Push model using huggingface_hub.
Browse files- config.json +70 -0
- model.safetensors +3 -0
config.json
ADDED
|
@@ -0,0 +1,70 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"seed": 1,
|
| 3 |
+
"precision": "float16",
|
| 4 |
+
"num_workers": 8,
|
| 5 |
+
"resume": false,
|
| 6 |
+
"tb_write_every_n_steps": 100,
|
| 7 |
+
"print_every_n_steps": 400,
|
| 8 |
+
"val_every_n_steps": 1600,
|
| 9 |
+
"lr": 1e-05,
|
| 10 |
+
"batch_size": 100,
|
| 11 |
+
"weight_decay": 0.0,
|
| 12 |
+
"warmup_fraction": 0.1,
|
| 13 |
+
"num_epochs": 10,
|
| 14 |
+
"num_steps": 500000,
|
| 15 |
+
"gradient_accumulation_steps": 24,
|
| 16 |
+
"gradient_clip_val": 1.0,
|
| 17 |
+
"early_stop_step": 3200,
|
| 18 |
+
"early_stop_threshold": -1.0,
|
| 19 |
+
"exp_dir": "/data/scratch/pyp/exp_pyp/VoiceCraft/gigaspeech/tts_enhanced_330M",
|
| 20 |
+
"dataset": "gigaspeech",
|
| 21 |
+
"dataset_dir": "/data/scratch/pyp/datasets/gigaspeech_phn_enc_manifest/xl",
|
| 22 |
+
"pseudo_epoch_size": 3000,
|
| 23 |
+
"phn_folder_name": "phonemes",
|
| 24 |
+
"encodec_folder_name": "encodec_16khz_4codebooks",
|
| 25 |
+
"manifest_name": "manifest_large16khz_lessambi",
|
| 26 |
+
"pad_x": 0,
|
| 27 |
+
"max_num_tokens": 20000,
|
| 28 |
+
"val_max_num_tokens": 6000,
|
| 29 |
+
"num_buckets": 10,
|
| 30 |
+
"dynamic_batching": 1,
|
| 31 |
+
"audio_max_length": 16.0,
|
| 32 |
+
"audio_min_length": 1.0,
|
| 33 |
+
"text_max_length": 400,
|
| 34 |
+
"text_min_length": 10.0,
|
| 35 |
+
"encodec_sr": 50,
|
| 36 |
+
"mask_len_min": 1,
|
| 37 |
+
"mask_len_max": 600,
|
| 38 |
+
"drop_long": 1,
|
| 39 |
+
"eos": 2051,
|
| 40 |
+
"reduced_eog": 1,
|
| 41 |
+
"special_first": 0,
|
| 42 |
+
"n_special": 4,
|
| 43 |
+
"codebook_weight": "[2,1,1,1]",
|
| 44 |
+
"empty_token": 2048,
|
| 45 |
+
"optimizer_name": "AdamW",
|
| 46 |
+
"reduce_lr_start_step": 3000,
|
| 47 |
+
"reduce_lr_start_epoch": 4,
|
| 48 |
+
"clipping_update_period": 1000,
|
| 49 |
+
"max_mask_portion": 0.9,
|
| 50 |
+
"max_n_spans": 3,
|
| 51 |
+
"shuffle_mask_embedding": 0,
|
| 52 |
+
"mask_sample_dist": "poisson1",
|
| 53 |
+
"min_gap": 5,
|
| 54 |
+
"n_codebooks": 4,
|
| 55 |
+
"text_vocab_size": 120,
|
| 56 |
+
"text_pad_token": 120,
|
| 57 |
+
"audio_vocab_size": 2048,
|
| 58 |
+
"eog": 2049,
|
| 59 |
+
"audio_pad_token": 2050,
|
| 60 |
+
"d_model": 1024,
|
| 61 |
+
"audio_embedding_dim": 1024,
|
| 62 |
+
"text_embedding_dropout": 0.0,
|
| 63 |
+
"audio_embedding_dropout": 0.0,
|
| 64 |
+
"text_positional_embedding_dropout": 0.0,
|
| 65 |
+
"audio_positional_embedding_dropout": 0.0,
|
| 66 |
+
"trm_dropout": 0.0,
|
| 67 |
+
"nhead": 16,
|
| 68 |
+
"num_decoder_layers": 24,
|
| 69 |
+
"load_model_from": "./pretrained_models/giga330M.pth"
|
| 70 |
+
}
|
model.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:aec2638d007ccb427de8b7778d9c35d3b1bf41b3aa86dd62c0bd5a4d182cc837
|
| 3 |
+
size 1293853416
|