-01-05 23:40:14,698 INFO [train.py:862] Epoch 20, batch 5500, loss[loss=0.1879, simple_loss=0.4524, pruned_loss=0.09429, ctc_loss=0.1311, over 13775.00 frames. ], tot_loss[loss=0.1809, simple_loss=0.4328, pruned_loss=0.08751, ctc_loss=0.1282, over 2869622.07 frames. ], batch size: 63, lr: 3.00e-03, 
2023-01-05 23:41:20,569 INFO [train.py:862] Epoch 20, batch 6000, loss[loss=0.1724, simple_loss=0.4309, pruned_loss=0.08025, ctc_loss=0.1195, over 14697.00 frames. ], tot_loss[loss=0.1842, simple_loss=0.435, pruned_loss=0.08952, ctc_loss=0.1316, over 2865557.91 frames. ], batch size: 44, lr: 3.00e-03, 
2023-01-05 23:42:26,087 INFO [train.py:862] Epoch 20, batch 6500, loss[loss=0.183, simple_loss=0.4562, pruned_loss=0.08652, ctc_loss=0.1266, over 14701.00 frames. ], tot_loss[loss=0.1799, simple_loss=0.4322, pruned_loss=0.08666, ctc_loss=0.1272, over 2873297.25 frames. ], batch size: 45, lr: 3.00e-03, 
2023-01-05 23:43:31,693 INFO [train.py:862] Epoch 20, batch 7000, loss[loss=0.1547, simple_loss=0.3923, pruned_loss=0.07404, ctc_loss=0.1052, over 14388.00 frames. ], tot_loss[loss=0.1802, simple_loss=0.4325, pruned_loss=0.08699, ctc_loss=0.1275, over 2867029.30 frames. ], batch size: 33, lr: 3.00e-03, 
2023-01-05 23:44:36,954 INFO [train.py:862] Epoch 20, batch 7500, loss[loss=0.1826, simple_loss=0.4437, pruned_loss=0.08632, ctc_loss=0.1288, over 14798.00 frames. ], tot_loss[loss=0.1819, simple_loss=0.4342, pruned_loss=0.08787, ctc_loss=0.1292, over 2854580.82 frames. ], batch size: 43, lr: 3.00e-03, 
2023-01-05 23:45:42,655 INFO [train.py:862] Epoch 20, batch 8000, loss[loss=0.2514, simple_loss=0.4958, pruned_loss=0.1371, ctc_loss=0.1941, over 13633.00 frames. ], tot_loss[loss=0.1805, simple_loss=0.4323, pruned_loss=0.08708, ctc_loss=0.1279, over 2855678.36 frames. ], batch size: 63, lr: 3.00e-03, 
2023-01-05 23:46:47,865 INFO [train.py:862] Epoch 20, batch 8500, loss[loss=0.2292, simple_loss=0.4812, pruned_loss=0.1207, ctc_loss=0.1726, over 14538.00 frames. ], tot_loss[loss=0.1802, simple_loss=0.4333, pruned_loss=0.08676, ctc_loss=0.1274, over 2867994.03 frames. ], batch size: 53, lr: 3.00e-03, 
2023-01-05 23:47:52,905 INFO [train.py:862] Epoch 20, batch 9000, loss[loss=0.2006, simple_loss=0.414, pruned_loss=0.1027, ctc_loss=0.1539, over 14400.00 frames. ], tot_loss[loss=0.1803, simple_loss=0.4323, pruned_loss=0.08698, ctc_loss=0.1277, over 2854455.66 frames. ], batch size: 33, lr: 3.00e-03, 
2023-01-05 23:47:52,905 INFO [train.py:887] Computing validation loss
2023-01-05 23:48:18,829 INFO [train.py:897] Epoch 20, validation: loss=0.197, simple_loss=0.4486, pruned_loss=0.09754, ctc_loss=0.1435, over 944034.00 frames. 
2023-01-05 23:48:18,829 INFO [train.py:898] Maximum memory allocated so far is 7283MB
2023-01-05 23:49:24,914 INFO [train.py:862] Epoch 20, batch 9500, loss[loss=0.1662, simple_loss=0.4339, pruned_loss=0.07543, ctc_loss=0.1121, over 13653.00 frames. ], tot_loss[loss=0.1809, simple_loss=0.4342, pruned_loss=0.08726, ctc_loss=0.1279, over 2866879.43 frames. ], batch size: 63, lr: 3.00e-03, 
2023-01-05 23:50:31,305 INFO [train.py:862] Epoch 20, batch 10000, loss[loss=0.1918, simple_loss=0.4397, pruned_loss=0.08933, ctc_loss=0.1414, over 14657.00 frames. ], tot_loss[loss=0.1829, simple_loss=0.4347, pruned_loss=0.08847, ctc_loss=0.1302, over 2857782.68 frames. ], batch size: 42, lr: 3.00e-03, 
2023-01-05 23:51:36,983 INFO [train.py:862] Epoch 20, batch 10500, loss[loss=0.1693, simple_loss=0.4332, pruned_loss=0.07874, ctc_loss=0.1153, over 13736.00 frames. ], tot_loss[loss=0.1818, simple_loss=0.4334, pruned_loss=0.08777, ctc_loss=0.1292, over 2857922.08 frames. ], batch size: 63, lr: 3.00e-03, 
2023-01-05 23:52:42,920 INFO [train.py:862] Epoch 20, batch 11000, loss[loss=0.2035, simple_loss=0.472, pruned_loss=0.09989, ctc_loss=0.1468, over 14508.00 frames. ], tot_loss[loss=0.1824, simple_loss=0.4347, pruned_loss=0.08815, ctc_loss=0.1296, over 2872332.55 frames. ], batch size: 53, lr: 3.00e-03, 
2023-01-05 23:53:49,297 INFO [train.py:862] Epoch 20, batch 11500, loss[loss=0.1519, simple_loss=0.373, pruned_loss=0.06775, ctc_loss=0.1081, over 14414.00 frames. ], tot_loss[loss=0.1824, simple_loss=0.434, pruned_loss=0.08848, ctc_loss=0.1297, over 2874121.97 frames. ], batch size: 33, lr: 3.00e-03, 
2023-01-05 23:54:55,150 INFO [train.py:862] Epoch 20, batch 12000, loss[loss=0.1353, simple_loss=0.3764, pruned_loss=0.05738, ctc_loss=0.08799, over 14397.00 frames. ], tot_loss[loss=0.1816, simple_loss=0.4336, pruned_loss=0.0875, ctc_loss=0.129, over 2851866.58 frames. ], batch size: 33, lr: 3.00e-03, 
2023-01-05 23:56:01,655 INFO [train.py:862] Epoch 20, batch 12500, loss[loss=0.1449, simple_loss=0.4141, pruned_loss=0.06275, ctc_loss=0.09143, over 14866.00 frames. ], tot_loss[loss=0.18, simple_loss=0.4331, pruned_loss=0.08651, ctc_loss=0.1273, over 2878788.89 frames. ], batch size: 40, lr: 3.00e-03, 
2023-01-05 23:57:07,439 INFO [train.py:862] Epoch 20, batch 13000, loss[loss=0.2038, simple_loss=0.4655, pruned_loss=0.1017, ctc_loss=0.1479, over 14825.00 frames. ], tot_loss[loss=0.1816, simple_loss=0.4341, pruned_loss=0.08773, ctc_loss=0.1287, over 2861996.19 frames. ], batch size: 41, lr: 3.00e-03, 
2023-01-05 23:58:13,226 INFO [train.py:862] Epoch 20, batch 13500, loss[loss=0.1595, simple_loss=0.4102, pruned_loss=0.07099, ctc_loss=0.1095, over 14669.00 frames. ], tot_loss[loss=0.1807, simple_loss=0.4342, pruned_loss=0.08697, ctc_loss=0.1279, over 2853241.20 frames. ], batch size: 40, lr: 3.00e-03, 
2023-01-05 23:59:05,576 INFO [checkpoint.py:75] Saving checkpoint to tiny_transducer_ctc/exp_2m_phone_halfdelay/checkpoint-360000.pt
2023-01-05 23:59:19,738 INFO [train.py:862] Epoch 20, batch 14000, loss[loss=0.155, simple_loss=0.3882, pruned_loss=0.07521, ctc_loss=0.106, over 14782.00 frames. ], tot_loss[loss=0.1831, simple_loss=0.4346, pruned_loss=0.08857, ctc_loss=0.1304, over 2853237.45 frames. ], batch size: 36, lr: 3.00e-03, 
2023-01-06 00:00:25,716 INFO [train.py:862] Epoch 20, batch 14500, loss[loss=0.1526, simple_loss=0.3862, pruned_loss=0.07448, ctc_loss=0.1033, over 14516.00 frames. ], tot_loss[loss=0.18, simple_loss=0.4314, pruned_loss=0.08661, ctc_loss=0.1275, over 2862603.12 frames. ], batch size: 34, lr: 3.00e-03, 
2023-01-06 00:01:32,199 INFO [train.py:862] Epoch 20, batch 15000, loss[loss=0.1926, simple_loss=0.4217, pruned_loss=0.09839, ctc_loss=0.1426, over 14519.00 frames. ], tot_loss[loss=0.1811, simple_loss=0.4337, pruned_loss=0.08735, ctc_loss=0.1284, over 2867305.87 frames. ], batch size: 36, lr: 3.00e-03, 
2023-01-06 00:02:38,235 INFO [train.py:862] Epoch 20, batch 15500, loss[loss=0.1877, simple_loss=0.4559, pruned_loss=0.09309, ctc_loss=0.1305, over 14676.00 frames. ], tot_loss[loss=0.1797, simple_loss=0.4328, pruned_loss=0.08641, ctc_loss=0.1269, over 2853432.07 frames. ], batch size: 47, lr: 3.00e-03, 
2023-01-06 00:03:44,502 INFO [train.py:862] Epoch 20, batch 16000, loss[loss=0.1629, simple_loss=0.3948, pruned_loss=0.07429, ctc_loss=0.1163, over 14529.00 frames. ], tot_loss[loss=0.1807, simple_loss=0.4323, pruned_loss=0.08735, ctc_loss=0.128, over 2853029.11 frames. ], batch size: 36, lr: 3.00e-03, 
2023-01-06 00:04:50,564 INFO [train.py:862] Epoch 20, batch 16500, loss[loss=0.1918, simple_loss=0.4559, pruned_loss=0.08995, ctc_loss=0.1377, over 14650.00 frames. ], tot_loss[loss=0.1805, simple_loss=0.4331, pruned_loss=0.08696, ctc_loss=0.1278, over 2862509.40 frames. ], batch size: 42, lr: 3.00e-03, 
2023-01-06 00:05:56,368 INFO [train.py:862] Epoch 20, batch 17000, loss[loss=0.1815, simple_loss=0.455, pruned_loss=0.08241, ctc_loss=0.1265, over 14693.00 frames. ], tot_loss[loss=0.1817, simple_loss=0.4324, pruned_loss=0.08804, ctc_loss=0.1292, over 2864445.57 frames. ], batch size: 47, lr: 3.00e-03, 
2023-01-06 00:07:02,829 INFO [train.py:862] Epoch 20, batch 17500, loss[loss=0.1723, simple_loss=0.4174, pruned_loss=0.08202, ctc_loss=0.1215, over 14718.00 frames. ], tot_loss[loss=0.1814, simple_loss=0.4332, pruned_loss=0.08755, ctc_loss=0.1288, over 2851063.47 frames. ], batch size: 38, lr: 3.00e-03, 
2023-01-06 00:08:07,080 INFO [train.py:862] Epoch 20, batch 18000, loss[loss=0.132, simple_loss=0.3759, pruned_loss=0.05636, ctc_loss=0.08384, over 14517.00 frames. ], tot_loss[loss=0.1803, simple_loss=0.4332, pruned_loss=0.08706, ctc_loss=0.1275, over 2885603.35 frames. ], batch size: 37, lr: 3.00e-03, 
2023-01-06 00:08:07,081 INFO [train.py:887] Computing validation loss
2023-01-06 00:08:32,996 INFO [train.py:897] Epoch 20, validation: loss=0.1942, simple_loss=0.4471, pruned_loss=0.09607, ctc_loss=0.1404, over 944034.00 frames. 
2023-01-06 00:08:32,997 INFO [train.py:898] Maximum memory allocated so far is 7283MB
2023-01-06 00:08:59,251 INFO [checkpoint.py:75] Saving checkpoint to tiny_transducer_ctc/exp_2m_phone_halfdelay/epoch-20.pt
2023-01-06 00:09:02,251 INFO [train.py:862] Epoch 21, batch 0, loss[loss=0.1936, simple_loss=0.4082, pruned_loss=0.09997, ctc_loss=0.1463, over 14411.00 frames. ], tot_loss[loss=0.1936, simple_loss=0.4082, pruned_loss=0.09997, ctc_loss=0.1463, over 14411.00 frames. ], batch size: 33, lr: 3.00e-03, 
2023-01-06 00:10:07,785 INFO [train.py:862] Epoch 21, batch 500, loss[loss=0.148, simple_loss=0.4353, pruned_loss=0.06347, ctc_loss=0.09094, over 14553.00 frames. ], tot_loss[loss=0.18, simple_loss=0.4334, pruned_loss=0.08648, ctc_loss=0.1273, over 2648712.06 frames. ], batch size: 50, lr: 3.00e-03, 
2023-01-06 00:11:13,136 INFO [train.py:862] Epoch 21, batch 1000, loss[loss=0.1796, simple_loss=0.4409, pruned_loss=0.08719, ctc_loss=0.1247, over 14317.00 frames. ], tot_loss[loss=0.181, simple_loss=0.4339, pruned_loss=0.08708, ctc_loss=0.1283, over 2837903.72 frames. ], batch size: 52, lr: 3.00e-03, 
2023-01-06 00:12:18,624 INFO [train.py:862] Epoch 21, batch 1500, loss[loss=0.1722, simple_loss=0.4449, pruned_loss=0.07706, ctc_loss=0.1177, over 14524.00 frames. ], tot_loss[loss=0.1799, simple_loss=0.4318, pruned_loss=0.08658, ctc_loss=0.1274, over 2869762.27 frames. ], batch size: 53, lr: 3.00e-03, 
2023-01-06 00:13:23,888 INFO [train.py:862] Epoch 21, batch 2000, loss[loss=0.1466, simple_loss=0.4274, pruned_loss=0.06415, ctc_loss=0.0903, over 14753.00 frames. ], tot_loss[loss=0.1794, simple_loss=0.4327, pruned_loss=0.08605, ctc_loss=0.1267, over 2870443.70 frames. ], batch size: 45, lr: 3.00e-03, 
2023-01-06 00:14:29,879 INFO [train.py:862] Epoch 21, batch 2500, loss[loss=0.1648, simple_loss=0.4309, pruned_loss=0.07617, ctc_loss=0.1104, over 14846.00 frames. ], tot_loss[loss=0.1786, simple_loss=0.4316, pruned_loss=0.08556, ctc_loss=0.126, over 2866556.29 frames. ], batch size: 42, lr: 3.00e-03, 
2023-01-06 00:15:35,565 INFO [train.py:862] Epoch 21, batch 3000, loss[loss=0.1892, simple_loss=0.4457, pruned_loss=0.09512, ctc_loss=0.1341, over 14666.00 frames. ], tot_loss[loss=0.1797, simple_loss=0.4327, pruned_loss=0.08633, ctc_loss=0.127, over 2867369.24 frames. ], batch size: 40, lr: 3.00e-03, 
2023-01-06 00:16:41,735 INFO [train.py:862] Epoch 21, batch 3500, loss[loss=0.1736, simple_loss=0.4601, pruned_loss=0.07509, ctc_loss=0.1173, over 14493.00 frames. ], tot_loss[loss=0.1801, simple_loss=0.4323, pruned_loss=0.08654, ctc_loss=0.1276, over 2864458.91 frames. ], batch size: 50, lr: 3.00e-03, 
2023-01-06 00:17:47,342 INFO [train.py:862] Epoch 21, batch 4000, loss[loss=0.2416, simple_loss=0.4787, pruned_loss=0.1292, ctc_loss=0.1872, over 12986.00 frames. ], tot_loss[loss=0.1811, simple_loss=0.4336, pruned_loss=0.08721, ctc_loss=0.1284, over 2871849.67 frames. ], batch size: 76, lr: 3.00e-03, 
2023-01-06 00:18:53,131 INFO [train.py:862] Epoch 21, batch 4500, loss[loss=0.1791, simple_loss=0.4268, pruned_loss=0.08622, ctc_loss=0.1274, over 14693.00 frames. ], tot_loss[loss=0.1826, simple_loss=0.434, pruned_loss=0.0884, ctc_loss=0.1299, over 2851373.11 frames. ], batch size: 39, lr: 3.00e-03, 
2023-01-06 00:19:59,482 INFO [train.py:862] Epoch 21, batch 5000, loss[loss=0.1709, simple_loss=0.4306, pruned_loss=0.07903, ctc_loss=0.118, over 14491.00 frames. ], tot_loss[loss=0.1805, simple_loss=0.4329, pruned_loss=0.08707, ctc_loss=0.1278, over 2867263.40 frames. ], batch size: 37, lr: 3.00e-03, 
2023-01-06 00:21:05,326 INFO [train.py:862] Epoch 21, batch 5500, loss[loss=0.1486, simple_loss=0.4038, pruned_loss=0.0696, ctc_loss=0.09593, over 14512.00 frames. ], tot_loss[loss=0.184, simple_loss=0.4354, pruned_loss=0.08902, ctc_loss=0.1315, over 2859462.19 frames. ], batch size: 37, lr: 3.00e-03, 
2023-01-06 00:21:29,367 INFO [checkpoint.py:75] Saving checkpoint to tiny_transducer_ctc/exp_2m_phone_halfdelay/checkpoint-370000.pt
2023-01-06 00:22:11,637 INFO [train.py:862] Epoch 21, batch 6000, loss[loss=0.1708, simple_loss=0.4234, pruned_loss=0.08213, ctc_loss=0.1181, over 14661.00 frames. ], tot_loss[loss=0.1811, simple_loss=0.4338, pruned_loss=0.08737, ctc_loss=0.1283, over 2868540.15 frames. ], batch size: 40, lr: 3.00e-03, 
2023-01-06 00:23:16,943 INFO [train.py:862] Epoch 21, batch 6500, loss[loss=0.1306, simple_loss=0.387, pruned_loss=0.05452, ctc_loss=0.0803, over 14721.00 frames. ], tot_loss[loss=0.1756, simple_loss=0.4295, pruned_loss=0.08364, ctc_loss=0.1229, over 2873885.51 frames. ], batch size: 38, lr: 3.00e-03, 
2023-01-06 00:24:22,382 INFO [train.py:862] Epoch 21, batch 7000, loss[loss=0.1627, simple_loss=0.4201, pruned_loss=0.07538, ctc_loss=0.1101, over 14647.00 frames. ], tot_loss[loss=0.1791, simple_loss=0.4317, pruned_loss=0.08608, ctc_loss=0.1264, over 2864543.86 frames. ], batch size: 42, lr: 3.00e-03, 
2023-01-06 00:25:28,626 INFO [train.py:862] Epoch 21, batch 7500, loss[loss=0.1567, simple_loss=0.4188, pruned_loss=0.07016, ctc_loss=0.104, over 14653.00 frames. ], tot_loss[loss=0.1791, simple_loss=0.4318, pruned_loss=0.08599, ctc_loss=0.1265, over 2885198.44 frames. ], batch size: 41, lr: 3.00e-03, 
2023-01-06 00:26:34,407 INFO [train.py:862] Epoch 21, batch 8000, loss[loss=0.1419, simple_loss=0.3757, pruned_loss=0.06529, ctc_loss=0.09415, over 14391.00 frames. ], tot_loss[loss=0.1791, simple_loss=0.4314, pruned_loss=0.08581, ctc_loss=0.1266, over 2871607.08 frames. ], batch size: 33, lr: 3.00e-03, 
2023-01-06 00:27:40,216 INFO [train.py:862] Epoch 21, batch 8500, loss[loss=0.1881, simple_loss=0.4069, pruned_loss=0.08972, ctc_loss=0.1431, over 14421.00 frames. ], tot_loss[loss=0.1811, simple_loss=0.4328, pruned_loss=0.0874, ctc_loss=0.1286, over 2871805.08 frames. ], batch size: 33, lr: 3.00e-03, 
2023-01-06 00:28:46,624 INFO [train.py:862] Epoch 21, batch 9000, loss[loss=0.2063, simple_loss=0.4632, pruned_loss=0.1027, ctc_loss=0.1514, over 14246.00 frames. ], tot_loss[loss=0.1786, simple_loss=0.4303, pruned_loss=0.08577, ctc_loss=0.1262, over 2849619.34 frames. ], batch size: 52, lr: 3.00e-03, 
2023-01-06 00:28:46,625 INFO [train.py:887] Computing validation loss
2023-01-06 00:29:12,146 INFO [train.py:897] Epoch 21, validation: loss=0.1965, simple_loss=0.4482, pruned_loss=0.09748, ctc_loss=0.1429, over 944034.00 frames. 
2023-01-06 00:29:12,147 INFO [train.py:898] Maximum memory allocated so far is 7283MB
2023-01-06 00:30:18,699 INFO [train.py:862] Epoch 21, batch 9500, loss[loss=0.1898, simple_loss=0.4594, pruned_loss=0.09061, ctc_loss=0.1338, over 14752.00 frames. ], tot_loss[loss=0.1809, simple_loss=0.433, pruned_loss=0.08743, ctc_loss=0.1281, over 2870830.34 frames. ], batch size: 46, lr: 3.00e-03, 
2023-01-06 00:31:25,785 INFO [train.py:862] Epoch 21, batch 10000, loss[loss=0.1716, simple_loss=0.4488, pruned_loss=0.07707, ctc_loss=0.116, over 14643.00 frames. ], tot_loss[loss=0.1789, simple_loss=0.4309, pruned_loss=0.08604, ctc_loss=0.1263, over 2860431.85 frames. ], batch size: 41, lr: 3.00e-03, 
2023-01-06 00:32:31,493 INFO [train.py:862] Epoch 21, batch 10500, loss[loss=0.164, simple_loss=0.4257, pruned_loss=0.07606, ctc_loss=0.1105, over 14571.00 frames. ], tot_loss[loss=0.179, simple_loss=0.4314, pruned_loss=0.08576, ctc_loss=0.1265, over 2859539.41 frames. ], batch size: 43, lr: 3.00e-03, 
2023-01-06 00:33:37,913 INFO [train.py:862] Epoch 21, batch 11000, loss[loss=0.1701, simple_loss=0.4117, pruned_loss=0.0803, ctc_loss=0.1203, over 14664.00 frames. ], tot_loss[loss=0.1802, simple_loss=0.4329, pruned_loss=0.08656, ctc_loss=0.1276, over 2852193.80 frames. ], batch size: 35, lr: 3.00e-03, 
2023-01-06 00:34:44,389 INFO [train.py:862] Epoch 21, batch 11500, loss[loss=0.168, simple_loss=0.4139, pruned_loss=0.07474, ctc_loss=0.1193, over 14881.00 frames. ], tot_loss[loss=0.1808, simple_loss=0.4326, pruned_loss=0.08716, ctc_loss=0.1282, over 2879498.82 frames. ], batch size: 40, lr: 3.00e-03, 
2023-01-06 00:35:51,176 INFO [train.py:862] Epoch 21, batch 12000, loss[loss=0.1825, simple_loss=0.4403, pruned_loss=0.08358, ctc_loss=0.1306, over 14524.00 frames. ], tot_loss[loss=0.1806, simple_loss=0.4335, pruned_loss=0.08687, ctc_loss=0.1279, over 2866880.58 frames. ], batch size: 53, lr: 3.00e-03, 
2023-01-06 00:36:57,502 INFO [train.py:862] Epoch 21, batch 12500, loss[loss=0.2514, simple_loss=0.4836, pruned_loss=0.1368, ctc_loss=0.197, over 10139.00 frames. ], tot_loss[loss=0.1801, simple_loss=0.4325, pruned_loss=0.08641, ctc_loss=0.1275, over 2859414.49 frames. ], batch size: 104, lr: 3.00e-03, 
2023-01-06 00:38:03,644 INFO [train.py:862] Epoch 21, batch 13000, loss[loss=0.2283, simple_loss=0.486, pruned_loss=0.1244, ctc_loss=0.1686, over 14536.00 frames. ], tot_loss[loss=0.1799, simple_loss=0.4335, pruned_loss=0.0867, ctc_loss=0.127, over 2873722.91 frames. ], batch size: 50, lr: 3.00e-03, 
2023-01-06 00:39:10,670 INFO [train.py:862] Epoch 21, batch 13500, loss[loss=0.1932, simple_loss=0.4497, pruned_loss=0.09697, ctc_loss=0.1381, over 14649.00 frames. ], tot_loss[loss=0.1803, simple_loss=0.4328, pruned_loss=0.08668, ctc_loss=0.1276, over 2875413.14 frames. ], batch size: 41, lr: 3.00e-03, 
2023-01-06 00:40:16,632 INFO [train.py:862] Epoch 21, batch 14000, loss[loss=0.1454, simple_loss=0.4103, pruned_loss=0.06334, ctc_loss=0.09261, over 14711.00 frames. ], tot_loss[loss=0.1804, simple_loss=0.4318, pruned_loss=0.08698, ctc_loss=0.1279, over 2852341.69 frames. ], batch size: 37, lr: 3.00e-03, 
2023-01-06 00:41:23,302 INFO [train.py:862] Epoch 21, batch 14500, loss[loss=0.1705, simple_loss=0.4417, pruned_loss=0.0775, ctc_loss=0.1158, over 13971.00 frames. ], tot_loss[loss=0.1809, simple_loss=0.4337, pruned_loss=0.08714, ctc_loss=0.1282, over 2858177.70 frames. ], batch size: 56, lr: 3.00e-03, 
2023-01-06 00:42:29,634 INFO [train.py:862] Epoch 21, batch 15000, loss[loss=0.1427, simple_loss=0.3755, pruned_loss=0.06365, ctc_loss=0.09607, over 14507.00 frames. ], tot_loss[loss=0.1809, simple_loss=0.4318, pruned_loss=0.08729, ctc_loss=0.1285, over 2858272.12 frames. ], batch size: 34, lr: 3.00e-03, 
2023-01-06 00:43:36,045 INFO [train.py:862] Epoch 21, batch 15500, loss[loss=0.1492, simple_loss=0.3856, pruned_loss=0.07028, ctc_loss=0.1003, over 14502.00 frames. ], tot_loss[loss=0.1804, simple_loss=0.4314, pruned_loss=0.08697, ctc_loss=0.1279, over 2873242.46 frames. ], batch size: 34, lr: 3.00e-03, 
2023-01-06 00:44:00,215 INFO [checkpoint.py:75] Saving checkpoint to tiny_transducer_ctc/exp_2m_phone_halfdelay/checkpoint-380000.pt
2023-01-06 00:44:43,155 INFO [train.py:862] Epoch 21, batch 16000, loss[loss=0.166, simple_loss=0.4075, pruned_loss=0.07678, ctc_loss=0.1169, over 14529.00 frames. ], tot_loss[loss=0.1778, simple_loss=0.4305, pruned_loss=0.08525, ctc_loss=0.1252, over 2877757.35 frames. ], batch size: 34, lr: 3.00e-03, 
2023-01-06 00:45:49,211 INFO [train.py:862] Epoch 21, batch 16500, loss[loss=0.1623, simple_loss=0.4348, pruned_loss=0.0751, ctc_loss=0.1065, over 14019.00 frames. ], tot_loss[loss=0.1786, simple_loss=0.4311, pruned_loss=0.08563, ctc_loss=0.126, over 2850707.23 frames. ], batch size: 56, lr: 3.00e-03, 
2023-01-06 00:46:56,075 INFO [train.py:862] Epoch 21, batch 17000, loss[loss=0.2348, simple_loss=0.4776, pruned_loss=0.1205, ctc_loss=0.1815, over 14762.00 frames. ], tot_loss[loss=0.1811, simple_loss=0.4324, pruned_loss=0.08717, ctc_loss=0.1287, over 2866428.38 frames. ], batch size: 45, lr: 3.00e-03, 
2023-01-06 00:48:02,336 INFO [train.py:862] Epoch 21, batch 17500, loss[loss=0.1955, simple_loss=0.4555, pruned_loss=0.09566, ctc_loss=0.1406, over 14667.00 frames. ], tot_loss[loss=0.1793, simple_loss=0.4315, pruned_loss=0.08626, ctc_loss=0.1267, over 2854658.89 frames. ], batch size: 44, lr: 3.00e-03, 
2023-01-06 00:49:08,094 INFO [train.py:862] Epoch 21, batch 18000, loss[loss=0.1715, simple_loss=0.4585, pruned_loss=0.07351, ctc_loss=0.1152, over 14735.00 frames. ], tot_loss[loss=0.1775, simple_loss=0.4309, pruned_loss=0.08498, ctc_loss=0.1248, over 2861907.03 frames. ], batch size: 45, lr: 3.00e-03, 
2023-01-06 00:49:08,095 INFO [train.py:887] Computing validation loss
2023-01-06 00:49:33,500 INFO [train.py:897] Epoch 21, validation: loss=0.198, simple_loss=0.4492, pruned_loss=0.09802, ctc_loss=0.1446, over 944034.00 frames. 
2023-01-06 00:49:33,500 INFO [train.py:898] Maximum memory allocated so far is 7283MB
2023-01-06 00:50:01,074 INFO [checkpoint.py:75] Saving checkpoint to tiny_transducer_ctc/exp_2m_phone_halfdelay/epoch-21.pt
2023-01-06 00:50:04,173 INFO [train.py:862] Epoch 22, batch 0, loss[loss=0.2292, simple_loss=0.4807, pruned_loss=0.1221, ctc_loss=0.1721, over 14579.00 frames. ], tot_loss[loss=0.2292, simple_loss=0.4807, pruned_loss=0.1221, ctc_loss=0.1721, over 14579.00 frames. ], batch size: 50, lr: 3.00e-03, 
2023-01-06 00:51:09,857 INFO [train.py:862] Epoch 22, batch 500, loss[loss=0.1818, simple_loss=0.4625, pruned_loss=0.08198, ctc_loss=0.1254, over 14852.00 frames. ], tot_loss[loss=0.1829, simple_loss=0.4354, pruned_loss=0.08849, ctc_loss=0.1301, over 2649989.36 frames. ], batch size: 42, lr: 3.00e-03, 
2023-01-06 00:52:15,406 INFO [train.py:862] Epoch 22, batch 1000, loss[loss=0.1823, simple_loss=0.4714, pruned_loss=0.08318, ctc_loss=0.1238, over 14584.00 frames. ], tot_loss[loss=0.1782, simple_loss=0.4311, pruned_loss=0.08527, ctc_loss=0.1257, over 2850159.54 frames. ], batch size: 43, lr: 3.00e-03, 
2023-01-06 00:53:21,367 INFO [train.py:862] Epoch 22, batch 1500, loss[loss=0.2013, simple_loss=0.4461, pruned_loss=0.1039, ctc_loss=0.1475, over 12933.00 frames. ], tot_loss[loss=0.1785, simple_loss=0.4304, pruned_loss=0.08564, ctc_loss=0.1261, over 2881981.82 frames. ], batch size: 77, lr: 3.00e-03, 
2023-01-06 00:54:26,689 INFO [train.py:862] Epoch 22, batch 2000, loss[loss=0.2517, simple_loss=0.5072, pruned_loss=0.1284, ctc_loss=0.1959, over 14846.00 frames. ], tot_loss[loss=0.1793, simple_loss=0.4317, pruned_loss=0.08596, ctc_loss=0.1268, over 2874062.12 frames. ], batch size: 42, lr: 3.00e-03, 
2023-01-06 00:55:32,336 INFO [train.py:862] Epoch 22, batch 2500, loss[loss=0.1951, simple_loss=0.4401, pruned_loss=0.0943, ctc_loss=0.144, over 10330.00 frames. ], tot_loss[loss=0.1788, simple_loss=0.4331, pruned_loss=0.08545, ctc_loss=0.1259, over 2866470.46 frames. ], batch size: 104, lr: 3.00e-03, 
2023-01-06 00:56:38,531 INFO [train.py:862] Epoch 22, batch 3000, loss[loss=0.1474, simple_loss=0.4142, pruned_loss=0.06332, ctc_loss=0.09474, over 14517.00 frames. ], tot_loss[loss=0.1797, simple_loss=0.4322, pruned_loss=0.08644, ctc_loss=0.127, over 2873840.52 frames. ], batch size: 38, lr: 3.00e-03, 
2023-01-06 00:57:44,688 INFO [train.py:862] Epoch 22, batch 3500, loss[loss=0.1606, simple_loss=0.4155, pruned_loss=0.07197, ctc_loss=0.1096, over 13706.00 frames. ], tot_loss[loss=0.1807, simple_loss=0.4322, pruned_loss=0.08701, ctc_loss=0.1282, over 2874541.68 frames. ], batch size: 63, lr: 3.00e-03, 
2023-01-06 00:58:50,241 INFO [train.py:862] Epoch 22, batch 4000, loss[loss=0.2361, simple_loss=0.4575, pruned_loss=0.1266, ctc_loss=0.185, over 10348.00 frames. ], tot_loss[loss=0.179, simple_loss=0.4319, pruned_loss=0.08586, ctc_loss=0.1264, over 2862128.95 frames. ], batch size: 104, lr: 3.00e-03, 
2023-01-06 00:59:56,835 INFO [train.py:862] Epoch 22, batch 4500, loss[loss=0.1896, simple_loss=0.457, pruned_loss=0.08976, ctc_loss=0.1345, over 14725.00 frames. ], tot_loss[loss=0.1802, simple_loss=0.4323, pruned_loss=0.08688, ctc_loss=0.1276, over 2870579.32 frames. ], batch size: 45, lr: 3.00e-03, 
2023-01-06 01:01:02,511 INFO [train.py:862] Epoch 22, batch 5000, loss[loss=0.173, simple_loss=0.4332, pruned_loss=0.08328, ctc_loss=0.1186, over 14879.00 frames. ], tot_loss[loss=0.1793, simple_loss=0.4329, pruned_loss=0.08627, ctc_loss=0.1265, over 2876997.36 frames. ], batch size: 39, lr: 3.00e-03, 
2023-01-06 01:02:08,812 INFO [train.py:862] Epoch 22, batch 5500, loss[loss=0.174, simple_loss=0.3962, pruned_loss=0.0856, ctc_loss=0.127, over 14557.00 frames. ], tot_loss[loss=0.1804, simple_loss=0.4325, pruned_loss=0.08702, ctc_loss=0.1278, over 2861610.10 frames. ], batch size: 34, lr: 3.00e-03, 
2023-01-06 01:03:14,557 INFO [train.py:862] Epoch 22, batch 6000, loss[loss=0.1707, simple_loss=0.4467, pruned_loss=0.07855, ctc_loss=0.1145, over 14730.00 frames. ], tot_loss[loss=0.1799, simple_loss=0.4343, pruned_loss=0.08632, ctc_loss=0.1269, over 2857830.83 frames. ], batch size: 46, lr: 3.00e-03, 
2023-01-06 01:04:20,268 INFO [train.py:862] Epoch 22, batch 6500, loss[loss=0.1575, simple_loss=0.4326, pruned_loss=0.0665, ctc_loss=0.1038, over 14514.00 frames. ], tot_loss[loss=0.1796, simple_loss=0.4319, pruned_loss=0.08658, ctc_loss=0.1269, over 2863260.82 frames. ], batch size: 50, lr: 3.00e-03, 
2023-01-06 01:05:26,045 INFO [train.py:862] Epoch 22, batch 7000, loss[loss=0.1818, simple_loss=0.4299, pruned_loss=0.08935, ctc_loss=0.1293, over 14823.00 frames. ], tot_loss[loss=0.1799, simple_loss=0.4325, pruned_loss=0.08676, ctc_loss=0.1272, over 2860959.53 frames. ], batch size: 41, lr: 3.00e-03, 
2023-01-06 01:06:27,281 INFO [checkpoint.py:75] Saving checkpoint to tiny_transducer_ctc/exp_2m_phone_halfdelay/checkpoint-390000.pt
2023-01-06 01:06:31,889 INFO [train.py:862] Epoch 22, batch 7500, loss[loss=0.1625, simple_loss=0.4155, pruned_loss=0.0761, ctc_loss=0.1106, over 14644.00 frames. ], tot_loss[loss=0.1792, simple_loss=0.4323, pruned_loss=0.08604, ctc_loss=0.1264, over 2866070.98 frames. ], batch size: 40, lr: 3.00e-03, 
2023-01-06 01:07:37,944 INFO [train.py:862] Epoch 22, batch 8000, loss[loss=0.1864, simple_loss=0.4124, pruned_loss=0.09447, ctc_loss=0.1374, over 14527.00 frames. ], tot_loss[loss=0.1794, simple_loss=0.4321, pruned_loss=0.08604, ctc_loss=0.1268, over 2870937.18 frames. ], batch size: 37, lr: 3.00e-03, 
2023-01-06 01:08:43,183 INFO [train.py:862] Epoch 22, batch 8500, loss[loss=0.1635, simple_loss=0.4045, pruned_loss=0.0779, ctc_loss=0.1135, over 14684.00 frames. ], tot_loss[loss=0.1804, simple_loss=0.4327, pruned_loss=0.08682, ctc_loss=0.1278, over 2867812.51 frames. ], batch size: 35, lr: 3.00e-03, 
2023-01-06 01:09:48,349 INFO [train.py:862] Epoch 22, batch 9000, loss[loss=0.1915, simple_loss=0.4648, pruned_loss=0.089, ctc_loss=0.1359, over 13656.00 frames. ], tot_loss[loss=0.1805, simple_loss=0.4335, pruned_loss=0.08664, ctc_loss=0.1279, over 2867348.22 frames. ], batch size: 63, lr: 3.00e-03, 
2023-01-06 01:09:48,350 INFO [train.py:887] Computing validation loss
2023-01-06 01:10:14,522 INFO [train.py:897] Epoch 22, validation: loss=0.1953, simple_loss=0.4477, pruned_loss=0.09615, ctc_loss=0.1418, over 944034.00 frames. 
2023-01-06 01:10:14,522 INFO [train.py:898] Maximum memory allocated so far is 7283MB
2023-01-06 01:11:21,185 INFO [train.py:862] Epoch 22, batch 9500, loss[loss=0.1604, simple_loss=0.4325, pruned_loss=0.07206, ctc_loss=0.1056, over 14187.00 frames. ], tot_loss[loss=0.1793, simple_loss=0.4314, pruned_loss=0.08614, ctc_loss=0.1268, over 2869631.88 frames. ], batch size: 56, lr: 3.00e-03, 
2023-01-06 01:12:28,372 INFO [train.py:862] Epoch 22, batch 10000, loss[loss=0.1647, simple_loss=0.4279, pruned_loss=0.07603, ctc_loss=0.111, over 13999.00 frames. ], tot_loss[loss=0.1787, simple_loss=0.4315, pruned_loss=0.08569, ctc_loss=0.1261, over 2852758.58 frames. ], batch size: 56, lr: 3.00e-03, 
2023-01-06 01:13:35,126 INFO [train.py:862] Epoch 22, batch 10500, loss[loss=0.1655, simple_loss=0.4464, pruned_loss=0.077, ctc_loss=0.1077, over 14649.00 frames. ], tot_loss[loss=0.1781, simple_loss=0.4311, pruned_loss=0.08518, ctc_loss=0.1255, over 2848106.74 frames. ], batch size: 41, lr: 3.00e-03, 
2023-01-06 01:14:41,997 INFO [train.py:862] Epoch 22, batch 11000, loss[loss=0.1393, simple_loss=0.4068, pruned_loss=0.06065, ctc_loss=0.08589, over 14704.00 frames. ], tot_loss[loss=0.1788, simple_loss=0.4325, pruned_loss=0.08576, ctc_loss=0.126, over 2870634.93 frames. ], batch size: 38, lr: 3.00e-03, 
2023-01-06 01:15:49,970 INFO [train.py:862] Epoch 22, batch 11500, loss[loss=0.2024, simple_loss=0.4655, pruned_loss=0.1019, ctc_loss=0.1457, over 14651.00 frames. ], tot_loss[loss=0.1773, simple_loss=0.4303, pruned_loss=0.08509, ctc_loss=0.1246, over 2848568.10 frames. ], batch size: 41, lr: 3.00e-03, 
2023-01-06 01:16:56,835 INFO [train.py:862] Epoch 22, batch 12000, loss[loss=0.1976, simple_loss=0.4459, pruned_loss=0.09438, ctc_loss=0.1462, over 12983.00 frames. ], tot_loss[loss=0.1784, simple_loss=0.4312, pruned_loss=0.08552, ctc_loss=0.1258, over 2864145.44 frames. ], batch size: 76, lr: 3.00e-03, 
2023-01-06 01:18:04,254 INFO [train.py:862] Epoch 22, batch 12500, loss[loss=0.2194, simple_loss=0.4679, pruned_loss=0.1167, ctc_loss=0.1631, over 14119.00 frames. ], tot_loss[loss=0.1821, simple_loss=0.435, pruned_loss=0.08797, ctc_loss=0.1292, over 2859566.75 frames. ], batch size: 56, lr: 3.00e-03, 
2023-01-06 01:19:11,726 INFO [train.py:862] Epoch 22, batch 13000, loss[loss=0.1549, simple_loss=0.411, pruned_loss=0.06902, ctc_loss=0.1036, over 14719.00 frames. ], tot_loss[loss=0.1774, simple_loss=0.4296, pruned_loss=0.08492, ctc_loss=0.125, over 2859078.67 frames. ], batch size: 37, lr: 3.00e-03, 
2023-01-06 01:20:18,920 INFO [train.py:862] Epoch 22, batch 13500, loss[loss=0.2159, simple_loss=0.467, pruned_loss=0.1075, ctc_loss=0.1623, over 10493.00 frames. ], tot_loss[loss=0.1782, simple_loss=0.4323, pruned_loss=0.08529, ctc_loss=0.1254, over 2858585.39 frames. ], batch size: 104, lr: 3.00e-03, 
2023-01-06 01:21:26,343 INFO [train.py:862] Epoch 22, batch 14000, loss[loss=0.2126, simple_loss=0.4477, pruned_loss=0.1146, ctc_loss=0.1587, over 14516.00 frames. ], tot_loss[loss=0.1792, simple_loss=0.4325, pruned_loss=0.08609, ctc_loss=0.1264, over 2863495.87 frames. ], batch size: 37, lr: 3.00e-03, 
2023-01-06 01:22:33,266 INFO [train.py:862] Epoch 22, batch 14500, loss[loss=0.1389, simple_loss=0.3728, pruned_loss=0.06287, ctc_loss=0.09154, over 14776.00 frames. ], tot_loss[loss=0.1791, simple_loss=0.4318, pruned_loss=0.08594, ctc_loss=0.1265, over 2870603.28 frames. ], batch size: 36, lr: 3.00e-03, 
2023-01-06 01:23:40,489 INFO [train.py:862] Epoch 22, batch 15000, loss[loss=0.1606, simple_loss=0.4396, pruned_loss=0.07326, ctc_loss=0.1038, over 14798.00 frames. ], tot_loss[loss=0.1786, simple_loss=0.431, pruned_loss=0.08553, ctc_loss=0.1262, over 2853939.85 frames. ], batch size: 43, lr: 3.00e-03, 
2023-01-06 01:24:47,982 INFO [train.py:862] Epoch 22, batch 15500, loss[loss=0.1521, simple_loss=0.4125, pruned_loss=0.07206, ctc_loss=0.09797, over 14516.00 frames. ], tot_loss[loss=0.1797, simple_loss=0.4319, pruned_loss=0.08637, ctc_loss=0.1272, over 2852103.24 frames. ], batch size: 38, lr: 3.00e-03, 
2023-01-06 01:25:55,363 INFO [train.py:862] Epoch 22, batch 16000, loss[loss=0.1456, simple_loss=0.3919, pruned_loss=0.06699, ctc_loss=0.09528, over 14704.00 frames. ], tot_loss[loss=0.1806, simple_loss=0.4317, pruned_loss=0.08721, ctc_loss=0.1281, over 2853268.73 frames. ], batch size: 37, lr: 3.00e-03, 
2023-01-06 01:27:02,288 INFO [train.py:862] Epoch 22, batch 16500, loss[loss=0.1942, simple_loss=0.4497, pruned_loss=0.09773, ctc_loss=0.1391, over 12934.00 frames. ], tot_loss[loss=0.1762, simple_loss=0.4297, pruned_loss=0.08402, ctc_loss=0.1236, over 2873069.58 frames. ], batch size: 76, lr: 3.00e-03, 
2023-01-06 01:28:08,717 INFO [train.py:862] Epoch 22, batch 17000, loss[loss=0.1352, simple_loss=0.395, pruned_loss=0.05643, ctc_loss=0.08427, over 14694.00 frames. ], tot_loss[loss=0.1788, simple_loss=0.4328, pruned_loss=0.0856, ctc_loss=0.126, over 2852323.60 frames. ], batch size: 38, lr: 3.00e-03, 
2023-01-06 01:29:11,900 INFO [checkpoint.py:75] Saving checkpoint to tiny_transducer_ctc/exp_2m_phone_halfdelay/checkpoint-400000.pt
2023-01-06 01:29:16,477 INFO [train.py:862] Epoch 22, batch 17500, loss[loss=0.1605, simple_loss=0.4279, pruned_loss=0.07273, ctc_loss=0.1064, over 14733.00 frames. ], tot_loss[loss=0.179, simple_loss=0.432, pruned_loss=0.08588, ctc_loss=0.1263, over 2858424.86 frames. ], batch size: 46, lr: 3.00e-03, 
2023-01-06 01:30:21,705 INFO [train.py:862] Epoch 22, batch 18000, loss[loss=0.1709, simple_loss=0.3998, pruned_loss=0.08487, ctc_loss=0.122, over 14764.00 frames. ], tot_loss[loss=0.1808, simple_loss=0.4328, pruned_loss=0.08739, ctc_loss=0.1281, over 2846041.46 frames. ], batch size: 36, lr: 3.00e-03, 
2023-01-06 01:30:21,706 INFO [train.py:887] Computing validation loss
2023-01-06 01:30:47,625 INFO [train.py:897] Epoch 22, validation: loss=0.2085, simple_loss=0.4549, pruned_loss=0.1045, ctc_loss=0.1556, over 944034.00 frames. 
2023-01-06 01:30:47,626 INFO [train.py:898] Maximum memory allocated so far is 7283MB
2023-01-06 01:31:15,313 INFO [checkpoint.py:75] Saving checkpoint to tiny_transducer_ctc/exp_2m_phone_halfdelay/epoch-22.pt
2023-01-06 01:31:18,312 INFO [train.py:862] Epoch 23, batch 0, loss[loss=0.2025, simple_loss=0.4254, pruned_loss=0.1052, ctc_loss=0.153, over 13660.00 frames. ], tot_loss[loss=0.2025, simple_loss=0.4254, pruned_loss=0.1052, ctc_loss=0.153, over 13660.00 frames. ], batch size: 30, lr: 3.00e-03, 
2023-01-06 01:32:25,705 INFO [train.py:862] Epoch 23, batch 500, loss[loss=0.1845, simple_loss=0.408, pruned_loss=0.09446, ctc_loss=0.1357, over 14647.00 frames. ], tot_loss[loss=0.1803, simple_loss=0.4333, pruned_loss=0.08671, ctc_loss=0.1276, over 2637800.74 frames. ], batch size: 35, lr: 3.00e-03, 
2023-01-06 01:33:33,590 INFO [train.py:862] Epoch 23, batch 1000, loss[loss=0.1507, simple_loss=0.4019, pruned_loss=0.0655, ctc_loss=0.1011, over 14518.00 frames. ], tot_loss[loss=0.1788, simple_loss=0.4321, pruned_loss=0.0854, ctc_loss=0.1262, over 2870230.84 frames. ], batch size: 36, lr: 3.00e-03, 
2023-01-06 01:34:41,941 INFO [train.py:862] Epoch 23, batch 1500, loss[loss=0.1613, simple_loss=0.4432, pruned_loss=0.06717, ctc_loss=0.1066, over 14676.00 frames. ], tot_loss[loss=0.1803, simple_loss=0.4321, pruned_loss=0.08701, ctc_loss=0.1277, over 2858887.43 frames. ], batch size: 47, lr: 3.00e-03, 
2023-01-06 01:35:49,750 INFO [train.py:862] Epoch 23, batch 2000, loss[loss=0.2368, simple_loss=0.4808, pruned_loss=0.124, ctc_loss=0.1821, over 14528.00 frames. ], tot_loss[loss=0.1777, simple_loss=0.4312, pruned_loss=0.08468, ctc_loss=0.1251, over 2869144.41 frames. ], batch size: 48, lr: 3.00e-03, 
2023-01-06 01:36:58,666 INFO [train.py:862] Epoch 23, batch 2500, loss[loss=0.1642, simple_loss=0.4227, pruned_loss=0.07778, ctc_loss=0.1107, over 14704.00 frames. ], tot_loss[loss=0.1766, simple_loss=0.4291, pruned_loss=0.08442, ctc_loss=0.1241, over 2874600.64 frames. ], batch size: 39, lr: 3.00e-03, 
2023-01-06 01:38:06,749 INFO [train.py:862] Epoch 23, batch 3000, loss[loss=0.1855, simple_loss=0.4307, pruned_loss=0.08824, ctc_loss=0.1349, over 12846.00 frames. ], tot_loss[loss=0.179, simple_loss=0.4319, pruned_loss=0.08574, ctc_loss=0.1264, over 2881710.63 frames. ], batch size: 76, lr: 3.00e-03, 
2023-01-06 01:39:14,425 INFO [train.py:862] Epoch 23, batch 3500, loss[loss=0.1526, simple_loss=0.3776, pruned_loss=0.07264, ctc_loss=0.106, over 14550.00 frames. ], tot_loss[loss=0.1799, simple_loss=0.433, pruned_loss=0.08636, ctc_loss=0.1271, over 2886156.42 frames. ], batch size: 34, lr: 3.00e-03, 
2023-01-06 01:40:22,128 INFO [train.py:862] Epoch 23, batch 4000, loss[loss=0.1538, simple_loss=0.3959, pruned_loss=0.07115, ctc_loss=0.1045, over 14807.00 frames. ], tot_loss[loss=0.1784, simple_loss=0.4307, pruned_loss=0.08565, ctc_loss=0.1259, over 2879188.91 frames. ], batch size: 36, lr: 3.00e-03, 
2023-01-06 01:41:30,894 INFO [train.py:862] Epoch 23, batch 4500, loss[loss=0.1558, simple_loss=0.3911, pruned_loss=0.07702, ctc_loss=0.1058, over 14515.00 frames. ], tot_loss[loss=0.1783, simple_loss=0.4294, pruned_loss=0.08554, ctc_loss=0.126, over 2861644.76 frames. ], batch size: 36, lr: 3.00e-03, 
2023-01-06 01:42:38,714 INFO [train.py:862] Epoch 23, batch 5000, loss[loss=0.1447, simple_loss=0.3986, pruned_loss=0.06433, ctc_loss=0.09375, over 14770.00 frames. ], tot_loss[loss=0.1792, simple_loss=0.4314, pruned_loss=0.08596, ctc_loss=0.1266, over 2866453.03 frames. ], batch size: 36, lr: 3.00e-03, 
2023-01-06 01:43:46,493 INFO [train.py:862] Epoch 23, batch 5500, loss[loss=0.1686, simple_loss=0.4273, pruned_loss=0.07486, ctc_loss=0.1173, over 14802.00 frames. ], tot_loss[loss=0.1786, simple_loss=0.4302, pruned_loss=0.08573, ctc_loss=0.1262, over 2856625.68 frames. ], batch size: 43, lr: 3.00e-03, 
2023-01-06 01:44:54,704 INFO [train.py:862] Epoch 23, batch 6000, loss[loss=0.1972, simple_loss=0.4515, pruned_loss=0.09392, ctc_loss=0.1447, over 10353.00 frames. ], tot_loss[loss=0.1775, simple_loss=0.4306, pruned_loss=0.08482, ctc_loss=0.1249, over 2855148.12 frames. ], batch size: 105, lr: 3.00e-03, 
2023-01-06 01:46:02,343 INFO [train.py:862] Epoch 23, batch 6500, loss[loss=0.161, simple_loss=0.4045, pruned_loss=0.07681, ctc_loss=0.1105, over 14531.00 frames. ], tot_loss[loss=0.1791, simple_loss=0.4311, pruned_loss=0.0862, ctc_loss=0.1266, over 2860277.43 frames. ], batch size: 36, lr: 3.00e-03, 
2023-01-06 01:47:10,820 INFO [train.py:862] Epoch 23, batch 7000, loss[loss=0.1521, simple_loss=0.4117, pruned_loss=0.06983, ctc_loss=0.09913, over 14708.00 frames. ], tot_loss[loss=0.1771, simple_loss=0.4289, pruned_loss=0.08484, ctc_loss=0.1247, over 2837602.90 frames. ], batch size: 38, lr: 3.00e-03, 
2023-01-06 01:48:18,962 INFO [train.py:862] Epoch 23, batch 7500, loss[loss=0.1807, simple_loss=0.439, pruned_loss=0.08813, ctc_loss=0.1262, over 14646.00 frames. ], tot_loss[loss=0.1787, simple_loss=0.4321, pruned_loss=0.08588, ctc_loss=0.1259, over 2854639.80 frames. ], batch size: 41, lr: 3.00e-03, 
2023-01-06 01:49:27,179 INFO [train.py:862] Epoch 23, batch 8000, loss[loss=0.2043, simple_loss=0.4637, pruned_loss=0.09657, ctc_loss=0.1511, over 14686.00 frames. ], tot_loss[loss=0.1786, simple_loss=0.4311, pruned_loss=0.08555, ctc_loss=0.126, over 2863064.02 frames. ], batch size: 47, lr: 3.00e-03, 
2023-01-06 01:50:34,686 INFO [train.py:862] Epoch 23, batch 8500, loss[loss=0.1457, simple_loss=0.4216, pruned_loss=0.06351, ctc_loss=0.0906, over 14646.00 frames. ], tot_loss[loss=0.1775, simple_loss=0.43, pruned_loss=0.08515, ctc_loss=0.125, over 2875806.11 frames. ], batch size: 40, lr: 3.00e-03, 
2023-01-06 01:51:42,347 INFO [train.py:862] Epoch 23, batch 9000, loss[loss=0.1846, simple_loss=0.4543, pruned_loss=0.08761, ctc_loss=0.1288, over 14497.00 frames. ], tot_loss[loss=0.1781, simple_loss=0.43, pruned_loss=0.08542, ctc_loss=0.1256, over 2860721.19 frames. ], batch size: 50, lr: 3.00e-03, 
2023-01-06 01:51:42,348 INFO [train.py:887] Computing validation loss
2023-01-06 01:52:08,771 INFO [train.py:897] Epoch 23, validation: loss=0.1963, simple_loss=0.4479, pruned_loss=0.09724, ctc_loss=0.1427, over 944034.00 frames. 
2023-01-06 01:52:08,772 INFO [train.py:898] Maximum memory allocated so far is 7351MB
2023-01-06 01:52:43,037 INFO [checkpoint.py:75] Saving checkpoint to tiny_transducer_ctc/exp_2m_phone_halfdelay/checkpoint-410000.pt
2023-01-06 01:53:17,832 INFO [train.py:862] Epoch 23, batch 9500, loss[loss=0.1708, simple_loss=0.425, pruned_loss=0.08068, ctc_loss=0.1184, over 14664.00 frames. ], tot_loss[loss=0.1799, simple_loss=0.4315, pruned_loss=0.08651, ctc_loss=0.1274, over 2850275.02 frames. ], batch size: 40, lr: 3.00e-03, 
2023-01-06 01:54:27,603 INFO [train.py:862] Epoch 23, batch 10000, loss[loss=0.1744, simple_loss=0.4021, pruned_loss=0.08513, ctc_loss=0.1265, over 14667.00 frames. ], tot_loss[loss=0.1784, simple_loss=0.4314, pruned_loss=0.08543, ctc_loss=0.1258, over 2869555.75 frames. ], batch size: 35, lr: 3.00e-03, 
2023-01-06 01:55:36,069 INFO [train.py:862] Epoch 23, batch 10500, loss[loss=0.1454, simple_loss=0.3918, pruned_loss=0.06676, ctc_loss=0.09512, over 14701.00 frames. ], tot_loss[loss=0.1774, simple_loss=0.4299, pruned_loss=0.08484, ctc_loss=0.1249, over 2866037.45 frames. ], batch size: 35, lr: 3.00e-03, 
2023-01-06 01:56:45,886 INFO [train.py:862] Epoch 23, batch 11000, loss[loss=0.2043, simple_loss=0.4832, pruned_loss=0.1013, ctc_loss=0.145, over 14496.00 frames. ], tot_loss[loss=0.1775, simple_loss=0.431, pruned_loss=0.08477, ctc_loss=0.1248, over 2862902.02 frames. ], batch size: 48, lr: 3.00e-03, 
2023-01-06 01:57:54,603 INFO [train.py:862] Epoch 23, batch 11500, loss[loss=0.1356, simple_loss=0.3834, pruned_loss=0.05811, ctc_loss=0.0867, over 14813.00 frames. ], tot_loss[loss=0.1774, simple_loss=0.4306, pruned_loss=0.08482, ctc_loss=0.1248, over 2859263.41 frames. ], batch size: 36, lr: 3.00e-03, 
2023-01-06 01:59:03,613 INFO [train.py:862] Epoch 23, batch 12000, loss[loss=0.2217, simple_loss=0.4677, pruned_loss=0.112, ctc_loss=0.1685, over 10233.00 frames. ], tot_loss[loss=0.1776, simple_loss=0.431, pruned_loss=0.08495, ctc_loss=0.125, over 2862501.12 frames. ], batch size: 106, lr: 3.00e-03, 
2023-01-06 02:00:13,199 INFO [train.py:862] Epoch 23, batch 12500, loss[loss=0.2325, simple_loss=0.4718, pruned_loss=0.1249, ctc_loss=0.1775, over 13649.00 frames. ], tot_loss[loss=0.179, simple_loss=0.432, pruned_loss=0.08583, ctc_loss=0.1264, over 2859724.71 frames. ], batch size: 63, lr: 3.00e-03, 
2023-01-06 02:01:22,647 INFO [train.py:862] Epoch 23, batch 13000, loss[loss=0.1365, simple_loss=0.3686, pruned_loss=0.06294, ctc_loss=0.08907, over 14519.00 frames. ], tot_loss[loss=0.178, simple_loss=0.4327, pruned_loss=0.08522, ctc_loss=0.1251, over 2859885.49 frames. ], batch size: 34, lr: 3.00e-03, 
2023-01-06 02:02:32,253 INFO [train.py:862] Epoch 23, batch 13500, loss[loss=0.2366, simple_loss=0.483, pruned_loss=0.1236, ctc_loss=0.1815, over 14698.00 frames. ], tot_loss[loss=0.1782, simple_loss=0.4313, pruned_loss=0.0853, ctc_loss=0.1255, over 2865290.79 frames. ], batch size: 44, lr: 3.00e-03, 
2023-01-06 02:03:40,860 INFO [train.py:862] Epoch 23, batch 14000, loss[loss=0.1883, simple_loss=0.4495, pruned_loss=0.08955, ctc_loss=0.1343, over 13521.00 frames. ], tot_loss[loss=0.1792, simple_loss=0.4318, pruned_loss=0.08616, ctc_loss=0.1266, over 2862312.72 frames. ], batch size: 63, lr: 3.00e-03, 
2023-01-06 02:04:50,353 INFO [train.py:862] Epoch 23, batch 14500, loss[loss=0.1641, simple_loss=0.4198, pruned_loss=0.07558, ctc_loss=0.1121, over 12852.00 frames. ], tot_loss[loss=0.1769, simple_loss=0.4298, pruned_loss=0.08442, ctc_loss=0.1245, over 2857998.59 frames. ], batch size: 76, lr: 3.00e-03, 
2023-01-06 02:05:57,760 INFO [train.py:862] Epoch 23, batch 15000, loss[loss=0.1587, simple_loss=0.4135, pruned_loss=0.07564, ctc_loss=0.1058, over 14869.00 frames. ], tot_loss[loss=0.1792, simple_loss=0.431, pruned_loss=0.08611, ctc_loss=0.1267, over 2861911.41 frames. ], batch size: 40, lr: 3.00e-03, 
2023-01-06 02:07:05,341 INFO [train.py:862] Epoch 23, batch 15500, loss[loss=0.1695, simple_loss=0.4516, pruned_loss=0.07843, ctc_loss=0.1117, over 14576.00 frames. ], tot_loss[loss=0.1779, simple_loss=0.4318, pruned_loss=0.08519, ctc_loss=0.1251, over 2863314.74 frames. ], batch size: 43, lr: 3.00e-03, 
2023-01-06 02:08:13,387 INFO [train.py:862] Epoch 23, batch 16000, loss[loss=0.1519, simple_loss=0.4206, pruned_loss=0.06631, ctc_loss=0.09847, over 14674.00 frames. ], tot_loss[loss=0.1782, simple_loss=0.4304, pruned_loss=0.08548, ctc_loss=0.1258, over 2858384.39 frames. ], batch size: 40, lr: 3.00e-03, 
2023-01-06 02:09:20,179 INFO [train.py:862] Epoch 23, batch 16500, loss[loss=0.1865, simple_loss=0.4391, pruned_loss=0.09214, ctc_loss=0.1328, over 14778.00 frames. ], tot_loss[loss=0.1775, simple_loss=0.4313, pruned_loss=0.08492, ctc_loss=0.1248, over 2875923.31 frames. ], batch size: 43, lr: 3.00e-03, 
2023-01-06 02:10:27,061 INFO [train.py:862] Epoch 23, batch 17000, loss[loss=0.181, simple_loss=0.424, pruned_loss=0.09142, ctc_loss=0.1285, over 14690.00 frames. ], tot_loss[loss=0.1798, simple_loss=0.4317, pruned_loss=0.08648, ctc_loss=0.1273, over 2859750.40 frames. ], batch size: 39, lr: 3.00e-03, 
2023-01-06 02:11:33,576 INFO [train.py:862] Epoch 23, batch 17500, loss[loss=0.1911, simple_loss=0.4432, pruned_loss=0.09423, ctc_loss=0.1376, over 12852.00 frames. ], tot_loss[loss=0.1773, simple_loss=0.4301, pruned_loss=0.08475, ctc_loss=0.1247, over 2871033.08 frames. ], batch size: 76, lr: 3.00e-03, 
2023-01-06 02:12:38,603 INFO [train.py:862] Epoch 23, batch 18000, loss[loss=0.2044, simple_loss=0.4286, pruned_loss=0.1089, ctc_loss=0.1535, over 14422.00 frames. ], tot_loss[loss=0.1791, simple_loss=0.4312, pruned_loss=0.08604, ctc_loss=0.1265, over 2850032.08 frames. ], batch size: 33, lr: 3.00e-03, 
2023-01-06 02:12:38,603 INFO [train.py:887] Computing validation loss
2023-01-06 02:13:04,449 INFO [train.py:897] Epoch 23, validation: loss=0.1964, simple_loss=0.4478, pruned_loss=0.09689, ctc_loss=0.143, over 944034.00 frames. 
2023-01-06 02:13:04,450 INFO [train.py:898] Maximum memory allocated so far is 7351MB
2023-01-06 02:13:31,915 INFO [checkpoint.py:75] Saving checkpoint to tiny_transducer_ctc/exp_2m_phone_halfdelay/epoch-23.pt
2023-01-06 02:13:34,695 INFO [train.py:862] Epoch 24, batch 0, loss[loss=0.1776, simple_loss=0.4106, pruned_loss=0.08741, ctc_loss=0.1283, over 14016.00 frames. ], tot_loss[loss=0.1776, simple_loss=0.4106, pruned_loss=0.08741, ctc_loss=0.1283, over 14016.00 frames. ], batch size: 31, lr: 3.00e-03, 
2023-01-06 02:14:40,549 INFO [train.py:862] Epoch 24, batch 500, loss[loss=0.183, simple_loss=0.4517, pruned_loss=0.08674, ctc_loss=0.1274, over 14542.00 frames. ], tot_loss[loss=0.179, simple_loss=0.4327, pruned_loss=0.08579, ctc_loss=0.1262, over 2639178.41 frames. ], batch size: 50, lr: 3.00e-03, 
2023-01-06 02:15:47,530 INFO [train.py:862] Epoch 24, batch 1000, loss[loss=0.2059, simple_loss=0.474, pruned_loss=0.1037, ctc_loss=0.1481, over 12975.00 frames. ], tot_loss[loss=0.1787, simple_loss=0.4309, pruned_loss=0.08565, ctc_loss=0.1263, over 2852001.04 frames. ], batch size: 76, lr: 3.00e-03, 
2023-01-06 02:15:51,576 INFO [checkpoint.py:75] Saving checkpoint to tiny_transducer_ctc/exp_2m_phone_halfdelay/checkpoint-420000.pt
2023-01-06 02:16:53,811 INFO [train.py:862] Epoch 24, batch 1500, loss[loss=0.1653, simple_loss=0.4193, pruned_loss=0.07327, ctc_loss=0.1149, over 14807.00 frames. ], tot_loss[loss=0.1779, simple_loss=0.4301, pruned_loss=0.0851, ctc_loss=0.1255, over 2872326.81 frames. ], batch size: 41, lr: 3.00e-03, 
2023-01-06 02:18:00,717 INFO [train.py:862] Epoch 24, batch 2000, loss[loss=0.1734, simple_loss=0.4469, pruned_loss=0.08257, ctc_loss=0.1165, over 14654.00 frames. ], tot_loss[loss=0.179, simple_loss=0.4321, pruned_loss=0.08597, ctc_loss=0.1263, over 2870372.16 frames. ], batch size: 40, lr: 3.00e-03, 
2023-01-06 02:19:07,219 INFO [train.py:862] Epoch 24, batch 2500, loss[loss=0.1493, simple_loss=0.4322, pruned_loss=0.06177, ctc_loss=0.09414, over 14646.00 frames. ], tot_loss[loss=0.1781, simple_loss=0.4331, pruned_loss=0.08491, ctc_loss=0.1252, over 2859156.55 frames. ], batch size: 41, lr: 3.00e-03, 
2023-01-06 02:20:14,209 INFO [train.py:862] Epoch 24, batch 3000, loss[loss=0.1914, simple_loss=0.4548, pruned_loss=0.09462, ctc_loss=0.1355, over 14527.00 frames. ], tot_loss[loss=0.1779, simple_loss=0.4318, pruned_loss=0.08495, ctc_loss=0.1253, over 2869410.47 frames. ], batch size: 53, lr: 3.00e-03, 
2023-01-06 02:21:21,844 INFO [train.py:862] Epoch 24, batch 3500, loss[loss=0.2083, simple_loss=0.4567, pruned_loss=0.107, ctc_loss=0.1538, over 14661.00 frames. ], tot_loss[loss=0.1775, simple_loss=0.4299, pruned_loss=0.08476, ctc_loss=0.1251, over 2861277.98 frames. ], batch size: 41, lr: 3.00e-03, 
2023-01-06 02:22:28,583 INFO [train.py:862] Epoch 24, batch 4000, loss[loss=0.1627, simple_loss=0.4061, pruned_loss=0.07362, ctc_loss=0.1138, over 14892.00 frames. ], tot_loss[loss=0.1779, simple_loss=0.4302, pruned_loss=0.08518, ctc_loss=0.1254, over 2871023.84 frames. ], batch size: 39, lr: 3.00e-03, 
2023-01-06 02:23:35,677 INFO [train.py:862] Epoch 24, batch 4500, loss[loss=0.218, simple_loss=0.4659, pruned_loss=0.1147, ctc_loss=0.1624, over 9952.00 frames. ], tot_loss[loss=0.179, simple_loss=0.4319, pruned_loss=0.0857, ctc_loss=0.1264, over 2868249.14 frames. ], batch size: 104, lr: 3.00e-03, 
2023-01-06 02:24:42,089 INFO [train.py:862] Epoch 24, batch 5000, loss[loss=0.1752, simple_loss=0.4452, pruned_loss=0.08178, ctc_loss=0.1198, over 14666.00 frames. ], tot_loss[loss=0.1775, simple_loss=0.4307, pruned_loss=0.0848, ctc_loss=0.1249, over 2876769.96 frames. ], batch size: 42, lr: 3.00e-03, 
2023-01-06 02:25:48,644 INFO [train.py:862] Epoch 24, batch 5500, loss[loss=0.2026, simple_loss=0.451, pruned_loss=0.1008, ctc_loss=0.1497, over 13585.00 frames. ], tot_loss[loss=0.1778, simple_loss=0.4306, pruned_loss=0.08518, ctc_loss=0.1253, over 2869119.86 frames. ], batch size: 63, lr: 3.00e-03, 
2023-01-06 02:26:55,561 INFO [train.py:862] Epoch 24, batch 6000, loss[loss=0.1945, simple_loss=0.4475, pruned_loss=0.09212, ctc_loss=0.1425, over 14249.00 frames. ], tot_loss[loss=0.1784, simple_loss=0.4315, pruned_loss=0.08558, ctc_loss=0.1257, over 2881725.33 frames. ], batch size: 52, lr: 3.00e-03, 
2023-01-06 02:28:02,132 INFO [train.py:862] Epoch 24, batch 6500, loss[loss=0.2055, simple_loss=0.4654, pruned_loss=0.1024, ctc_loss=0.15, over 14674.00 frames. ], tot_loss[loss=0.1752, simple_loss=0.4291, pruned_loss=0.08347, ctc_loss=0.1226, over 2882571.25 frames. ], batch size: 42, lr: 3.00e-03, 
2023-01-06 02:29:08,785 INFO [train.py:862] Epoch 24, batch 7000, loss[loss=0.1926, simple_loss=0.4419, pruned_loss=0.0947, ctc_loss=0.1399, over 14582.00 frames. ], tot_loss[loss=0.1792, simple_loss=0.4306, pruned_loss=0.08648, ctc_loss=0.1266, over 2857073.57 frames. ], batch size: 43, lr: 3.00e-03, 
2023-01-06 02:30:16,046 INFO [train.py:862] Epoch 24, batch 7500, loss[loss=0.1676, simple_loss=0.4123, pruned_loss=0.08114, ctc_loss=0.1163, over 14545.00 frames. ], tot_loss[loss=0.1774, simple_loss=0.4311, pruned_loss=0.0848, ctc_loss=0.1248, over 2870077.91 frames. ], batch size: 36, lr: 3.00e-03, 
2023-01-06 02:31:22,427 INFO [train.py:862] Epoch 24, batch 8000, loss[loss=0.181, simple_loss=0.4498, pruned_loss=0.0896, ctc_loss=0.1238, over 14122.00 frames. ], tot_loss[loss=0.1757, simple_loss=0.4298, pruned_loss=0.08344, ctc_loss=0.1232, over 2859754.02 frames. ], batch size: 56, lr: 3.00e-03, 
2023-01-06 02:32:29,571 INFO [train.py:862] Epoch 24, batch 8500, loss[loss=0.1643, simple_loss=0.4389, pruned_loss=0.07368, ctc_loss=0.1091, over 14545.00 frames. ], tot_loss[loss=0.1786, simple_loss=0.4312, pruned_loss=0.08579, ctc_loss=0.126, over 2846488.25 frames. ], batch size: 48, lr: 3.00e-03, 
2023-01-06 02:33:36,140 INFO [train.py:862] Epoch 24, batch 9000, loss[loss=0.1729, simple_loss=0.4364, pruned_loss=0.07876, ctc_loss=0.1197, over 14655.00 frames. ], tot_loss[loss=0.178, simple_loss=0.4297, pruned_loss=0.08545, ctc_loss=0.1256, over 2861714.05 frames. ], batch size: 40, lr: 3.00e-03, 
2023-01-06 02:33:36,141 INFO [train.py:887] Computing validation loss
2023-01-06 02:34:02,573 INFO [train.py:897] Epoch 24, validation: loss=0.1942, simple_loss=0.4466, pruned_loss=0.09568, ctc_loss=0.1408, over 944034.00 frames. 
2023-01-06 02:34:02,573 INFO [train.py:898] Maximum memory allocated so far is 7351MB
2023-01-06 02:35:09,055 INFO [train.py:862] Epoch 24, batch 9500, loss[loss=0.1617, simple_loss=0.4255, pruned_loss=0.07433, ctc_loss=0.108, over 14675.00 frames. ], tot_loss[loss=0.1769, simple_loss=0.429, pruned_loss=0.08464, ctc_loss=0.1245, over 2865903.20 frames. ], batch size: 47, lr: 3.00e-03, 
2023-01-06 02:36:16,152 INFO [train.py:862] Epoch 24, batch 10000, loss[loss=0.1824, simple_loss=0.4589, pruned_loss=0.08661, ctc_loss=0.1252, over 14584.00 frames. ], tot_loss[loss=0.1777, simple_loss=0.4313, pruned_loss=0.08509, ctc_loss=0.1249, over 2853069.61 frames. ], batch size: 48, lr: 3.00e-03, 
2023-01-06 02:37:23,086 INFO [train.py:862] Epoch 24, batch 10500, loss[loss=0.1608, simple_loss=0.4244, pruned_loss=0.07194, ctc_loss=0.1079, over 14693.00 frames. ], tot_loss[loss=0.179, simple_loss=0.4306, pruned_loss=0.08612, ctc_loss=0.1266, over 2839550.56 frames. ], batch size: 38, lr: 3.00e-03, 
2023-01-06 02:38:30,129 INFO [train.py:862] Epoch 24, batch 11000, loss[loss=0.1754, simple_loss=0.427, pruned_loss=0.0806, ctc_loss=0.1245, over 14750.00 frames. ], tot_loss[loss=0.1795, simple_loss=0.4325, pruned_loss=0.08601, ctc_loss=0.1269, over 2825771.63 frames. ], batch size: 45, lr: 3.00e-03, 
2023-01-06 02:38:34,295 INFO [checkpoint.py:75] Saving checkpoint to tiny_transducer_ctc/exp_2m_phone_halfdelay/checkpoint-430000.pt
2023-01-06 02:39:38,236 INFO [train.py:862] Epoch 24, batch 11500, loss[loss=0.15, simple_loss=0.3957, pruned_loss=0.07084, ctc_loss=0.09906, over 14796.00 frames. ], tot_loss[loss=0.1805, simple_loss=0.4337, pruned_loss=0.08695, ctc_loss=0.1277, over 2871536.01 frames. ], batch size: 36, lr: 3.00e-03, 
2023-01-06 02:40:45,625 INFO [train.py:862] Epoch 24, batch 12000, loss[loss=0.1772, simple_loss=0.3945, pruned_loss=0.08672, ctc_loss=0.1315, over 14456.00 frames. ], tot_loss[loss=0.1788, simple_loss=0.4313, pruned_loss=0.08576, ctc_loss=0.1262, over 2872075.25 frames. ], batch size: 33, lr: 3.00e-03, 
2023-01-06 02:41:53,539 INFO [train.py:862] Epoch 24, batch 12500, loss[loss=0.1931, simple_loss=0.4481, pruned_loss=0.09152, ctc_loss=0.1406, over 13635.00 frames. ], tot_loss[loss=0.1798, simple_loss=0.4326, pruned_loss=0.08638, ctc_loss=0.1272, over 2847704.34 frames. ], batch size: 63, lr: 3.00e-03, 
2023-01-06 02:43:00,485 INFO [train.py:862] Epoch 24, batch 13000, loss[loss=0.1332, simple_loss=0.3711, pruned_loss=0.06056, ctc_loss=0.08474, over 14527.00 frames. ], tot_loss[loss=0.1783, simple_loss=0.4311, pruned_loss=0.08546, ctc_loss=0.1257, over 2860737.66 frames. ], batch size: 34, lr: 3.00e-03, 
2023-01-06 02:44:07,951 INFO [train.py:862] Epoch 24, batch 13500, loss[loss=0.1964, simple_loss=0.463, pruned_loss=0.09655, ctc_loss=0.14, over 14650.00 frames. ], tot_loss[loss=0.1774, simple_loss=0.4308, pruned_loss=0.08479, ctc_loss=0.1248, over 2863933.72 frames. ], batch size: 40, lr: 3.00e-03, 
2023-01-06 02:45:15,151 INFO [train.py:862] Epoch 24, batch 14000, loss[loss=0.173, simple_loss=0.456, pruned_loss=0.07568, ctc_loss=0.1169, over 14549.00 frames. ], tot_loss[loss=0.175, simple_loss=0.4284, pruned_loss=0.08318, ctc_loss=0.1226, over 2868521.56 frames. ], batch size: 50, lr: 3.00e-03, 
2023-01-06 02:46:22,175 INFO [train.py:862] Epoch 24, batch 14500, loss[loss=0.1719, simple_loss=0.4344, pruned_loss=0.08081, ctc_loss=0.1179, over 13161.00 frames. ], tot_loss[loss=0.1766, simple_loss=0.4297, pruned_loss=0.08421, ctc_loss=0.1241, over 2875519.97 frames. ], batch size: 76, lr: 3.00e-03, 
2023-01-06 02:47:29,639 INFO [train.py:862] Epoch 24, batch 15000, loss[loss=0.2455, simple_loss=0.4847, pruned_loss=0.1344, ctc_loss=0.1892, over 14652.00 frames. ], tot_loss[loss=0.1773, simple_loss=0.4292, pruned_loss=0.08497, ctc_loss=0.125, over 2858896.51 frames. ], batch size: 41, lr: 3.00e-03, 
2023-01-06 02:48:36,821 INFO [train.py:862] Epoch 24, batch 15500, loss[loss=0.171, simple_loss=0.3957, pruned_loss=0.08327, ctc_loss=0.1238, over 14535.00 frames. ], tot_loss[loss=0.1789, simple_loss=0.4315, pruned_loss=0.08578, ctc_loss=0.1263, over 2865365.43 frames. ], batch size: 34, lr: 3.00e-03, 
2023-01-06 02:49:43,612 INFO [train.py:862] Epoch 24, batch 16000, loss[loss=0.1867, simple_loss=0.4533, pruned_loss=0.09003, ctc_loss=0.1309, over 14689.00 frames. ], tot_loss[loss=0.1779, simple_loss=0.4308, pruned_loss=0.08509, ctc_loss=0.1254, over 2874276.50 frames. ], batch size: 47, lr: 3.00e-03, 
2023-01-06 02:50:50,953 INFO [train.py:862] Epoch 24, batch 16500, loss[loss=0.1938, simple_loss=0.458, pruned_loss=0.09557, ctc_loss=0.1378, over 14414.00 frames. ], tot_loss[loss=0.1768, simple_loss=0.4299, pruned_loss=0.08446, ctc_loss=0.1242, over 2833256.90 frames. ], batch size: 50, lr: 3.00e-03, 
2023-01-06 02:51:58,600 INFO [train.py:862] Epoch 24, batch 17000, loss[loss=0.1577, simple_loss=0.3942, pruned_loss=0.07576, ctc_loss=0.1084, over 14791.00 frames. ], tot_loss[loss=0.1779, simple_loss=0.4306, pruned_loss=0.08521, ctc_loss=0.1254, over 2859236.32 frames. ], batch size: 36, lr: 3.00e-03, 
2023-01-06 02:53:06,255 INFO [train.py:862] Epoch 24, batch 17500, loss[loss=0.2203, simple_loss=0.4691, pruned_loss=0.1076, ctc_loss=0.168, over 14845.00 frames. ], tot_loss[loss=0.1776, simple_loss=0.4309, pruned_loss=0.08495, ctc_loss=0.125, over 2845919.98 frames. ], batch size: 40, lr: 3.00e-03, 
2023-01-06 02:54:11,891 INFO [train.py:862] Epoch 24, batch 18000, loss[loss=0.1836, simple_loss=0.439, pruned_loss=0.0914, ctc_loss=0.129, over 13098.00 frames. ], tot_loss[loss=0.1761, simple_loss=0.4299, pruned_loss=0.08395, ctc_loss=0.1235, over 2873306.41 frames. ], batch size: 76, lr: 3.00e-03, 
2023-01-06 02:54:11,892 INFO [train.py:887] Computing validation loss
2023-01-06 02:54:38,224 INFO [train.py:897] Epoch 24, validation: loss=0.1957, simple_loss=0.4471, pruned_loss=0.09646, ctc_loss=0.1424, over 944034.00 frames. 
2023-01-06 02:54:38,224 INFO [train.py:898] Maximum memory allocated so far is 7351MB
2023-01-06 02:55:05,752 INFO [checkpoint.py:75] Saving checkpoint to tiny_transducer_ctc/exp_2m_phone_halfdelay/epoch-24.pt
2023-01-06 02:55:08,885 INFO [train.py:862] Epoch 25, batch 0, loss[loss=0.2153, simple_loss=0.4414, pruned_loss=0.1134, ctc_loss=0.1644, over 14500.00 frames. ], tot_loss[loss=0.2153, simple_loss=0.4414, pruned_loss=0.1134, ctc_loss=0.1644, over 14500.00 frames. ], batch size: 37, lr: 3.00e-03, 
2023-01-06 02:56:16,581 INFO [train.py:862] Epoch 25, batch 500, loss[loss=0.1575, simple_loss=0.3781, pruned_loss=0.07365, ctc_loss=0.1124, over 14662.00 frames. ], tot_loss[loss=0.1794, simple_loss=0.432, pruned_loss=0.08644, ctc_loss=0.1266, over 2652907.35 frames. ], batch size: 35, lr: 3.00e-03, 
2023-01-06 02:57:22,846 INFO [train.py:862] Epoch 25, batch 1000, loss[loss=0.1905, simple_loss=0.4567, pruned_loss=0.09444, ctc_loss=0.1338, over 14845.00 frames. ], tot_loss[loss=0.1775, simple_loss=0.43, pruned_loss=0.08485, ctc_loss=0.1251, over 2851675.97 frames. ], batch size: 42, lr: 3.00e-03, 
2023-01-06 02:58:30,370 INFO [train.py:862] Epoch 25, batch 1500, loss[loss=0.2153, simple_loss=0.4587, pruned_loss=0.1121, ctc_loss=0.1613, over 14800.00 frames. ], tot_loss[loss=0.1773, simple_loss=0.4306, pruned_loss=0.08451, ctc_loss=0.1248, over 2863988.51 frames. ], batch size: 43, lr: 3.00e-03, 
2023-01-06 02:59:36,927 INFO [train.py:862] Epoch 25, batch 2000, loss[loss=0.1325, simple_loss=0.3759, pruned_loss=0.05773, ctc_loss=0.08394, over 14420.00 frames. ], tot_loss[loss=0.1765, simple_loss=0.4299, pruned_loss=0.0842, ctc_loss=0.1239, over 2868836.29 frames. ], batch size: 33, lr: 3.00e-03, 
2023-01-06 03:00:43,987 INFO [train.py:862] Epoch 25, batch 2500, loss[loss=0.1791, simple_loss=0.4468, pruned_loss=0.08418, ctc_loss=0.124, over 14695.00 frames. ], tot_loss[loss=0.1766, simple_loss=0.4316, pruned_loss=0.08424, ctc_loss=0.1238, over 2873830.72 frames. ], batch size: 44, lr: 3.00e-03, 
2023-01-06 03:01:26,320 INFO [checkpoint.py:75] Saving checkpoint to tiny_transducer_ctc/exp_2m_phone_halfdelay/checkpoint-440000.pt
2023-01-06 03:01:51,071 INFO [train.py:862] Epoch 25, batch 3000, loss[loss=0.186, simple_loss=0.4322, pruned_loss=0.0893, ctc_loss=0.1348, over 14697.00 frames. ], tot_loss[loss=0.1759, simple_loss=0.4298, pruned_loss=0.08369, ctc_loss=0.1233, over 2871166.24 frames. ], batch size: 37, lr: 3.00e-03, 
2023-01-06 03:02:58,201 INFO [train.py:862] Epoch 25, batch 3500, loss[loss=0.1468, simple_loss=0.3911, pruned_loss=0.06183, ctc_loss=0.09936, over 14777.00 frames. ], tot_loss[loss=0.1773, simple_loss=0.4299, pruned_loss=0.08475, ctc_loss=0.1248, over 2870182.28 frames. ], batch size: 36, lr: 3.00e-03, 
2023-01-06 03:04:05,035 INFO [train.py:862] Epoch 25, batch 4000, loss[loss=0.1733, simple_loss=0.4383, pruned_loss=0.08057, ctc_loss=0.1192, over 14836.00 frames. ], tot_loss[loss=0.177, simple_loss=0.4293, pruned_loss=0.08454, ctc_loss=0.1246, over 2870473.10 frames. ], batch size: 42, lr: 3.00e-03, 
2023-01-06 03:05:12,159 INFO [train.py:862] Epoch 25, batch 4500, loss[loss=0.1438, simple_loss=0.3873, pruned_loss=0.06733, ctc_loss=0.0936, over 14666.00 frames. ], tot_loss[loss=0.1759, simple_loss=0.4284, pruned_loss=0.08387, ctc_loss=0.1235, over 2865333.33 frames. ], batch size: 35, lr: 3.00e-03, 
2023-01-06 03:06:19,244 INFO [train.py:862] Epoch 25, batch 5000, loss[loss=0.1917, simple_loss=0.436, pruned_loss=0.09231, ctc_loss=0.1409, over 10139.00 frames. ], tot_loss[loss=0.1774, simple_loss=0.4307, pruned_loss=0.08502, ctc_loss=0.1248, over 2869287.40 frames. ], batch size: 104, lr: 3.00e-03, 
2023-01-06 03:07:27,177 INFO [train.py:862] Epoch 25, batch 5500, loss[loss=0.204, simple_loss=0.4596, pruned_loss=0.1026, ctc_loss=0.149, over 14581.00 frames. ], tot_loss[loss=0.1791, simple_loss=0.4319, pruned_loss=0.08591, ctc_loss=0.1265, over 2863107.37 frames. ], batch size: 43, lr: 3.00e-03, 
2023-01-06 03:08:33,610 INFO [train.py:862] Epoch 25, batch 6000, loss[loss=0.2038, simple_loss=0.4439, pruned_loss=0.1026, ctc_loss=0.152, over 14781.00 frames. ], tot_loss[loss=0.1784, simple_loss=0.431, pruned_loss=0.08561, ctc_loss=0.1258, over 2868125.23 frames. ], batch size: 43, lr: 3.00e-03, 
2023-01-06 03:09:41,354 INFO [train.py:862] Epoch 25, batch 6500, loss[loss=0.2456, simple_loss=0.4883, pruned_loss=0.1303, ctc_loss=0.1905, over 13082.00 frames. ], tot_loss[loss=0.1771, simple_loss=0.4297, pruned_loss=0.08465, ctc_loss=0.1246, over 2868440.39 frames. ], batch size: 76, lr: 3.00e-03, 
2023-01-06 03:10:49,631 INFO [train.py:862] Epoch 25, batch 7000, loss[loss=0.1508, simple_loss=0.3885, pruned_loss=0.06931, ctc_loss=0.1025, over 14781.00 frames. ], tot_loss[loss=0.177, simple_loss=0.4299, pruned_loss=0.08452, ctc_loss=0.1245, over 2854591.98 frames. ], batch size: 36, lr: 3.00e-03, 
2023-01-06 03:11:56,878 INFO [train.py:862] Epoch 25, batch 7500, loss[loss=0.1743, simple_loss=0.4084, pruned_loss=0.08299, ctc_loss=0.1259, over 14527.00 frames. ], tot_loss[loss=0.177, simple_loss=0.4297, pruned_loss=0.0847, ctc_loss=0.1245, over 2870509.17 frames. ], batch size: 36, lr: 3.00e-03, 
2023-01-06 03:13:04,573 INFO [train.py:862] Epoch 25, batch 8000, loss[loss=0.1506, simple_loss=0.4393, pruned_loss=0.0641, ctc_loss=0.09352, over 14585.00 frames. ], tot_loss[loss=0.176, simple_loss=0.4298, pruned_loss=0.08375, ctc_loss=0.1234, over 2846962.14 frames. ], batch size: 43, lr: 3.00e-03, 
2023-01-06 03:14:12,879 INFO [train.py:862] Epoch 25, batch 8500, loss[loss=0.2265, simple_loss=0.4833, pruned_loss=0.1092, ctc_loss=0.1732, over 14722.00 frames. ], tot_loss[loss=0.1789, simple_loss=0.4319, pruned_loss=0.08562, ctc_loss=0.1264, over 2870919.59 frames. ], batch size: 46, lr: 3.00e-03, 
2023-01-06 03:15:20,780 INFO [train.py:862] Epoch 25, batch 9000, loss[loss=0.1849, simple_loss=0.4575, pruned_loss=0.08839, ctc_loss=0.1282, over 14574.00 frames. ], tot_loss[loss=0.1771, simple_loss=0.4301, pruned_loss=0.08472, ctc_loss=0.1246, over 2858421.03 frames. ], batch size: 43, lr: 3.00e-03, 
2023-01-06 03:15:20,782 INFO [train.py:887] Computing validation loss
2023-01-06 03:15:47,644 INFO [train.py:897] Epoch 25, validation: loss=0.1931, simple_loss=0.4458, pruned_loss=0.09542, ctc_loss=0.1394, over 944034.00 frames. 
2023-01-06 03:15:47,645 INFO [train.py:898] Maximum memory allocated so far is 7351MB
2023-01-06 03:16:55,231 INFO [train.py:862] Epoch 25, batch 9500, loss[loss=0.1568, simple_loss=0.4285, pruned_loss=0.07199, ctc_loss=0.1013, over 14725.00 frames. ], tot_loss[loss=0.179, simple_loss=0.4321, pruned_loss=0.08595, ctc_loss=0.1263, over 2859612.79 frames. ], batch size: 45, lr: 3.00e-03, 
2023-01-06 03:18:02,794 INFO [train.py:862] Epoch 25, batch 10000, loss[loss=0.1401, simple_loss=0.4217, pruned_loss=0.05572, ctc_loss=0.08586, over 14741.00 frames. ], tot_loss[loss=0.1792, simple_loss=0.4326, pruned_loss=0.08594, ctc_loss=0.1265, over 2864132.25 frames. ], batch size: 45, lr: 3.00e-03, 
2023-01-06 03:19:10,648 INFO [train.py:862] Epoch 25, batch 10500, loss[loss=0.1895, simple_loss=0.4496, pruned_loss=0.09437, ctc_loss=0.1339, over 14510.00 frames. ], tot_loss[loss=0.1787, simple_loss=0.4315, pruned_loss=0.08559, ctc_loss=0.1261, over 2867473.71 frames. ], batch size: 50, lr: 3.00e-03, 
2023-01-06 03:20:18,808 INFO [train.py:862] Epoch 25, batch 11000, loss[loss=0.1629, simple_loss=0.3925, pruned_loss=0.07857, ctc_loss=0.1149, over 14529.00 frames. ], tot_loss[loss=0.1768, simple_loss=0.4296, pruned_loss=0.0846, ctc_loss=0.1243, over 2863983.45 frames. ], batch size: 34, lr: 3.00e-03, 
2023-01-06 03:21:26,595 INFO [train.py:862] Epoch 25, batch 11500, loss[loss=0.1593, simple_loss=0.4073, pruned_loss=0.07408, ctc_loss=0.1085, over 14535.00 frames. ], tot_loss[loss=0.1763, simple_loss=0.4294, pruned_loss=0.08421, ctc_loss=0.1238, over 2862112.01 frames. ], batch size: 36, lr: 3.00e-03, 
2023-01-06 03:22:35,102 INFO [train.py:862] Epoch 25, batch 12000, loss[loss=0.1683, simple_loss=0.4414, pruned_loss=0.07674, ctc_loss=0.1129, over 14105.00 frames. ], tot_loss[loss=0.1765, simple_loss=0.4306, pruned_loss=0.08413, ctc_loss=0.1238, over 2861651.18 frames. ], batch size: 56, lr: 3.00e-03, 
2023-01-06 03:23:43,226 INFO [train.py:862] Epoch 25, batch 12500, loss[loss=0.1572, simple_loss=0.4351, pruned_loss=0.06565, ctc_loss=0.1032, over 14790.00 frames. ], tot_loss[loss=0.1788, simple_loss=0.4299, pruned_loss=0.08582, ctc_loss=0.1265, over 2847552.14 frames. ], batch size: 43, lr: 3.00e-03, 
2023-01-06 03:24:25,745 INFO [checkpoint.py:75] Saving checkpoint to tiny_transducer_ctc/exp_2m_phone_halfdelay/checkpoint-450000.pt
2023-01-06 03:24:50,822 INFO [train.py:862] Epoch 25, batch 13000, loss[loss=0.2167, simple_loss=0.452, pruned_loss=0.1062, ctc_loss=0.1671, over 9978.00 frames. ], tot_loss[loss=0.1777, simple_loss=0.4321, pruned_loss=0.08458, ctc_loss=0.125, over 2847982.44 frames. ], batch size: 105, lr: 3.00e-03, 
2023-01-06 03:25:58,788 INFO [train.py:862] Epoch 25, batch 13500, loss[loss=0.1756, simple_loss=0.427, pruned_loss=0.07841, ctc_loss=0.1258, over 13014.00 frames. ], tot_loss[loss=0.1773, simple_loss=0.4313, pruned_loss=0.08446, ctc_loss=0.1247, over 2856284.60 frames. ], batch size: 76, lr: 3.00e-03, 
2023-01-06 03:27:05,930 INFO [train.py:862] Epoch 25, batch 14000, loss[loss=0.1824, simple_loss=0.4559, pruned_loss=0.08557, ctc_loss=0.1261, over 14802.00 frames. ], tot_loss[loss=0.1764, simple_loss=0.4296, pruned_loss=0.08401, ctc_loss=0.1239, over 2855141.24 frames. ], batch size: 43, lr: 3.00e-03, 
2023-01-06 03:28:13,238 INFO [train.py:862] Epoch 25, batch 14500, loss[loss=0.1868, simple_loss=0.4567, pruned_loss=0.08728, ctc_loss=0.1317, over 14645.00 frames. ], tot_loss[loss=0.1759, simple_loss=0.4301, pruned_loss=0.08372, ctc_loss=0.1232, over 2874352.52 frames. ], batch size: 41, lr: 3.00e-03, 
2023-01-06 03:29:20,797 INFO [train.py:862] Epoch 25, batch 15000, loss[loss=0.152, simple_loss=0.3945, pruned_loss=0.06995, ctc_loss=0.1026, over 14528.00 frames. ], tot_loss[loss=0.1767, simple_loss=0.4292, pruned_loss=0.08434, ctc_loss=0.1244, over 2851293.74 frames. ], batch size: 34, lr: 3.00e-03, 
2023-01-06 03:30:28,117 INFO [train.py:862] Epoch 25, batch 15500, loss[loss=0.1787, simple_loss=0.4105, pruned_loss=0.08809, ctc_loss=0.1295, over 14530.00 frames. ], tot_loss[loss=0.1813, simple_loss=0.4339, pruned_loss=0.08762, ctc_loss=0.1284, over 2865198.07 frames. ], batch size: 36, lr: 3.00e-03, 
2023-01-06 03:31:36,224 INFO [train.py:862] Epoch 25, batch 16000, loss[loss=0.1733, simple_loss=0.4311, pruned_loss=0.08381, ctc_loss=0.1193, over 14583.00 frames. ], tot_loss[loss=0.1759, simple_loss=0.4289, pruned_loss=0.08399, ctc_loss=0.1234, over 2854965.28 frames. ], batch size: 43, lr: 3.00e-03, 
2023-01-06 03:32:43,212 INFO [train.py:862] Epoch 25, batch 16500, loss[loss=0.1923, simple_loss=0.4504, pruned_loss=0.09414, ctc_loss=0.1378, over 14143.00 frames. ], tot_loss[loss=0.1786, simple_loss=0.4302, pruned_loss=0.08577, ctc_loss=0.1262, over 2854007.55 frames. ], batch size: 56, lr: 3.00e-03, 
2023-01-06 03:33:50,718 INFO [train.py:862] Epoch 25, batch 17000, loss[loss=0.1667, simple_loss=0.4329, pruned_loss=0.0752, ctc_loss=0.1132, over 14806.00 frames. ], tot_loss[loss=0.176, simple_loss=0.4304, pruned_loss=0.0839, ctc_loss=0.1233, over 2860617.47 frames. ], batch size: 43, lr: 3.00e-03, 
2023-01-06 03:34:58,444 INFO [train.py:862] Epoch 25, batch 17500, loss[loss=0.1811, simple_loss=0.455, pruned_loss=0.08475, ctc_loss=0.1249, over 14798.00 frames. ], tot_loss[loss=0.1754, simple_loss=0.4281, pruned_loss=0.08383, ctc_loss=0.1229, over 2851870.96 frames. ], batch size: 43, lr: 3.00e-03, 
2023-01-06 03:36:04,316 INFO [train.py:862] Epoch 25, batch 18000, loss[loss=0.179, simple_loss=0.4463, pruned_loss=0.0791, ctc_loss=0.1261, over 14577.00 frames. ], tot_loss[loss=0.1779, simple_loss=0.4312, pruned_loss=0.08531, ctc_loss=0.1252, over 2856029.29 frames. ], batch size: 43, lr: 3.00e-03, 
2023-01-06 03:36:04,316 INFO [train.py:887] Computing validation loss
2023-01-06 03:36:31,174 INFO [train.py:897] Epoch 25, validation: loss=0.1947, simple_loss=0.447, pruned_loss=0.09632, ctc_loss=0.1411, over 944034.00 frames. 
2023-01-06 03:36:31,175 INFO [train.py:898] Maximum memory allocated so far is 7351MB
2023-01-06 03:36:58,887 INFO [checkpoint.py:75] Saving checkpoint to tiny_transducer_ctc/exp_2m_phone_halfdelay/epoch-25.pt
2023-01-06 03:37:02,051 INFO [train.py:862] Epoch 26, batch 0, loss[loss=0.2048, simple_loss=0.4474, pruned_loss=0.1012, ctc_loss=0.1533, over 14799.00 frames. ], tot_loss[loss=0.2048, simple_loss=0.4474, pruned_loss=0.1012, ctc_loss=0.1533, over 14799.00 frames. ], batch size: 43, lr: 3.00e-03, 
2023-01-06 03:38:09,485 INFO [train.py:862] Epoch 26, batch 500, loss[loss=0.1888, simple_loss=0.462, pruned_loss=0.08697, ctc_loss=0.1334, over 14536.00 frames. ], tot_loss[loss=0.1768, simple_loss=0.4303, pruned_loss=0.08418, ctc_loss=0.1243, over 2630923.93 frames. ], batch size: 50, lr: 3.00e-03, 
2023-01-06 03:39:17,270 INFO [train.py:862] Epoch 26, batch 1000, loss[loss=0.1914, simple_loss=0.4517, pruned_loss=0.09491, ctc_loss=0.136, over 14647.00 frames. ], tot_loss[loss=0.1782, simple_loss=0.4315, pruned_loss=0.08523, ctc_loss=0.1256, over 2851855.11 frames. ], batch size: 40, lr: 3.00e-03, 
2023-01-06 03:40:24,370 INFO [train.py:862] Epoch 26, batch 1500, loss[loss=0.1701, simple_loss=0.4423, pruned_loss=0.07769, ctc_loss=0.1149, over 14177.00 frames. ], tot_loss[loss=0.176, simple_loss=0.4294, pruned_loss=0.08383, ctc_loss=0.1235, over 2879362.09 frames. ], batch size: 56, lr: 3.00e-03, 
2023-01-06 03:41:31,157 INFO [train.py:862] Epoch 26, batch 2000, loss[loss=0.1716, simple_loss=0.4397, pruned_loss=0.08163, ctc_loss=0.1159, over 13692.00 frames. ], tot_loss[loss=0.1749, simple_loss=0.4295, pruned_loss=0.08307, ctc_loss=0.1222, over 2861277.25 frames. ], batch size: 63, lr: 3.00e-03, 
2023-01-06 03:42:38,444 INFO [train.py:862] Epoch 26, batch 2500, loss[loss=0.1633, simple_loss=0.4258, pruned_loss=0.07225, ctc_loss=0.1111, over 14845.00 frames. ], tot_loss[loss=0.1766, simple_loss=0.4319, pruned_loss=0.08377, ctc_loss=0.1238, over 2865617.74 frames. ], batch size: 41, lr: 3.00e-03, 
2023-01-06 03:43:45,251 INFO [train.py:862] Epoch 26, batch 3000, loss[loss=0.2026, simple_loss=0.4318, pruned_loss=0.1067, ctc_loss=0.1512, over 14703.00 frames. ], tot_loss[loss=0.175, simple_loss=0.4277, pruned_loss=0.08328, ctc_loss=0.1226, over 2867580.16 frames. ], batch size: 39, lr: 3.00e-03, 
2023-01-06 03:44:52,301 INFO [train.py:862] Epoch 26, batch 3500, loss[loss=0.2196, simple_loss=0.4659, pruned_loss=0.1111, ctc_loss=0.1662, over 14651.00 frames. ], tot_loss[loss=0.1795, simple_loss=0.4326, pruned_loss=0.08609, ctc_loss=0.1268, over 2866085.38 frames. ], batch size: 41, lr: 3.00e-03, 
2023-01-06 03:45:59,698 INFO [train.py:862] Epoch 26, batch 4000, loss[loss=0.161, simple_loss=0.4204, pruned_loss=0.07499, ctc_loss=0.1078, over 14651.00 frames. ], tot_loss[loss=0.1784, simple_loss=0.4301, pruned_loss=0.08557, ctc_loss=0.126, over 2872271.51 frames. ], batch size: 41, lr: 3.00e-03, 
2023-01-06 03:47:06,483 INFO [train.py:862] Epoch 26, batch 4500, loss[loss=0.1638, simple_loss=0.4217, pruned_loss=0.07924, ctc_loss=0.1097, over 14583.00 frames. ], tot_loss[loss=0.1787, simple_loss=0.4317, pruned_loss=0.08566, ctc_loss=0.1261, over 2863258.43 frames. ], batch size: 43, lr: 3.00e-03, 
2023-01-06 03:47:20,479 INFO [checkpoint.py:75] Saving checkpoint to tiny_transducer_ctc/exp_2m_phone_halfdelay/checkpoint-460000.pt
2023-01-06 03:48:13,992 INFO [train.py:862] Epoch 26, batch 5000, loss[loss=0.1673, simple_loss=0.4269, pruned_loss=0.08099, ctc_loss=0.1127, over 14824.00 frames. ], tot_loss[loss=0.1751, simple_loss=0.4285, pruned_loss=0.0831, ctc_loss=0.1227, over 2858048.41 frames. ], batch size: 41, lr: 3.00e-03, 
2023-01-06 03:49:20,713 INFO [train.py:862] Epoch 26, batch 5500, loss[loss=0.1732, simple_loss=0.4199, pruned_loss=0.08575, ctc_loss=0.1207, over 14755.00 frames. ], tot_loss[loss=0.177, simple_loss=0.4307, pruned_loss=0.08444, ctc_loss=0.1243, over 2870416.79 frames. ], batch size: 36, lr: 3.00e-03, 
2023-01-06 03:50:27,678 INFO [train.py:862] Epoch 26, batch 6000, loss[loss=0.1748, simple_loss=0.4544, pruned_loss=0.07825, ctc_loss=0.1188, over 14717.00 frames. ], tot_loss[loss=0.1775, simple_loss=0.4303, pruned_loss=0.08498, ctc_loss=0.1249, over 2882436.35 frames. ], batch size: 45, lr: 3.00e-03, 
2023-01-06 03:51:34,634 INFO [train.py:862] Epoch 26, batch 6500, loss[loss=0.154, simple_loss=0.3872, pruned_loss=0.0706, ctc_loss=0.1068, over 14792.00 frames. ], tot_loss[loss=0.1769, simple_loss=0.4305, pruned_loss=0.08457, ctc_loss=0.1243, over 2871931.53 frames. ], batch size: 36, lr: 3.00e-03, 
2023-01-06 03:52:41,710 INFO [train.py:862] Epoch 26, batch 7000, loss[loss=0.1739, simple_loss=0.4294, pruned_loss=0.08181, ctc_loss=0.1213, over 14871.00 frames. ], tot_loss[loss=0.1749, simple_loss=0.4281, pruned_loss=0.08319, ctc_loss=0.1224, over 2876067.04 frames. ], batch size: 40, lr: 3.00e-03, 
2023-01-06 03:53:49,260 INFO [train.py:862] Epoch 26, batch 7500, loss[loss=0.1519, simple_loss=0.4223, pruned_loss=0.06442, ctc_loss=0.09882, over 14759.00 frames. ], tot_loss[loss=0.1749, simple_loss=0.4281, pruned_loss=0.08344, ctc_loss=0.1224, over 2865375.27 frames. ], batch size: 46, lr: 3.00e-03, 
2023-01-06 03:54:56,385 INFO [train.py:862] Epoch 26, batch 8000, loss[loss=0.2001, simple_loss=0.4518, pruned_loss=0.09576, ctc_loss=0.1479, over 10355.00 frames. ], tot_loss[loss=0.1771, simple_loss=0.4298, pruned_loss=0.08455, ctc_loss=0.1246, over 2853885.12 frames. ], batch size: 103, lr: 3.00e-03, 
2023-01-06 03:56:02,949 INFO [train.py:862] Epoch 26, batch 8500, loss[loss=0.1904, simple_loss=0.4562, pruned_loss=0.0945, ctc_loss=0.1338, over 14750.00 frames. ], tot_loss[loss=0.1767, simple_loss=0.4301, pruned_loss=0.08426, ctc_loss=0.1241, over 2857552.06 frames. ], batch size: 45, lr: 3.00e-03, 
2023-01-06 03:57:10,200 INFO [train.py:862] Epoch 26, batch 9000, loss[loss=0.2067, simple_loss=0.4671, pruned_loss=0.1053, ctc_loss=0.1501, over 14638.00 frames. ], tot_loss[loss=0.1751, simple_loss=0.4284, pruned_loss=0.0835, ctc_loss=0.1225, over 2855309.64 frames. ], batch size: 42, lr: 3.00e-03, 
2023-01-06 03:57:10,201 INFO [train.py:887] Computing validation loss
2023-01-06 03:57:36,243 INFO [train.py:897] Epoch 26, validation: loss=0.1922, simple_loss=0.4448, pruned_loss=0.09392, ctc_loss=0.139, over 944034.00 frames. 
2023-01-06 03:57:36,244 INFO [train.py:898] Maximum memory allocated so far is 7351MB
2023-01-06 03:58:44,190 INFO [train.py:862] Epoch 26, batch 9500, loss[loss=0.1702, simple_loss=0.447, pruned_loss=0.0738, ctc_loss=0.1158, over 14703.00 frames. ], tot_loss[loss=0.1756, simple_loss=0.4284, pruned_loss=0.08373, ctc_loss=0.1232, over 2857584.80 frames. ], batch size: 47, lr: 3.00e-03, 
2023-01-06 03:59:51,861 INFO [train.py:862] Epoch 26, batch 10000, loss[loss=0.1682, simple_loss=0.4365, pruned_loss=0.07738, ctc_loss=0.1136, over 14112.00 frames. ], tot_loss[loss=0.1772, simple_loss=0.4296, pruned_loss=0.08455, ctc_loss=0.1248, over 2866234.90 frames. ], batch size: 56, lr: 3.00e-03, 
2023-01-06 04:00:59,688 INFO [train.py:862] Epoch 26, batch 10500, loss[loss=0.1599, simple_loss=0.3978, pruned_loss=0.07582, ctc_loss=0.1107, over 13973.00 frames. ], tot_loss[loss=0.1769, simple_loss=0.4297, pruned_loss=0.0843, ctc_loss=0.1245, over 2864672.94 frames. ], batch size: 31, lr: 3.00e-03, 
2023-01-06 04:02:07,485 INFO [train.py:862] Epoch 26, batch 11000, loss[loss=0.1839, simple_loss=0.4474, pruned_loss=0.08854, ctc_loss=0.1289, over 13096.00 frames. ], tot_loss[loss=0.1769, simple_loss=0.4314, pruned_loss=0.08414, ctc_loss=0.1242, over 2854594.44 frames. ], batch size: 76, lr: 3.00e-03, 
2023-01-06 04:03:14,291 INFO [train.py:862] Epoch 26, batch 11500, loss[loss=0.1827, simple_loss=0.4465, pruned_loss=0.08453, ctc_loss=0.1291, over 14548.00 frames. ], tot_loss[loss=0.1775, simple_loss=0.4302, pruned_loss=0.08482, ctc_loss=0.1251, over 2861961.07 frames. ], batch size: 53, lr: 3.00e-03, 
2023-01-06 04:04:21,741 INFO [train.py:862] Epoch 26, batch 12000, loss[loss=0.1713, simple_loss=0.4377, pruned_loss=0.07841, ctc_loss=0.1174, over 14804.00 frames. ], tot_loss[loss=0.1752, simple_loss=0.4283, pruned_loss=0.08336, ctc_loss=0.1228, over 2866090.87 frames. ], batch size: 43, lr: 3.00e-03, 
2023-01-06 04:05:28,900 INFO [train.py:862] Epoch 26, batch 12500, loss[loss=0.1614, simple_loss=0.4076, pruned_loss=0.07326, ctc_loss=0.1119, over 14728.00 frames. ], tot_loss[loss=0.1793, simple_loss=0.433, pruned_loss=0.08588, ctc_loss=0.1266, over 2870231.24 frames. ], batch size: 37, lr: 3.00e-03, 
2023-01-06 04:06:36,073 INFO [train.py:862] Epoch 26, batch 13000, loss[loss=0.1813, simple_loss=0.453, pruned_loss=0.08178, ctc_loss=0.1268, over 14734.00 frames. ], tot_loss[loss=0.1771, simple_loss=0.4306, pruned_loss=0.08461, ctc_loss=0.1245, over 2859122.96 frames. ], batch size: 46, lr: 3.00e-03, 
2023-01-06 04:07:43,668 INFO [train.py:862] Epoch 26, batch 13500, loss[loss=0.1778, simple_loss=0.4517, pruned_loss=0.07893, ctc_loss=0.1233, over 14530.00 frames. ], tot_loss[loss=0.1767, simple_loss=0.4306, pruned_loss=0.08422, ctc_loss=0.1241, over 2845970.07 frames. ], batch size: 50, lr: 3.00e-03, 
2023-01-06 04:08:50,582 INFO [train.py:862] Epoch 26, batch 14000, loss[loss=0.1758, simple_loss=0.4484, pruned_loss=0.07929, ctc_loss=0.1211, over 14765.00 frames. ], tot_loss[loss=0.1761, simple_loss=0.4295, pruned_loss=0.08398, ctc_loss=0.1235, over 2876126.00 frames. ], batch size: 46, lr: 3.00e-03, 
2023-01-06 04:09:58,353 INFO [train.py:862] Epoch 26, batch 14500, loss[loss=0.1686, simple_loss=0.4202, pruned_loss=0.08067, ctc_loss=0.1163, over 14705.00 frames. ], tot_loss[loss=0.1769, simple_loss=0.4288, pruned_loss=0.08448, ctc_loss=0.1246, over 2858496.83 frames. ], batch size: 39, lr: 3.00e-03, 
2023-01-06 04:10:11,588 INFO [checkpoint.py:75] Saving checkpoint to tiny_transducer_ctc/exp_2m_phone_halfdelay/checkpoint-470000.pt
2023-01-06 04:11:05,069 INFO [train.py:862] Epoch 26, batch 15000, loss[loss=0.1934, simple_loss=0.4597, pruned_loss=0.0911, ctc_loss=0.1388, over 14547.00 frames. ], tot_loss[loss=0.1748, simple_loss=0.4269, pruned_loss=0.08317, ctc_loss=0.1225, over 2858766.07 frames. ], batch size: 48, lr: 3.00e-03, 
2023-01-06 04:12:12,436 INFO [train.py:862] Epoch 26, batch 15500, loss[loss=0.2204, simple_loss=0.4342, pruned_loss=0.115, ctc_loss=0.1725, over 14711.00 frames. ], tot_loss[loss=0.176, simple_loss=0.4293, pruned_loss=0.08382, ctc_loss=0.1235, over 2880471.70 frames. ], batch size: 38, lr: 3.00e-03, 
2023-01-06 04:13:19,453 INFO [train.py:862] Epoch 26, batch 16000, loss[loss=0.1569, simple_loss=0.3987, pruned_loss=0.07404, ctc_loss=0.107, over 14531.00 frames. ], tot_loss[loss=0.1747, simple_loss=0.4294, pruned_loss=0.08313, ctc_loss=0.122, over 2864756.47 frames. ], batch size: 36, lr: 3.00e-03, 
2023-01-06 04:14:26,090 INFO [train.py:862] Epoch 26, batch 16500, loss[loss=0.2005, simple_loss=0.4576, pruned_loss=0.1027, ctc_loss=0.1444, over 12790.00 frames. ], tot_loss[loss=0.1741, simple_loss=0.4277, pruned_loss=0.08288, ctc_loss=0.1215, over 2854929.93 frames. ], batch size: 76, lr: 3.00e-03, 
2023-01-06 04:15:33,767 INFO [train.py:862] Epoch 26, batch 17000, loss[loss=0.1995, simple_loss=0.4686, pruned_loss=0.09826, ctc_loss=0.1425, over 14756.00 frames. ], tot_loss[loss=0.176, simple_loss=0.4294, pruned_loss=0.08413, ctc_loss=0.1234, over 2864720.61 frames. ], batch size: 45, lr: 3.00e-03, 
2023-01-06 04:16:40,339 INFO [train.py:862] Epoch 26, batch 17500, loss[loss=0.1604, simple_loss=0.435, pruned_loss=0.07199, ctc_loss=0.1051, over 14827.00 frames. ], tot_loss[loss=0.1747, simple_loss=0.4292, pruned_loss=0.08292, ctc_loss=0.1221, over 2869227.17 frames. ], batch size: 41, lr: 3.00e-03, 
2023-01-06 04:17:46,012 INFO [train.py:862] Epoch 26, batch 18000, loss[loss=0.2232, simple_loss=0.4832, pruned_loss=0.1128, ctc_loss=0.1669, over 14527.00 frames. ], tot_loss[loss=0.1772, simple_loss=0.4304, pruned_loss=0.08469, ctc_loss=0.1246, over 2854076.52 frames. ], batch size: 50, lr: 3.00e-03, 
2023-01-06 04:17:46,013 INFO [train.py:887] Computing validation loss
2023-01-06 04:18:11,887 INFO [train.py:897] Epoch 26, validation: loss=0.1944, simple_loss=0.4466, pruned_loss=0.09631, ctc_loss=0.1408, over 944034.00 frames. 
2023-01-06 04:18:11,887 INFO [train.py:898] Maximum memory allocated so far is 7351MB
2023-01-06 04:18:39,427 INFO [checkpoint.py:75] Saving checkpoint to tiny_transducer_ctc/exp_2m_phone_halfdelay/epoch-26.pt
2023-01-06 04:18:42,248 INFO [train.py:862] Epoch 27, batch 0, loss[loss=0.2088, simple_loss=0.4299, pruned_loss=0.1073, ctc_loss=0.1602, over 14710.00 frames. ], tot_loss[loss=0.2088, simple_loss=0.4299, pruned_loss=0.1073, ctc_loss=0.1602, over 14710.00 frames. ], batch size: 37, lr: 3.00e-03, 
2023-01-06 04:19:50,224 INFO [train.py:862] Epoch 27, batch 500, loss[loss=0.2479, simple_loss=0.4982, pruned_loss=0.1345, ctc_loss=0.1898, over 14667.00 frames. ], tot_loss[loss=0.1763, simple_loss=0.4291, pruned_loss=0.08422, ctc_loss=0.1238, over 2641361.41 frames. ], batch size: 47, lr: 3.00e-03, 
2023-01-06 04:20:57,052 INFO [train.py:862] Epoch 27, batch 1000, loss[loss=0.2015, simple_loss=0.4663, pruned_loss=0.09492, ctc_loss=0.1472, over 14753.00 frames. ], tot_loss[loss=0.1757, simple_loss=0.4309, pruned_loss=0.08358, ctc_loss=0.1228, over 2857782.25 frames. ], batch size: 46, lr: 3.00e-03, 
2023-01-06 04:22:03,995 INFO [train.py:862] Epoch 27, batch 1500, loss[loss=0.1862, simple_loss=0.3937, pruned_loss=0.09332, ctc_loss=0.1417, over 14009.00 frames. ], tot_loss[loss=0.1758, simple_loss=0.4284, pruned_loss=0.08359, ctc_loss=0.1236, over 2876634.73 frames. ], batch size: 31, lr: 3.00e-03, 
2023-01-06 04:23:11,812 INFO [train.py:862] Epoch 27, batch 2000, loss[loss=0.1823, simple_loss=0.4133, pruned_loss=0.09281, ctc_loss=0.1322, over 14668.00 frames. ], tot_loss[loss=0.1758, simple_loss=0.43, pruned_loss=0.0837, ctc_loss=0.1232, over 2876667.08 frames. ], batch size: 35, lr: 3.00e-03, 
2023-01-06 04:24:19,219 INFO [train.py:862] Epoch 27, batch 2500, loss[loss=0.1867, simple_loss=0.413, pruned_loss=0.08969, ctc_loss=0.1398, over 14712.00 frames. ], tot_loss[loss=0.176, simple_loss=0.4291, pruned_loss=0.0838, ctc_loss=0.1235, over 2871081.29 frames. ], batch size: 37, lr: 3.00e-03, 
2023-01-06 04:25:26,550 INFO [train.py:862] Epoch 27, batch 3000, loss[loss=0.1579, simple_loss=0.4196, pruned_loss=0.07321, ctc_loss=0.1043, over 14715.00 frames. ], tot_loss[loss=0.1777, simple_loss=0.4312, pruned_loss=0.08489, ctc_loss=0.1251, over 2861322.19 frames. ], batch size: 38, lr: 3.00e-03, 
2023-01-06 04:26:33,887 INFO [train.py:862] Epoch 27, batch 3500, loss[loss=0.1975, simple_loss=0.4465, pruned_loss=0.09645, ctc_loss=0.1452, over 13759.00 frames. ], tot_loss[loss=0.1765, simple_loss=0.429, pruned_loss=0.08417, ctc_loss=0.1242, over 2851314.28 frames. ], batch size: 63, lr: 3.00e-03, 
2023-01-06 04:27:40,850 INFO [train.py:862] Epoch 27, batch 4000, loss[loss=0.2191, simple_loss=0.4881, pruned_loss=0.1122, ctc_loss=0.1604, over 14687.00 frames. ], tot_loss[loss=0.1753, simple_loss=0.4282, pruned_loss=0.08356, ctc_loss=0.1229, over 2864568.67 frames. ], batch size: 47, lr: 3.00e-03, 
2023-01-06 04:28:48,346 INFO [train.py:862] Epoch 27, batch 4500, loss[loss=0.1769, simple_loss=0.423, pruned_loss=0.08367, ctc_loss=0.1262, over 14654.00 frames. ], tot_loss[loss=0.1776, simple_loss=0.4307, pruned_loss=0.08489, ctc_loss=0.1251, over 2873389.88 frames. ], batch size: 40, lr: 3.00e-03, 
2023-01-06 04:29:55,295 INFO [train.py:862] Epoch 27, batch 5000, loss[loss=0.1569, simple_loss=0.4405, pruned_loss=0.06798, ctc_loss=0.1006, over 14697.00 frames. ], tot_loss[loss=0.1775, simple_loss=0.4296, pruned_loss=0.08511, ctc_loss=0.1251, over 2868231.50 frames. ], batch size: 47, lr: 3.00e-03, 
2023-01-06 04:31:02,161 INFO [train.py:862] Epoch 27, batch 5500, loss[loss=0.1367, simple_loss=0.3726, pruned_loss=0.06236, ctc_loss=0.08875, over 14426.00 frames. ], tot_loss[loss=0.1772, simple_loss=0.4313, pruned_loss=0.08447, ctc_loss=0.1246, over 2859493.37 frames. ], batch size: 33, lr: 3.00e-03, 
2023-01-06 04:32:09,505 INFO [train.py:862] Epoch 27, batch 6000, loss[loss=0.1771, simple_loss=0.4512, pruned_loss=0.08304, ctc_loss=0.1208, over 14739.00 frames. ], tot_loss[loss=0.1752, simple_loss=0.4284, pruned_loss=0.08352, ctc_loss=0.1227, over 2866519.25 frames. ], batch size: 46, lr: 3.00e-03, 
2023-01-06 04:33:00,893 INFO [checkpoint.py:75] Saving checkpoint to tiny_transducer_ctc/exp_2m_phone_halfdelay/checkpoint-480000.pt
2023-01-06 04:33:16,186 INFO [train.py:862] Epoch 27, batch 6500, loss[loss=0.1849, simple_loss=0.4514, pruned_loss=0.08724, ctc_loss=0.13, over 14842.00 frames. ], tot_loss[loss=0.1733, simple_loss=0.427, pruned_loss=0.08215, ctc_loss=0.1209, over 2870837.04 frames. ], batch size: 42, lr: 3.00e-03, 
2023-01-06 04:34:23,486 INFO [train.py:862] Epoch 27, batch 7000, loss[loss=0.181, simple_loss=0.4422, pruned_loss=0.08468, ctc_loss=0.1275, over 14040.00 frames. ], tot_loss[loss=0.1753, simple_loss=0.4287, pruned_loss=0.08354, ctc_loss=0.1227, over 2863209.59 frames. ], batch size: 56, lr: 3.00e-03, 
2023-01-06 04:35:30,154 INFO [train.py:862] Epoch 27, batch 7500, loss[loss=0.1337, simple_loss=0.3972, pruned_loss=0.05745, ctc_loss=0.08132, over 14667.00 frames. ], tot_loss[loss=0.1737, simple_loss=0.4278, pruned_loss=0.08249, ctc_loss=0.1212, over 2878047.59 frames. ], batch size: 40, lr: 3.00e-03, 
2023-01-06 04:36:36,834 INFO [train.py:862] Epoch 27, batch 8000, loss[loss=0.1941, simple_loss=0.4453, pruned_loss=0.1001, ctc_loss=0.139, over 10324.00 frames. ], tot_loss[loss=0.1754, simple_loss=0.4286, pruned_loss=0.08359, ctc_loss=0.1229, over 2862200.26 frames. ], batch size: 104, lr: 3.00e-03, 
2023-01-06 04:37:44,457 INFO [train.py:862] Epoch 27, batch 8500, loss[loss=0.1591, simple_loss=0.3899, pruned_loss=0.07593, ctc_loss=0.1112, over 14661.00 frames. ], tot_loss[loss=0.1785, simple_loss=0.4323, pruned_loss=0.08536, ctc_loss=0.1258, over 2866396.92 frames. ], batch size: 35, lr: 3.00e-03, 
2023-01-06 04:38:51,250 INFO [train.py:862] Epoch 27, batch 9000, loss[loss=0.1787, simple_loss=0.4572, pruned_loss=0.08455, ctc_loss=0.121, over 14223.00 frames. ], tot_loss[loss=0.1754, simple_loss=0.4293, pruned_loss=0.08339, ctc_loss=0.1229, over 2875005.63 frames. ], batch size: 52, lr: 3.00e-03, 
2023-01-06 04:38:51,251 INFO [train.py:887] Computing validation loss
2023-01-06 04:39:17,878 INFO [train.py:897] Epoch 27, validation: loss=0.1925, simple_loss=0.4454, pruned_loss=0.09443, ctc_loss=0.139, over 944034.00 frames. 
2023-01-06 04:39:17,879 INFO [train.py:898] Maximum memory allocated so far is 7351MB
2023-01-06 04:40:25,344 INFO [train.py:862] Epoch 27, batch 9500, loss[loss=0.197, simple_loss=0.4131, pruned_loss=0.1028, ctc_loss=0.1488, over 14429.00 frames. ], tot_loss[loss=0.1756, simple_loss=0.428, pruned_loss=0.08383, ctc_loss=0.1232, over 2852868.86 frames. ], batch size: 33, lr: 3.00e-03, 
2023-01-06 04:41:32,698 INFO [train.py:862] Epoch 27, batch 10000, loss[loss=0.1567, simple_loss=0.3985, pruned_loss=0.07249, ctc_loss=0.1074, over 14510.00 frames. ], tot_loss[loss=0.1755, simple_loss=0.4293, pruned_loss=0.08336, ctc_loss=0.123, over 2866635.63 frames. ], batch size: 34, lr: 3.00e-03, 
2023-01-06 04:42:39,917 INFO [train.py:862] Epoch 27, batch 10500, loss[loss=0.1857, simple_loss=0.4432, pruned_loss=0.0846, ctc_loss=0.134, over 14801.00 frames. ], tot_loss[loss=0.1773, simple_loss=0.4307, pruned_loss=0.08488, ctc_loss=0.1246, over 2857229.95 frames. ], batch size: 43, lr: 3.00e-03, 
2023-01-06 04:43:47,127 INFO [train.py:862] Epoch 27, batch 11000, loss[loss=0.2074, simple_loss=0.4435, pruned_loss=0.1023, ctc_loss=0.1574, over 14102.00 frames. ], tot_loss[loss=0.1766, simple_loss=0.4303, pruned_loss=0.0842, ctc_loss=0.1241, over 2872298.82 frames. ], batch size: 56, lr: 3.00e-03, 
2023-01-06 04:44:54,785 INFO [train.py:862] Epoch 27, batch 11500, loss[loss=0.165, simple_loss=0.424, pruned_loss=0.07232, ctc_loss=0.1138, over 14658.00 frames. ], tot_loss[loss=0.1772, simple_loss=0.4301, pruned_loss=0.08479, ctc_loss=0.1247, over 2859770.63 frames. ], batch size: 40, lr: 3.00e-03, 
2023-01-06 04:46:01,790 INFO [train.py:862] Epoch 27, batch 12000, loss[loss=0.2104, simple_loss=0.4629, pruned_loss=0.09912, ctc_loss=0.1589, over 14590.00 frames. ], tot_loss[loss=0.1758, simple_loss=0.4292, pruned_loss=0.08395, ctc_loss=0.1232, over 2862166.28 frames. ], batch size: 43, lr: 3.00e-03, 
2023-01-06 04:47:09,375 INFO [train.py:862] Epoch 27, batch 12500, loss[loss=0.1865, simple_loss=0.4453, pruned_loss=0.09158, ctc_loss=0.1317, over 14803.00 frames. ], tot_loss[loss=0.1777, simple_loss=0.4314, pruned_loss=0.08484, ctc_loss=0.125, over 2864034.36 frames. ], batch size: 43, lr: 3.00e-03, 
2023-01-06 04:48:16,682 INFO [train.py:862] Epoch 27, batch 13000, loss[loss=0.1458, simple_loss=0.3929, pruned_loss=0.06406, ctc_loss=0.09663, over 14685.00 frames. ], tot_loss[loss=0.1768, simple_loss=0.4297, pruned_loss=0.08452, ctc_loss=0.1243, over 2864789.22 frames. ], batch size: 39, lr: 3.00e-03, 
2023-01-06 04:49:23,583 INFO [train.py:862] Epoch 27, batch 13500, loss[loss=0.1735, simple_loss=0.4492, pruned_loss=0.07901, ctc_loss=0.1177, over 14820.00 frames. ], tot_loss[loss=0.1744, simple_loss=0.429, pruned_loss=0.08269, ctc_loss=0.1218, over 2867809.85 frames. ], batch size: 41, lr: 3.00e-03, 
2023-01-06 04:50:31,814 INFO [train.py:862] Epoch 27, batch 14000, loss[loss=0.1451, simple_loss=0.3907, pruned_loss=0.06828, ctc_loss=0.09433, over 14792.00 frames. ], tot_loss[loss=0.174, simple_loss=0.4283, pruned_loss=0.08246, ctc_loss=0.1214, over 2865382.30 frames. ], batch size: 36, lr: 3.00e-03, 
2023-01-06 04:51:39,089 INFO [train.py:862] Epoch 27, batch 14500, loss[loss=0.1615, simple_loss=0.4305, pruned_loss=0.07411, ctc_loss=0.1067, over 14682.00 frames. ], tot_loss[loss=0.1751, simple_loss=0.4288, pruned_loss=0.08333, ctc_loss=0.1226, over 2860827.33 frames. ], batch size: 44, lr: 3.00e-03, 
2023-01-06 04:52:47,128 INFO [train.py:862] Epoch 27, batch 15000, loss[loss=0.2064, simple_loss=0.4756, pruned_loss=0.09877, ctc_loss=0.1506, over 14529.00 frames. ], tot_loss[loss=0.175, simple_loss=0.4276, pruned_loss=0.08337, ctc_loss=0.1227, over 2866430.42 frames. ], batch size: 53, lr: 3.00e-03, 
2023-01-06 04:53:54,276 INFO [train.py:862] Epoch 27, batch 15500, loss[loss=0.1475, simple_loss=0.3867, pruned_loss=0.06773, ctc_loss=0.09882, over 14705.00 frames. ], tot_loss[loss=0.1765, simple_loss=0.4306, pruned_loss=0.08402, ctc_loss=0.1239, over 2847893.39 frames. ], batch size: 37, lr: 3.00e-03, 
2023-01-06 04:55:01,756 INFO [train.py:862] Epoch 27, batch 16000, loss[loss=0.1345, simple_loss=0.3827, pruned_loss=0.06044, ctc_loss=0.08426, over 14680.00 frames. ], tot_loss[loss=0.1743, simple_loss=0.4277, pruned_loss=0.08277, ctc_loss=0.1219, over 2841128.45 frames. ], batch size: 35, lr: 3.00e-03, 
2023-01-06 04:55:53,034 INFO [checkpoint.py:75] Saving checkpoint to tiny_transducer_ctc/exp_2m_phone_halfdelay/checkpoint-490000.pt
2023-01-06 04:56:09,186 INFO [train.py:862] Epoch 27, batch 16500, loss[loss=0.1817, simple_loss=0.4553, pruned_loss=0.08439, ctc_loss=0.1259, over 14546.00 frames. ], tot_loss[loss=0.1735, simple_loss=0.428, pruned_loss=0.08246, ctc_loss=0.1209, over 2872603.45 frames. ], batch size: 48, lr: 3.00e-03, 
2023-01-06 04:57:16,476 INFO [train.py:862] Epoch 27, batch 17000, loss[loss=0.1705, simple_loss=0.427, pruned_loss=0.07789, ctc_loss=0.1187, over 13746.00 frames. ], tot_loss[loss=0.1756, simple_loss=0.4298, pruned_loss=0.08355, ctc_loss=0.123, over 2842391.25 frames. ], batch size: 63, lr: 3.00e-03, 
2023-01-06 04:58:24,155 INFO [train.py:862] Epoch 27, batch 17500, loss[loss=0.1561, simple_loss=0.4135, pruned_loss=0.07429, ctc_loss=0.1026, over 14667.00 frames. ], tot_loss[loss=0.1749, simple_loss=0.4279, pruned_loss=0.08332, ctc_loss=0.1224, over 2874204.19 frames. ], batch size: 40, lr: 3.00e-03, 
2023-01-06 04:59:30,154 INFO [train.py:862] Epoch 27, batch 18000, loss[loss=0.1568, simple_loss=0.3966, pruned_loss=0.0752, ctc_loss=0.1067, over 14441.00 frames. ], tot_loss[loss=0.1731, simple_loss=0.4265, pruned_loss=0.08219, ctc_loss=0.1207, over 2874788.20 frames. ], batch size: 33, lr: 3.00e-03, 
2023-01-06 04:59:30,156 INFO [train.py:887] Computing validation loss
2023-01-06 04:59:56,571 INFO [train.py:897] Epoch 27, validation: loss=0.1944, simple_loss=0.4465, pruned_loss=0.09561, ctc_loss=0.141, over 944034.00 frames. 
2023-01-06 04:59:56,571 INFO [train.py:898] Maximum memory allocated so far is 7351MB
2023-01-06 05:00:24,082 INFO [checkpoint.py:75] Saving checkpoint to tiny_transducer_ctc/exp_2m_phone_halfdelay/epoch-27.pt
2023-01-06 05:00:26,866 INFO [train.py:862] Epoch 28, batch 0, loss[loss=0.1998, simple_loss=0.4441, pruned_loss=0.1002, ctc_loss=0.1474, over 14895.00 frames. ], tot_loss[loss=0.1998, simple_loss=0.4441, pruned_loss=0.1002, ctc_loss=0.1474, over 14895.00 frames. ], batch size: 39, lr: 3.00e-03, 
2023-01-06 05:01:34,322 INFO [train.py:862] Epoch 28, batch 500, loss[loss=0.1672, simple_loss=0.3875, pruned_loss=0.08082, ctc_loss=0.1212, over 14554.00 frames. ], tot_loss[loss=0.1787, simple_loss=0.4326, pruned_loss=0.08539, ctc_loss=0.1259, over 2639729.81 frames. ], batch size: 34, lr: 3.00e-03, 
2023-01-06 05:02:41,983 INFO [train.py:862] Epoch 28, batch 1000, loss[loss=0.188, simple_loss=0.444, pruned_loss=0.0915, ctc_loss=0.1342, over 14737.00 frames. ], tot_loss[loss=0.1757, simple_loss=0.4295, pruned_loss=0.08345, ctc_loss=0.1231, over 2849031.45 frames. ], batch size: 45, lr: 3.00e-03, 
2023-01-06 05:03:49,302 INFO [train.py:862] Epoch 28, batch 1500, loss[loss=0.1409, simple_loss=0.3788, pruned_loss=0.06472, ctc_loss=0.09233, over 14549.00 frames. ], tot_loss[loss=0.1747, simple_loss=0.4283, pruned_loss=0.08292, ctc_loss=0.1223, over 2865784.79 frames. ], batch size: 34, lr: 3.00e-03, 
2023-01-06 05:04:56,219 INFO [train.py:862] Epoch 28, batch 2000, loss[loss=0.1582, simple_loss=0.4175, pruned_loss=0.07117, ctc_loss=0.106, over 14126.00 frames. ], tot_loss[loss=0.176, simple_loss=0.4296, pruned_loss=0.0838, ctc_loss=0.1234, over 2872785.89 frames. ], batch size: 56, lr: 3.00e-03, 
2023-01-06 05:06:03,121 INFO [train.py:862] Epoch 28, batch 2500, loss[loss=0.2173, simple_loss=0.4666, pruned_loss=0.1095, ctc_loss=0.1635, over 14824.00 frames. ], tot_loss[loss=0.1722, simple_loss=0.4266, pruned_loss=0.08121, ctc_loss=0.1198, over 2869368.04 frames. ], batch size: 41, lr: 3.00e-03, 
2023-01-06 05:07:10,737 INFO [train.py:862] Epoch 28, batch 3000, loss[loss=0.1595, simple_loss=0.3929, pruned_loss=0.07318, ctc_loss=0.1122, over 14038.00 frames. ], tot_loss[loss=0.1761, simple_loss=0.4292, pruned_loss=0.084, ctc_loss=0.1236, over 2862247.79 frames. ], batch size: 31, lr: 3.00e-03, 
2023-01-06 05:08:17,958 INFO [train.py:862] Epoch 28, batch 3500, loss[loss=0.2166, simple_loss=0.4763, pruned_loss=0.1136, ctc_loss=0.1587, over 10557.00 frames. ], tot_loss[loss=0.1761, simple_loss=0.4291, pruned_loss=0.08377, ctc_loss=0.1238, over 2871576.23 frames. ], batch size: 105, lr: 3.00e-03, 
2023-01-06 05:09:25,013 INFO [train.py:862] Epoch 28, batch 4000, loss[loss=0.1608, simple_loss=0.4031, pruned_loss=0.07614, ctc_loss=0.1107, over 14706.00 frames. ], tot_loss[loss=0.1762, simple_loss=0.4297, pruned_loss=0.08396, ctc_loss=0.1236, over 2859994.40 frames. ], batch size: 37, lr: 3.00e-03, 
2023-01-06 05:10:32,552 INFO [train.py:862] Epoch 28, batch 4500, loss[loss=0.1979, simple_loss=0.4187, pruned_loss=0.09934, ctc_loss=0.1504, over 14692.00 frames. ], tot_loss[loss=0.1765, simple_loss=0.4309, pruned_loss=0.08404, ctc_loss=0.1238, over 2858917.27 frames. ], batch size: 39, lr: 3.00e-03, 
2023-01-06 05:11:39,335 INFO [train.py:862] Epoch 28, batch 5000, loss[loss=0.1548, simple_loss=0.3932, pruned_loss=0.06947, ctc_loss=0.1071, over 14753.00 frames. ], tot_loss[loss=0.1739, simple_loss=0.4271, pruned_loss=0.08252, ctc_loss=0.1216, over 2877776.04 frames. ], batch size: 36, lr: 3.00e-03, 
2023-01-06 05:12:46,874 INFO [train.py:862] Epoch 28, batch 5500, loss[loss=0.2161, simple_loss=0.4561, pruned_loss=0.1108, ctc_loss=0.1636, over 9956.00 frames. ], tot_loss[loss=0.1759, simple_loss=0.4298, pruned_loss=0.08378, ctc_loss=0.1233, over 2844240.85 frames. ], batch size: 104, lr: 3.00e-03, 
2023-01-06 05:13:54,347 INFO [train.py:862] Epoch 28, batch 6000, loss[loss=0.1454, simple_loss=0.3962, pruned_loss=0.06443, ctc_loss=0.09522, over 14695.00 frames. ], tot_loss[loss=0.1759, simple_loss=0.4296, pruned_loss=0.08393, ctc_loss=0.1233, over 2863620.63 frames. ], batch size: 37, lr: 3.00e-03, 
2023-01-06 05:15:01,651 INFO [train.py:862] Epoch 28, batch 6500, loss[loss=0.2131, simple_loss=0.4709, pruned_loss=0.1055, ctc_loss=0.1583, over 13747.00 frames. ], tot_loss[loss=0.1765, simple_loss=0.4299, pruned_loss=0.08411, ctc_loss=0.1239, over 2868026.83 frames. ], batch size: 63, lr: 3.00e-03, 
2023-01-06 05:16:09,716 INFO [train.py:862] Epoch 28, batch 7000, loss[loss=0.1986, simple_loss=0.4555, pruned_loss=0.09956, ctc_loss=0.1434, over 14646.00 frames. ], tot_loss[loss=0.1767, simple_loss=0.4294, pruned_loss=0.0846, ctc_loss=0.1242, over 2861528.91 frames. ], batch size: 41, lr: 3.00e-03, 
2023-01-06 05:17:17,255 INFO [train.py:862] Epoch 28, batch 7500, loss[loss=0.1614, simple_loss=0.4045, pruned_loss=0.07457, ctc_loss=0.1119, over 14517.00 frames. ], tot_loss[loss=0.1753, simple_loss=0.4305, pruned_loss=0.08301, ctc_loss=0.1226, over 2864508.91 frames. ], batch size: 37, lr: 3.00e-03, 
2023-01-06 05:18:25,006 INFO [train.py:862] Epoch 28, batch 8000, loss[loss=0.142, simple_loss=0.3836, pruned_loss=0.06153, ctc_loss=0.0943, over 14773.00 frames. ], tot_loss[loss=0.1742, simple_loss=0.4277, pruned_loss=0.08252, ctc_loss=0.1219, over 2871041.88 frames. ], batch size: 36, lr: 3.00e-03, 
2023-01-06 05:18:47,274 INFO [checkpoint.py:75] Saving checkpoint to tiny_transducer_ctc/exp_2m_phone_halfdelay/checkpoint-500000.pt
2023-01-06 05:19:32,658 INFO [train.py:862] Epoch 28, batch 8500, loss[loss=0.1252, simple_loss=0.3928, pruned_loss=0.04747, ctc_loss=0.07431, over 14722.00 frames. ], tot_loss[loss=0.1761, simple_loss=0.4289, pruned_loss=0.08407, ctc_loss=0.1237, over 2861217.80 frames. ], batch size: 37, lr: 3.00e-03, 
2023-01-06 05:20:40,081 INFO [train.py:862] Epoch 28, batch 9000, loss[loss=0.1706, simple_loss=0.429, pruned_loss=0.08195, ctc_loss=0.1167, over 14653.00 frames. ], tot_loss[loss=0.1739, simple_loss=0.4279, pruned_loss=0.08236, ctc_loss=0.1214, over 2872821.60 frames. ], batch size: 40, lr: 3.00e-03, 
2023-01-06 05:20:40,082 INFO [train.py:887] Computing validation loss
2023-01-06 05:21:06,575 INFO [train.py:897] Epoch 28, validation: loss=0.1921, simple_loss=0.4449, pruned_loss=0.09407, ctc_loss=0.1388, over 944034.00 frames. 
2023-01-06 05:21:06,576 INFO [train.py:898] Maximum memory allocated so far is 7351MB
2023-01-06 05:22:13,932 INFO [train.py:862] Epoch 28, batch 9500, loss[loss=0.1573, simple_loss=0.4457, pruned_loss=0.06978, ctc_loss=0.09934, over 14689.00 frames. ], tot_loss[loss=0.1759, simple_loss=0.4297, pruned_loss=0.08366, ctc_loss=0.1234, over 2874510.37 frames. ], batch size: 44, lr: 3.00e-03, 
2023-01-06 05:23:21,671 INFO [train.py:862] Epoch 28, batch 10000, loss[loss=0.1754, simple_loss=0.4411, pruned_loss=0.08193, ctc_loss=0.1209, over 14826.00 frames. ], tot_loss[loss=0.1744, simple_loss=0.4285, pruned_loss=0.08279, ctc_loss=0.1219, over 2866031.24 frames. ], batch size: 41, lr: 3.00e-03, 
2023-01-06 05:24:29,793 INFO [train.py:862] Epoch 28, batch 10500, loss[loss=0.2112, simple_loss=0.4499, pruned_loss=0.1083, ctc_loss=0.1589, over 9738.00 frames. ], tot_loss[loss=0.1773, simple_loss=0.4307, pruned_loss=0.08453, ctc_loss=0.1248, over 2868104.43 frames. ], batch size: 104, lr: 3.00e-03, 
2023-01-06 05:25:37,025 INFO [train.py:862] Epoch 28, batch 11000, loss[loss=0.1968, simple_loss=0.4312, pruned_loss=0.09855, ctc_loss=0.1466, over 14660.00 frames. ], tot_loss[loss=0.1757, simple_loss=0.4287, pruned_loss=0.08381, ctc_loss=0.1232, over 2866401.57 frames. ], batch size: 40, lr: 3.00e-03, 
2023-01-06 05:26:44,962 INFO [train.py:862] Epoch 28, batch 11500, loss[loss=0.1693, simple_loss=0.4411, pruned_loss=0.07709, ctc_loss=0.1143, over 14801.00 frames. ], tot_loss[loss=0.1755, simple_loss=0.4305, pruned_loss=0.08349, ctc_loss=0.1226, over 2863851.46 frames. ], batch size: 43, lr: 3.00e-03, 
2023-01-06 05:27:52,493 INFO [train.py:862] Epoch 28, batch 12000, loss[loss=0.1668, simple_loss=0.3965, pruned_loss=0.08219, ctc_loss=0.1181, over 14044.00 frames. ], tot_loss[loss=0.1741, simple_loss=0.4274, pruned_loss=0.08264, ctc_loss=0.1217, over 2870759.57 frames. ], batch size: 31, lr: 3.00e-03, 
2023-01-06 05:29:00,741 INFO [train.py:862] Epoch 28, batch 12500, loss[loss=0.1408, simple_loss=0.3941, pruned_loss=0.05519, ctc_loss=0.09298, over 14518.00 frames. ], tot_loss[loss=0.1744, simple_loss=0.4291, pruned_loss=0.08274, ctc_loss=0.1217, over 2859571.16 frames. ], batch size: 38, lr: 3.00e-03, 
2023-01-06 05:30:08,575 INFO [train.py:862] Epoch 28, batch 13000, loss[loss=0.1613, simple_loss=0.3932, pruned_loss=0.07539, ctc_loss=0.1138, over 13996.00 frames. ], tot_loss[loss=0.1754, simple_loss=0.4292, pruned_loss=0.08351, ctc_loss=0.1228, over 2832011.44 frames. ], batch size: 31, lr: 3.00e-03, 
2023-01-06 05:31:16,821 INFO [train.py:862] Epoch 28, batch 13500, loss[loss=0.1757, simple_loss=0.4548, pruned_loss=0.08007, ctc_loss=0.1193, over 14549.00 frames. ], tot_loss[loss=0.1767, simple_loss=0.4309, pruned_loss=0.08422, ctc_loss=0.124, over 2852009.69 frames. ], batch size: 48, lr: 3.00e-03, 
2023-01-06 05:32:24,314 INFO [train.py:862] Epoch 28, batch 14000, loss[loss=0.1764, simple_loss=0.4531, pruned_loss=0.08328, ctc_loss=0.1192, over 14662.00 frames. ], tot_loss[loss=0.1755, simple_loss=0.4295, pruned_loss=0.08335, ctc_loss=0.123, over 2867668.11 frames. ], batch size: 47, lr: 3.00e-03, 
2023-01-06 05:33:31,671 INFO [train.py:862] Epoch 28, batch 14500, loss[loss=0.2053, simple_loss=0.457, pruned_loss=0.1049, ctc_loss=0.1503, over 14667.00 frames. ], tot_loss[loss=0.1749, simple_loss=0.4284, pruned_loss=0.08326, ctc_loss=0.1224, over 2868585.65 frames. ], batch size: 44, lr: 3.00e-03, 
2023-01-06 05:34:39,659 INFO [train.py:862] Epoch 28, batch 15000, loss[loss=0.2088, simple_loss=0.4702, pruned_loss=0.1081, ctc_loss=0.1512, over 14508.00 frames. ], tot_loss[loss=0.1758, simple_loss=0.4278, pruned_loss=0.08397, ctc_loss=0.1235, over 2861223.15 frames. ], batch size: 50, lr: 3.00e-03, 
2023-01-06 05:35:46,973 INFO [train.py:862] Epoch 28, batch 15500, loss[loss=0.1532, simple_loss=0.4373, pruned_loss=0.06486, ctc_loss=0.0974, over 14558.00 frames. ], tot_loss[loss=0.1753, simple_loss=0.4293, pruned_loss=0.08325, ctc_loss=0.1227, over 2869571.78 frames. ], batch size: 48, lr: 3.00e-03, 
2023-01-06 05:36:54,541 INFO [train.py:862] Epoch 28, batch 16000, loss[loss=0.1975, simple_loss=0.4508, pruned_loss=0.09714, ctc_loss=0.1439, over 14678.00 frames. ], tot_loss[loss=0.174, simple_loss=0.4274, pruned_loss=0.0827, ctc_loss=0.1215, over 2856121.27 frames. ], batch size: 47, lr: 3.00e-03, 
2023-01-06 05:38:02,040 INFO [train.py:862] Epoch 28, batch 16500, loss[loss=0.1627, simple_loss=0.4382, pruned_loss=0.07412, ctc_loss=0.1068, over 14562.00 frames. ], tot_loss[loss=0.1741, simple_loss=0.4298, pruned_loss=0.08242, ctc_loss=0.1213, over 2873327.99 frames. ], batch size: 53, lr: 3.00e-03, 
2023-01-06 05:39:09,417 INFO [train.py:862] Epoch 28, batch 17000, loss[loss=0.2443, simple_loss=0.4891, pruned_loss=0.1291, ctc_loss=0.1888, over 14708.00 frames. ], tot_loss[loss=0.1747, simple_loss=0.4287, pruned_loss=0.08293, ctc_loss=0.1221, over 2859553.60 frames. ], batch size: 44, lr: 3.00e-03, 
2023-01-06 05:40:16,827 INFO [train.py:862] Epoch 28, batch 17500, loss[loss=0.137, simple_loss=0.3871, pruned_loss=0.06031, ctc_loss=0.08686, over 14807.00 frames. ], tot_loss[loss=0.1731, simple_loss=0.4273, pruned_loss=0.0821, ctc_loss=0.1206, over 2863150.91 frames. ], batch size: 36, lr: 3.00e-03, 
2023-01-06 05:41:22,751 INFO [train.py:862] Epoch 28, batch 18000, loss[loss=0.1583, simple_loss=0.4008, pruned_loss=0.07098, ctc_loss=0.1099, over 14529.00 frames. ], tot_loss[loss=0.1748, simple_loss=0.4288, pruned_loss=0.08295, ctc_loss=0.1223, over 2862795.96 frames. ], batch size: 36, lr: 3.00e-03, 
2023-01-06 05:41:22,751 INFO [train.py:887] Computing validation loss
2023-01-06 05:41:49,346 INFO [train.py:897] Epoch 28, validation: loss=0.1915, simple_loss=0.4447, pruned_loss=0.09411, ctc_loss=0.138, over 944034.00 frames. 
2023-01-06 05:41:49,347 INFO [train.py:898] Maximum memory allocated so far is 7351MB
2023-01-06 05:42:10,927 INFO [checkpoint.py:75] Saving checkpoint to tiny_transducer_ctc/exp_2m_phone_halfdelay/checkpoint-510000.pt
2023-01-06 05:42:16,765 INFO [checkpoint.py:75] Saving checkpoint to tiny_transducer_ctc/exp_2m_phone_halfdelay/epoch-28.pt
2023-01-06 05:42:20,000 INFO [train.py:862] Epoch 29, batch 0, loss[loss=0.2221, simple_loss=0.474, pruned_loss=0.1099, ctc_loss=0.1686, over 14042.00 frames. ], tot_loss[loss=0.2221, simple_loss=0.474, pruned_loss=0.1099, ctc_loss=0.1686, over 14042.00 frames. ], batch size: 56, lr: 3.00e-03, 
2023-01-06 05:43:27,256 INFO [train.py:862] Epoch 29, batch 500, loss[loss=0.1799, simple_loss=0.4231, pruned_loss=0.0892, ctc_loss=0.1281, over 14887.00 frames. ], tot_loss[loss=0.1761, simple_loss=0.4301, pruned_loss=0.08377, ctc_loss=0.1235, over 2644017.21 frames. ], batch size: 40, lr: 3.00e-03, 
2023-01-06 05:44:34,359 INFO [train.py:862] Epoch 29, batch 1000, loss[loss=0.1646, simple_loss=0.3827, pruned_loss=0.08171, ctc_loss=0.1181, over 14531.00 frames. ], tot_loss[loss=0.1743, simple_loss=0.4285, pruned_loss=0.08269, ctc_loss=0.1218, over 2859038.61 frames. ], batch size: 34, lr: 3.00e-03, 
2023-01-06 05:45:42,119 INFO [train.py:862] Epoch 29, batch 1500, loss[loss=0.2037, simple_loss=0.4773, pruned_loss=0.09849, ctc_loss=0.1466, over 14514.00 frames. ], tot_loss[loss=0.1751, simple_loss=0.4297, pruned_loss=0.08326, ctc_loss=0.1224, over 2877958.23 frames. ], batch size: 50, lr: 3.00e-03, 
2023-01-06 05:46:49,633 INFO [train.py:862] Epoch 29, batch 2000, loss[loss=0.1584, simple_loss=0.4015, pruned_loss=0.07389, ctc_loss=0.1085, over 14705.00 frames. ], tot_loss[loss=0.1746, simple_loss=0.4277, pruned_loss=0.08316, ctc_loss=0.1222, over 2875643.25 frames. ], batch size: 37, lr: 3.00e-03, 
2023-01-06 05:47:56,722 INFO [train.py:862] Epoch 29, batch 2500, loss[loss=0.1671, simple_loss=0.4314, pruned_loss=0.08002, ctc_loss=0.112, over 14797.00 frames. ], tot_loss[loss=0.1745, simple_loss=0.4281, pruned_loss=0.08282, ctc_loss=0.1221, over 2862680.26 frames. ], batch size: 43, lr: 3.00e-03, 
2023-01-06 05:49:04,561 INFO [train.py:862] Epoch 29, batch 3000, loss[loss=0.2072, simple_loss=0.4484, pruned_loss=0.1076, ctc_loss=0.1538, over 10422.00 frames. ], tot_loss[loss=0.1763, simple_loss=0.4311, pruned_loss=0.08381, ctc_loss=0.1235, over 2860690.30 frames. ], batch size: 104, lr: 3.00e-03, 
2023-01-06 05:50:11,929 INFO [train.py:862] Epoch 29, batch 3500, loss[loss=0.2031, simple_loss=0.444, pruned_loss=0.101, ctc_loss=0.1517, over 9988.00 frames. ], tot_loss[loss=0.1753, simple_loss=0.429, pruned_loss=0.08329, ctc_loss=0.1227, over 2871731.97 frames. ], batch size: 104, lr: 3.00e-03, 
2023-01-06 05:51:19,265 INFO [train.py:862] Epoch 29, batch 4000, loss[loss=0.1266, simple_loss=0.3672, pruned_loss=0.05783, ctc_loss=0.07736, over 14524.00 frames. ], tot_loss[loss=0.1746, simple_loss=0.4292, pruned_loss=0.08276, ctc_loss=0.1219, over 2876500.72 frames. ], batch size: 34, lr: 3.00e-03, 
2023-01-06 05:52:27,434 INFO [train.py:862] Epoch 29, batch 4500, loss[loss=0.1834, simple_loss=0.4448, pruned_loss=0.08783, ctc_loss=0.129, over 12935.00 frames. ], tot_loss[loss=0.1737, simple_loss=0.4267, pruned_loss=0.08222, ctc_loss=0.1215, over 2870953.59 frames. ], batch size: 76, lr: 3.00e-03, 
2023-01-06 05:53:34,776 INFO [train.py:862] Epoch 29, batch 5000, loss[loss=0.1656, simple_loss=0.4086, pruned_loss=0.07751, ctc_loss=0.1158, over 14881.00 frames. ], tot_loss[loss=0.1753, simple_loss=0.429, pruned_loss=0.08329, ctc_loss=0.1228, over 2875955.43 frames. ], batch size: 39, lr: 3.00e-03, 
2023-01-06 05:54:41,369 INFO [train.py:862] Epoch 29, batch 5500, loss[loss=0.1494, simple_loss=0.4155, pruned_loss=0.0636, ctc_loss=0.09718, over 14831.00 frames. ], tot_loss[loss=0.1754, simple_loss=0.4309, pruned_loss=0.08322, ctc_loss=0.1225, over 2863069.50 frames. ], batch size: 41, lr: 3.00e-03, 
2023-01-06 05:55:48,681 INFO [train.py:862] Epoch 29, batch 6000, loss[loss=0.1483, simple_loss=0.4105, pruned_loss=0.06971, ctc_loss=0.09397, over 14662.00 frames. ], tot_loss[loss=0.1753, simple_loss=0.4295, pruned_loss=0.0835, ctc_loss=0.1226, over 2857393.04 frames. ], batch size: 40, lr: 3.00e-03, 
2023-01-06 05:56:56,073 INFO [train.py:862] Epoch 29, batch 6500, loss[loss=0.203, simple_loss=0.4511, pruned_loss=0.1037, ctc_loss=0.1489, over 13791.00 frames. ], tot_loss[loss=0.1747, simple_loss=0.4278, pruned_loss=0.08308, ctc_loss=0.1223, over 2857767.63 frames. ], batch size: 64, lr: 3.00e-03, 
2023-01-06 05:58:03,728 INFO [train.py:862] Epoch 29, batch 7000, loss[loss=0.1889, simple_loss=0.4604, pruned_loss=0.08649, ctc_loss=0.1341, over 14748.00 frames. ], tot_loss[loss=0.1752, simple_loss=0.4285, pruned_loss=0.08344, ctc_loss=0.1227, over 2867924.17 frames. ], batch size: 45, lr: 3.00e-03, 
2023-01-06 05:59:11,166 INFO [train.py:862] Epoch 29, batch 7500, loss[loss=0.1909, simple_loss=0.4564, pruned_loss=0.09671, ctc_loss=0.1334, over 13934.00 frames. ], tot_loss[loss=0.1749, simple_loss=0.428, pruned_loss=0.08312, ctc_loss=0.1225, over 2879183.32 frames. ], batch size: 56, lr: 3.00e-03, 
2023-01-06 06:00:18,797 INFO [train.py:862] Epoch 29, batch 8000, loss[loss=0.1527, simple_loss=0.3827, pruned_loss=0.07057, ctc_loss=0.106, over 14409.00 frames. ], tot_loss[loss=0.1764, simple_loss=0.43, pruned_loss=0.08402, ctc_loss=0.1238, over 2854044.81 frames. ], batch size: 33, lr: 3.00e-03, 
2023-01-06 06:01:26,669 INFO [train.py:862] Epoch 29, batch 8500, loss[loss=0.1875, simple_loss=0.4556, pruned_loss=0.0893, ctc_loss=0.132, over 14813.00 frames. ], tot_loss[loss=0.1733, simple_loss=0.4288, pruned_loss=0.08206, ctc_loss=0.1206, over 2863934.36 frames. ], batch size: 41, lr: 3.00e-03, 
2023-01-06 06:02:33,891 INFO [train.py:862] Epoch 29, batch 9000, loss[loss=0.1744, simple_loss=0.4628, pruned_loss=0.08037, ctc_loss=0.1155, over 14721.00 frames. ], tot_loss[loss=0.1736, simple_loss=0.4271, pruned_loss=0.08227, ctc_loss=0.1212, over 2874743.05 frames. ], batch size: 45, lr: 3.00e-03, 
2023-01-06 06:02:33,892 INFO [train.py:887] Computing validation loss
2023-01-06 06:03:00,581 INFO [train.py:897] Epoch 29, validation: loss=0.1925, simple_loss=0.4455, pruned_loss=0.0953, ctc_loss=0.1387, over 944034.00 frames. 
2023-01-06 06:03:00,582 INFO [train.py:898] Maximum memory allocated so far is 7351MB
2023-01-06 06:04:07,776 INFO [train.py:862] Epoch 29, batch 9500, loss[loss=0.1473, simple_loss=0.4178, pruned_loss=0.06656, ctc_loss=0.09244, over 14668.00 frames. ], tot_loss[loss=0.1751, simple_loss=0.4284, pruned_loss=0.0833, ctc_loss=0.1226, over 2866866.24 frames. ], batch size: 41, lr: 3.00e-03, 
2023-01-06 06:05:08,589 INFO [checkpoint.py:75] Saving checkpoint to tiny_transducer_ctc/exp_2m_phone_halfdelay/checkpoint-520000.pt
2023-01-06 06:05:14,949 INFO [train.py:862] Epoch 29, batch 10000, loss[loss=0.1908, simple_loss=0.4288, pruned_loss=0.09876, ctc_loss=0.1383, over 14702.00 frames. ], tot_loss[loss=0.1748, simple_loss=0.4286, pruned_loss=0.08308, ctc_loss=0.1223, over 2866397.23 frames. ], batch size: 39, lr: 3.00e-03, 
2023-01-06 06:06:22,527 INFO [train.py:862] Epoch 29, batch 10500, loss[loss=0.2174, simple_loss=0.4684, pruned_loss=0.1067, ctc_loss=0.1644, over 14589.00 frames. ], tot_loss[loss=0.1749, simple_loss=0.4284, pruned_loss=0.08325, ctc_loss=0.1224, over 2854680.52 frames. ], batch size: 43, lr: 3.00e-03, 
2023-01-06 06:07:29,443 INFO [train.py:862] Epoch 29, batch 11000, loss[loss=0.1875, simple_loss=0.4455, pruned_loss=0.08609, ctc_loss=0.1355, over 14559.00 frames. ], tot_loss[loss=0.1739, simple_loss=0.4277, pruned_loss=0.08247, ctc_loss=0.1214, over 2859437.38 frames. ], batch size: 48, lr: 3.00e-03, 
2023-01-06 06:08:37,243 INFO [train.py:862] Epoch 29, batch 11500, loss[loss=0.1862, simple_loss=0.4444, pruned_loss=0.09015, ctc_loss=0.1321, over 14655.00 frames. ], tot_loss[loss=0.1744, simple_loss=0.4287, pruned_loss=0.08267, ctc_loss=0.1218, over 2861035.89 frames. ], batch size: 41, lr: 3.00e-03, 
2023-01-06 06:09:44,277 INFO [train.py:862] Epoch 29, batch 12000, loss[loss=0.1442, simple_loss=0.4148, pruned_loss=0.05842, ctc_loss=0.09204, over 14811.00 frames. ], tot_loss[loss=0.175, simple_loss=0.429, pruned_loss=0.08314, ctc_loss=0.1225, over 2864132.25 frames. ], batch size: 41, lr: 3.00e-03, 
2023-01-06 06:10:51,954 INFO [train.py:862] Epoch 29, batch 12500, loss[loss=0.1675, simple_loss=0.4112, pruned_loss=0.08015, ctc_loss=0.1169, over 14782.00 frames. ], tot_loss[loss=0.1756, simple_loss=0.4292, pruned_loss=0.08355, ctc_loss=0.123, over 2860796.66 frames. ], batch size: 36, lr: 3.00e-03, 
2023-01-06 06:11:59,338 INFO [train.py:862] Epoch 29, batch 13000, loss[loss=0.1627, simple_loss=0.4064, pruned_loss=0.08096, ctc_loss=0.1106, over 14073.00 frames. ], tot_loss[loss=0.1724, simple_loss=0.4265, pruned_loss=0.08143, ctc_loss=0.1199, over 2872232.61 frames. ], batch size: 31, lr: 3.00e-03, 
2023-01-06 06:13:06,367 INFO [train.py:862] Epoch 29, batch 13500, loss[loss=0.1605, simple_loss=0.39, pruned_loss=0.07713, ctc_loss=0.1127, over 14415.00 frames. ], tot_loss[loss=0.1738, simple_loss=0.4286, pruned_loss=0.082, ctc_loss=0.1213, over 2860478.75 frames. ], batch size: 33, lr: 3.00e-03, 
2023-01-06 06:14:13,609 INFO [train.py:862] Epoch 29, batch 14000, loss[loss=0.1735, simple_loss=0.4147, pruned_loss=0.08787, ctc_loss=0.1213, over 14875.00 frames. ], tot_loss[loss=0.1756, simple_loss=0.4285, pruned_loss=0.08359, ctc_loss=0.1232, over 2867949.95 frames. ], batch size: 40, lr: 3.00e-03, 
2023-01-06 06:15:21,046 INFO [train.py:862] Epoch 29, batch 14500, loss[loss=0.1709, simple_loss=0.4524, pruned_loss=0.07591, ctc_loss=0.1146, over 14587.00 frames. ], tot_loss[loss=0.1742, simple_loss=0.4289, pruned_loss=0.08243, ctc_loss=0.1217, over 2863917.31 frames. ], batch size: 43, lr: 3.00e-03, 
2023-01-06 06:16:28,790 INFO [train.py:862] Epoch 29, batch 15000, loss[loss=0.1886, simple_loss=0.4349, pruned_loss=0.09013, ctc_loss=0.1376, over 14660.00 frames. ], tot_loss[loss=0.1739, simple_loss=0.4281, pruned_loss=0.08239, ctc_loss=0.1214, over 2857876.90 frames. ], batch size: 40, lr: 3.00e-03, 
2023-01-06 06:17:36,004 INFO [train.py:862] Epoch 29, batch 15500, loss[loss=0.1539, simple_loss=0.4198, pruned_loss=0.07107, ctc_loss=0.0995, over 14507.00 frames. ], tot_loss[loss=0.1746, simple_loss=0.429, pruned_loss=0.08282, ctc_loss=0.122, over 2840856.58 frames. ], batch size: 37, lr: 3.00e-03, 
2023-01-06 06:18:42,954 INFO [train.py:862] Epoch 29, batch 16000, loss[loss=0.1529, simple_loss=0.3855, pruned_loss=0.07293, ctc_loss=0.1045, over 14666.00 frames. ], tot_loss[loss=0.1736, simple_loss=0.4272, pruned_loss=0.08239, ctc_loss=0.1211, over 2862457.60 frames. ], batch size: 35, lr: 3.00e-03, 
2023-01-06 06:19:51,032 INFO [train.py:862] Epoch 29, batch 16500, loss[loss=0.1938, simple_loss=0.4311, pruned_loss=0.09662, ctc_loss=0.143, over 14703.00 frames. ], tot_loss[loss=0.1735, simple_loss=0.4276, pruned_loss=0.08206, ctc_loss=0.1211, over 2868216.09 frames. ], batch size: 38, lr: 3.00e-03, 
2023-01-06 06:20:58,529 INFO [train.py:862] Epoch 29, batch 17000, loss[loss=0.1547, simple_loss=0.4079, pruned_loss=0.07048, ctc_loss=0.1034, over 14517.00 frames. ], tot_loss[loss=0.174, simple_loss=0.4268, pruned_loss=0.08252, ctc_loss=0.1217, over 2862471.42 frames. ], batch size: 37, lr: 3.00e-03, 
2023-01-06 06:22:05,871 INFO [train.py:862] Epoch 29, batch 17500, loss[loss=0.1471, simple_loss=0.4206, pruned_loss=0.06339, ctc_loss=0.09289, over 14795.00 frames. ], tot_loss[loss=0.1747, simple_loss=0.4284, pruned_loss=0.08308, ctc_loss=0.1222, over 2864872.29 frames. ], batch size: 43, lr: 3.00e-03, 
2023-01-06 06:23:11,483 INFO [train.py:862] Epoch 29, batch 18000, loss[loss=0.1986, simple_loss=0.4339, pruned_loss=0.09754, ctc_loss=0.1489, over 10187.00 frames. ], tot_loss[loss=0.1773, simple_loss=0.43, pruned_loss=0.0848, ctc_loss=0.1247, over 2853596.05 frames. ], batch size: 104, lr: 3.00e-03, 
2023-01-06 06:23:11,484 INFO [train.py:887] Computing validation loss
2023-01-06 06:23:38,277 INFO [train.py:897] Epoch 29, validation: loss=0.1906, simple_loss=0.4441, pruned_loss=0.09291, ctc_loss=0.1374, over 944034.00 frames. 
2023-01-06 06:23:38,278 INFO [train.py:898] Maximum memory allocated so far is 7351MB
2023-01-06 06:24:05,440 INFO [checkpoint.py:75] Saving checkpoint to tiny_transducer_ctc/exp_2m_phone_halfdelay/epoch-29.pt
2023-01-06 06:24:08,216 INFO [train.py:862] Epoch 30, batch 0, loss[loss=0.2056, simple_loss=0.4496, pruned_loss=0.1056, ctc_loss=0.152, over 14647.00 frames. ], tot_loss[loss=0.2056, simple_loss=0.4496, pruned_loss=0.1056, ctc_loss=0.152, over 14647.00 frames. ], batch size: 40, lr: 3.00e-03, 
2023-01-06 06:25:15,826 INFO [train.py:862] Epoch 30, batch 500, loss[loss=0.1978, simple_loss=0.4657, pruned_loss=0.09948, ctc_loss=0.1402, over 14555.00 frames. ], tot_loss[loss=0.1738, simple_loss=0.4278, pruned_loss=0.08212, ctc_loss=0.1214, over 2640320.64 frames. ], batch size: 53, lr: 3.00e-03, 
2023-01-06 06:26:22,935 INFO [train.py:862] Epoch 30, batch 1000, loss[loss=0.2142, simple_loss=0.4696, pruned_loss=0.1055, ctc_loss=0.1602, over 14723.00 frames. ], tot_loss[loss=0.1757, simple_loss=0.4301, pruned_loss=0.08345, ctc_loss=0.1231, over 2832565.80 frames. ], batch size: 47, lr: 3.00e-03, 
2023-01-06 06:27:30,181 INFO [train.py:862] Epoch 30, batch 1500, loss[loss=0.2133, simple_loss=0.423, pruned_loss=0.1184, ctc_loss=0.1633, over 14380.00 frames. ], tot_loss[loss=0.1761, simple_loss=0.4286, pruned_loss=0.08397, ctc_loss=0.1238, over 2857445.39 frames. ], batch size: 33, lr: 3.00e-03, 
2023-01-06 06:28:02,093 INFO [checkpoint.py:75] Saving checkpoint to tiny_transducer_ctc/exp_2m_phone_halfdelay/checkpoint-530000.pt
2023-01-06 06:28:37,531 INFO [train.py:862] Epoch 30, batch 2000, loss[loss=0.1531, simple_loss=0.4255, pruned_loss=0.06855, ctc_loss=0.09821, over 14578.00 frames. ], tot_loss[loss=0.1737, simple_loss=0.428, pruned_loss=0.08239, ctc_loss=0.1211, over 2879763.02 frames. ], batch size: 43, lr: 3.00e-03, 
2023-01-06 06:29:44,630 INFO [train.py:862] Epoch 30, batch 2500, loss[loss=0.1239, simple_loss=0.3912, pruned_loss=0.04871, ctc_loss=0.07233, over 14711.00 frames. ], tot_loss[loss=0.1728, simple_loss=0.4284, pruned_loss=0.08149, ctc_loss=0.1202, over 2881437.42 frames. ], batch size: 39, lr: 3.00e-03, 
2023-01-06 06:30:51,920 INFO [train.py:862] Epoch 30, batch 3000, loss[loss=0.1503, simple_loss=0.4103, pruned_loss=0.06776, ctc_loss=0.09771, over 14696.00 frames. ], tot_loss[loss=0.173, simple_loss=0.4257, pruned_loss=0.08189, ctc_loss=0.1208, over 2871450.01 frames. ], batch size: 39, lr: 3.00e-03, 
2023-01-06 06:31:59,438 INFO [train.py:862] Epoch 30, batch 3500, loss[loss=0.1462, simple_loss=0.4137, pruned_loss=0.06503, ctc_loss=0.09228, over 14527.00 frames. ], tot_loss[loss=0.1749, simple_loss=0.4284, pruned_loss=0.08322, ctc_loss=0.1223, over 2875798.75 frames. ], batch size: 38, lr: 3.00e-03, 
2023-01-06 06:33:06,705 INFO [train.py:862] Epoch 30, batch 4000, loss[loss=0.1629, simple_loss=0.4254, pruned_loss=0.07154, ctc_loss=0.1109, over 14522.00 frames. ], tot_loss[loss=0.1738, simple_loss=0.4279, pruned_loss=0.08242, ctc_loss=0.1212, over 2876699.65 frames. ], batch size: 53, lr: 3.00e-03, 
2023-01-06 06:34:14,701 INFO [train.py:862] Epoch 30, batch 4500, loss[loss=0.1296, simple_loss=0.3733, pruned_loss=0.05528, ctc_loss=0.08144, over 14691.00 frames. ], tot_loss[loss=0.1749, simple_loss=0.4291, pruned_loss=0.08291, ctc_loss=0.1224, over 2849629.06 frames. ], batch size: 35, lr: 3.00e-03, 
2023-01-06 06:35:22,385 INFO [train.py:862] Epoch 30, batch 5000, loss[loss=0.1532, simple_loss=0.4059, pruned_loss=0.06892, ctc_loss=0.1024, over 14714.00 frames. ], tot_loss[loss=0.1745, simple_loss=0.4278, pruned_loss=0.08271, ctc_loss=0.1221, over 2849230.52 frames. ], batch size: 37, lr: 3.00e-03, 
2023-01-06 06:36:30,507 INFO [train.py:862] Epoch 30, batch 5500, loss[loss=0.1628, simple_loss=0.4197, pruned_loss=0.07646, ctc_loss=0.1098, over 14721.00 frames. ], tot_loss[loss=0.1765, simple_loss=0.4292, pruned_loss=0.08432, ctc_loss=0.124, over 2866420.03 frames. ], batch size: 37, lr: 3.00e-03, 
2023-01-06 06:37:37,576 INFO [train.py:862] Epoch 30, batch 6000, loss[loss=0.1588, simple_loss=0.4121, pruned_loss=0.075, ctc_loss=0.1065, over 14650.00 frames. ], tot_loss[loss=0.1743, simple_loss=0.4286, pruned_loss=0.08263, ctc_loss=0.1217, over 2874813.35 frames. ], batch size: 40, lr: 3.00e-03, 
2023-01-06 06:38:44,594 INFO [train.py:862] Epoch 30, batch 6500, loss[loss=0.1814, simple_loss=0.4448, pruned_loss=0.08544, ctc_loss=0.1273, over 13765.00 frames. ], tot_loss[loss=0.1743, simple_loss=0.4279, pruned_loss=0.08281, ctc_loss=0.1218, over 2866244.42 frames. ], batch size: 63, lr: 3.00e-03, 
2023-01-06 06:39:52,215 INFO [train.py:862] Epoch 30, batch 7000, loss[loss=0.221, simple_loss=0.4294, pruned_loss=0.1226, ctc_loss=0.1712, over 14679.00 frames. ], tot_loss[loss=0.1754, simple_loss=0.429, pruned_loss=0.08368, ctc_loss=0.1228, over 2862850.79 frames. ], batch size: 35, lr: 3.00e-03, 
2023-01-06 06:40:59,438 INFO [train.py:862] Epoch 30, batch 7500, loss[loss=0.1528, simple_loss=0.4004, pruned_loss=0.06641, ctc_loss=0.1041, over 14713.00 frames. ], tot_loss[loss=0.1729, simple_loss=0.4279, pruned_loss=0.08173, ctc_loss=0.1203, over 2861941.39 frames. ], batch size: 37, lr: 3.00e-03, 
2023-01-06 06:42:06,698 INFO [train.py:862] Epoch 30, batch 8000, loss[loss=0.1965, simple_loss=0.4604, pruned_loss=0.0964, ctc_loss=0.1407, over 14565.00 frames. ], tot_loss[loss=0.1753, simple_loss=0.4296, pruned_loss=0.08334, ctc_loss=0.1227, over 2855106.01 frames. ], batch size: 43, lr: 3.00e-03, 
2023-01-06 06:43:13,807 INFO [train.py:862] Epoch 30, batch 8500, loss[loss=0.1751, simple_loss=0.3911, pruned_loss=0.08624, ctc_loss=0.1294, over 14052.00 frames. ], tot_loss[loss=0.1732, simple_loss=0.4265, pruned_loss=0.08206, ctc_loss=0.1209, over 2879777.48 frames. ], batch size: 31, lr: 3.00e-03, 
2023-01-06 06:44:20,886 INFO [train.py:862] Epoch 30, batch 9000, loss[loss=0.1699, simple_loss=0.4466, pruned_loss=0.07774, ctc_loss=0.1137, over 14796.00 frames. ], tot_loss[loss=0.174, simple_loss=0.4285, pruned_loss=0.08238, ctc_loss=0.1215, over 2872309.52 frames. ], batch size: 43, lr: 3.00e-03, 
2023-01-06 06:44:20,887 INFO [train.py:887] Computing validation loss
2023-01-06 06:44:47,238 INFO [train.py:897] Epoch 30, validation: loss=0.1899, simple_loss=0.4435, pruned_loss=0.09243, ctc_loss=0.1366, over 944034.00 frames. 
2023-01-06 06:44:47,238 INFO [train.py:898] Maximum memory allocated so far is 7351MB
2023-01-06 06:45:54,270 INFO [train.py:862] Epoch 30, batch 9500, loss[loss=0.2035, simple_loss=0.4417, pruned_loss=0.1017, ctc_loss=0.1524, over 14705.00 frames. ], tot_loss[loss=0.1748, simple_loss=0.4282, pruned_loss=0.08315, ctc_loss=0.1223, over 2851724.70 frames. ], batch size: 38, lr: 3.00e-03, 
2023-01-06 06:47:02,408 INFO [train.py:862] Epoch 30, batch 10000, loss[loss=0.1761, simple_loss=0.4339, pruned_loss=0.08456, ctc_loss=0.1223, over 14662.00 frames. ], tot_loss[loss=0.1746, simple_loss=0.4282, pruned_loss=0.08291, ctc_loss=0.1221, over 2860583.59 frames. ], batch size: 42, lr: 3.00e-03, 
2023-01-06 06:48:09,614 INFO [train.py:862] Epoch 30, batch 10500, loss[loss=0.1316, simple_loss=0.3735, pruned_loss=0.05898, ctc_loss=0.08264, over 14685.00 frames. ], tot_loss[loss=0.1745, simple_loss=0.4279, pruned_loss=0.08276, ctc_loss=0.1221, over 2864903.59 frames. ], batch size: 35, lr: 3.00e-03, 
2023-01-06 06:49:17,236 INFO [train.py:862] Epoch 30, batch 11000, loss[loss=0.1696, simple_loss=0.4158, pruned_loss=0.08068, ctc_loss=0.1186, over 14864.00 frames. ], tot_loss[loss=0.1725, simple_loss=0.427, pruned_loss=0.08141, ctc_loss=0.12, over 2867640.07 frames. ], batch size: 40, lr: 3.00e-03, 
2023-01-06 06:50:25,233 INFO [train.py:862] Epoch 30, batch 11500, loss[loss=0.1501, simple_loss=0.4093, pruned_loss=0.07075, ctc_loss=0.09645, over 14514.00 frames. ], tot_loss[loss=0.1753, simple_loss=0.43, pruned_loss=0.08341, ctc_loss=0.1226, over 2862952.16 frames. ], batch size: 38, lr: 3.00e-03, 
2023-01-06 06:50:57,465 INFO [checkpoint.py:75] Saving checkpoint to tiny_transducer_ctc/exp_2m_phone_halfdelay/checkpoint-540000.pt
2023-01-06 06:51:33,287 INFO [train.py:862] Epoch 30, batch 12000, loss[loss=0.2077, simple_loss=0.4443, pruned_loss=0.1027, ctc_loss=0.1575, over 10459.00 frames. ], tot_loss[loss=0.1735, simple_loss=0.4277, pruned_loss=0.08211, ctc_loss=0.121, over 2845663.84 frames. ], batch size: 104, lr: 3.00e-03, 
2023-01-06 06:52:41,425 INFO [train.py:862] Epoch 30, batch 12500, loss[loss=0.2133, simple_loss=0.4633, pruned_loss=0.1103, ctc_loss=0.1581, over 10246.00 frames. ], tot_loss[loss=0.1752, simple_loss=0.4287, pruned_loss=0.08323, ctc_loss=0.1228, over 2874022.27 frames. ], batch size: 103, lr: 3.00e-03, 
2023-01-06 06:53:48,951 INFO [train.py:862] Epoch 30, batch 13000, loss[loss=0.1418, simple_loss=0.3905, pruned_loss=0.06574, ctc_loss=0.0908, over 14721.00 frames. ], tot_loss[loss=0.1737, simple_loss=0.4279, pruned_loss=0.08229, ctc_loss=0.1212, over 2867533.70 frames. ], batch size: 37, lr: 3.00e-03, 
2023-01-06 06:54:56,771 INFO [train.py:862] Epoch 30, batch 13500, loss[loss=0.2288, simple_loss=0.4725, pruned_loss=0.1174, ctc_loss=0.1753, over 14671.00 frames. ], tot_loss[loss=0.1741, simple_loss=0.4277, pruned_loss=0.08236, ctc_loss=0.1217, over 2845806.76 frames. ], batch size: 44, lr: 3.00e-03, 
2023-01-06 06:56:04,869 INFO [train.py:862] Epoch 30, batch 14000, loss[loss=0.151, simple_loss=0.3984, pruned_loss=0.0693, ctc_loss=0.1006, over 14795.00 frames. ], tot_loss[loss=0.175, simple_loss=0.4283, pruned_loss=0.08311, ctc_loss=0.1226, over 2865556.27 frames. ], batch size: 36, lr: 3.00e-03, 
2023-01-06 06:57:12,426 INFO [train.py:862] Epoch 30, batch 14500, loss[loss=0.1533, simple_loss=0.3878, pruned_loss=0.07214, ctc_loss=0.1049, over 14672.00 frames. ], tot_loss[loss=0.1756, simple_loss=0.4296, pruned_loss=0.08368, ctc_loss=0.1229, over 2847028.66 frames. ], batch size: 35, lr: 3.00e-03, 
2023-01-06 06:58:20,257 INFO [train.py:862] Epoch 30, batch 15000, loss[loss=0.1704, simple_loss=0.4381, pruned_loss=0.08313, ctc_loss=0.1139, over 14827.00 frames. ], tot_loss[loss=0.1739, simple_loss=0.4274, pruned_loss=0.0826, ctc_loss=0.1214, over 2855451.05 frames. ], batch size: 41, lr: 3.00e-03, 
2023-01-06 06:59:27,815 INFO [train.py:862] Epoch 30, batch 15500, loss[loss=0.1431, simple_loss=0.4113, pruned_loss=0.05553, ctc_loss=0.09255, over 14699.00 frames. ], tot_loss[loss=0.1758, simple_loss=0.4291, pruned_loss=0.08384, ctc_loss=0.1233, over 2843000.12 frames. ], batch size: 47, lr: 3.00e-03, 
2023-01-06 07:00:35,877 INFO [train.py:862] Epoch 30, batch 16000, loss[loss=0.1852, simple_loss=0.4576, pruned_loss=0.08963, ctc_loss=0.1281, over 14568.00 frames. ], tot_loss[loss=0.1743, simple_loss=0.4284, pruned_loss=0.08286, ctc_loss=0.1217, over 2847387.80 frames. ], batch size: 48, lr: 3.00e-03, 
2023-01-06 07:01:42,769 INFO [train.py:862] Epoch 30, batch 16500, loss[loss=0.1771, simple_loss=0.4329, pruned_loss=0.07956, ctc_loss=0.1261, over 14552.00 frames. ], tot_loss[loss=0.1711, simple_loss=0.4259, pruned_loss=0.0805, ctc_loss=0.1187, over 2875999.80 frames. ], batch size: 48, lr: 3.00e-03, 
2023-01-06 07:02:50,436 INFO [train.py:862] Epoch 30, batch 17000, loss[loss=0.1673, simple_loss=0.4448, pruned_loss=0.07815, ctc_loss=0.1102, over 14728.00 frames. ], tot_loss[loss=0.1736, simple_loss=0.4283, pruned_loss=0.0825, ctc_loss=0.1209, over 2870893.86 frames. ], batch size: 45, lr: 3.00e-03, 
2023-01-06 07:03:57,695 INFO [train.py:862] Epoch 30, batch 17500, loss[loss=0.1348, simple_loss=0.3585, pruned_loss=0.06096, ctc_loss=0.08962, over 14503.00 frames. ], tot_loss[loss=0.1743, simple_loss=0.4272, pruned_loss=0.08281, ctc_loss=0.1219, over 2857752.29 frames. ], batch size: 34, lr: 3.00e-03, 
2023-01-06 07:05:03,587 INFO [train.py:862] Epoch 30, batch 18000, loss[loss=0.1967, simple_loss=0.4529, pruned_loss=0.09755, ctc_loss=0.1421, over 13728.00 frames. ], tot_loss[loss=0.1728, simple_loss=0.4267, pruned_loss=0.08197, ctc_loss=0.1203, over 2864851.07 frames. ], batch size: 63, lr: 3.00e-03, 
2023-01-06 07:05:03,588 INFO [train.py:887] Computing validation loss
2023-01-06 07:05:29,726 INFO [train.py:897] Epoch 30, validation: loss=0.193, simple_loss=0.4456, pruned_loss=0.09449, ctc_loss=0.1398, over 944034.00 frames. 
2023-01-06 07:05:29,727 INFO [train.py:898] Maximum memory allocated so far is 7351MB
2023-01-06 07:05:57,611 INFO [checkpoint.py:75] Saving checkpoint to tiny_transducer_ctc/exp_2m_phone_halfdelay/epoch-30.pt
2023-01-06 07:05:58,066 INFO [train.py:1118] Done!