--- license: mit pipeline_tag: text-generation library_name: transformers language: [ 'en', 'am', 'ar', 'as', 'az', 'be', 'bg', 'bn', 'br', 'bs', 'ca', 'cs', 'cy', 'da', 'de', 'el', 'eo', 'es', 'et', 'eu', 'fa', 'ff', 'fi', 'fr', 'fy', 'ga', 'gd', 'gl', 'gn', 'gu', 'ha', 'he', 'hi', 'hr', 'ht', 'hu', 'hy', 'id', 'ig', 'is', 'it', 'ja', 'jv', 'ka', 'kk', 'km', 'kn', 'ko', 'ku', 'ky', 'la', 'lg', 'li', 'ln', 'lo', 'lt', 'lv', 'mg', 'mk', 'ml', 'mn', 'mr', 'ms', 'my', 'ne', 'nl', 'no', 'ns', 'om', 'or', 'pa', 'pl', 'ps', 'pt', 'qu', 'rm', 'ro', 'ru', 'sa', 'si', 'sc', 'sd', 'sk', 'sl', 'so', 'sq', 'sr', 'ss', 'su', 'sv', 'sw', 'ta', 'te', 'th', 'tl', 'tn', 'tr', 'ug', 'uk', 'ur', 'uz', 'vi', 'wo', 'xh', 'yi', 'yo', 'zu', ] datasets: # core - base - ontocord/fineweb-permissive-multilingual-2m - distily/c4_multilingual_1M - data-silence/sumnews - xu-song/cc100-samples - badrex/llm-emoji-dataset - fblgit/simple-math - Gusarich/math-expressions-1m - neuralwork/arxiver - christopher/rosetta-code - nampdn-ai/tiny-codes - JeanKaddour/minipile # core - instruct - NousResearch/hermes-function-calling-v1 - simplescaling/s1K-1.1 # base - instruct - mlabonne/open-perfectblend - allenai/tulu-3-sft-mixture - rombodawg/Everything_Instruct_Multilingual # base - reason - open-r1/OpenR1-Math-220k - open-thoughts/OpenThoughts-114k - cognitivecomputations/dolphin-r1 - simplescaling/s1K-1.1 tags: - chat - core - base - instruct - reason --- # tangled-alpha-0.5-core ![logo](./misc/logo.jpg) ```bash time python -B prepare_core_datasets.py ``` ``` i=0, min_len=0, max_len=1073741824, block_size=4097, chunk_size=16388000, len(dataset)=1287403, len(dataset) * block_size=5274490091 Total number of tokens in the optimized dataset '../core-data-0-0-1073741824-4097-4000' is 5274490091 ``` ```bash CUDA_VISIBLE_DEVICES=0 CUDA_LAUNCH_BLOCKING=0 PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True litgpt pretrain --config pretrain_core_model.yaml ``` ``` Seed set to 23 Time to instantiate model: 0.31 seconds. Total parameters: 201,359,872 Verifying settings ... Measured TFLOPs: 7072.06 Epoch 1 | iter 256 step 1 | loss train: 11.961, val: n/a | iter time: 406.23 ms (step) remaining time: 3 days, 13:55:33 Epoch 1 | iter 512 step 2 | loss train: 11.953, val: n/a | iter time: 358.84 ms (step) remaining time: 3 days, 0:49:32 Epoch 1 | iter 768 step 3 | loss train: 11.943, val: n/a | iter time: 357.16 ms (step) remaining time: 2 days, 20:38:36 Epoch 1 | iter 1024 step 4 | loss train: 11.907, val: n/a | iter time: 355.69 ms (step) remaining time: 2 days, 18:31:54 Epoch 1 | iter 1280 step 5 | loss train: 11.854, val: n/a | iter time: 358.32 ms (step) remaining time: 2 days, 17:13:13 Epoch 1 | iter 1536 step 6 | loss train: 11.789, val: n/a | iter time: 355.59 ms (step) remaining time: 2 days, 16:18:25 Epoch 1 | iter 1792 step 7 | loss train: 11.703, val: n/a | iter time: 354.88 ms (step) remaining time: 2 days, 15:37:56 Epoch 1 | iter 2048 step 8 | loss train: 11.586, val: n/a | iter time: 354.07 ms (step) remaining time: 2 days, 15:06:45 Epoch 1 | iter 2304 step 9 | loss train: 11.451, val: n/a | iter time: 352.89 ms (step) remaining time: 2 days, 14:41:54 Epoch 1 | iter 2560 step 10 | loss train: 11.347, val: n/a | iter time: 355.58 ms (step) remaining time: 2 days, 14:21:38 Epoch 1 | iter 2816 step 11 | loss train: 11.271, val: n/a | iter time: 351.01 ms (step) remaining time: 2 days, 14:04:43 Epoch 1 | iter 3072 step 12 | loss train: 11.194, val: n/a | iter time: 351.91 ms (step) remaining time: 2 days, 13:50:26 Epoch 1 | iter 3328 step 13 | loss train: 11.151, val: n/a | iter time: 353.02 ms (step) remaining time: 2 days, 13:38:04 Epoch 1 | iter 3584 step 14 | loss train: 11.097, val: n/a | iter time: 353.75 ms (step) remaining time: 2 days, 13:27:21 Epoch 1 | iter 3840 step 15 | loss train: 11.064, val: n/a | iter time: 358.31 ms (step) remaining time: 2 days, 13:17:48 Epoch 1 | iter 4096 step 16 | loss train: 11.008, val: n/a | iter time: 351.95 ms (step) remaining time: 2 days, 13:09:17 Epoch 1 | iter 4352 step 17 | loss train: 10.997, val: n/a | iter time: 352.26 ms (step) remaining time: 2 days, 13:01:35 Epoch 1 | iter 4608 step 18 | loss train: 10.951, val: n/a | iter time: 352.57 ms (step) remaining time: 2 days, 12:54:35 Epoch 1 | iter 4864 step 19 | loss train: 10.902, val: n/a | iter time: 354.73 ms (step) remaining time: 2 days, 12:48:13 Epoch 1 | iter 5120 step 20 | loss train: 10.877, val: n/a | iter time: 354.47 ms (step) remaining time: 2 days, 12:43:19 Epoch 1 | iter 5376 step 21 | loss train: 10.830, val: n/a | iter time: 353.78 ms (step) remaining time: 2 days, 12:37:49 Epoch 1 | iter 5632 step 22 | loss train: 10.809, val: n/a | iter time: 355.03 ms (step) remaining time: 2 days, 12:32:44 Epoch 1 | iter 5888 step 23 | loss train: 10.727, val: n/a | iter time: 351.49 ms (step) remaining time: 2 days, 12:27:56 Epoch 1 | iter 6144 step 24 | loss train: 10.707, val: n/a | iter time: 351.58 ms (step) remaining time: 2 days, 12:23:24 Epoch 1 | iter 6400 step 25 | loss train: 10.643, val: n/a | iter time: 350.84 ms (step) remaining time: 2 days, 12:19:10 Epoch 1 | iter 6656 step 26 | loss train: 10.649, val: n/a | iter time: 355.14 ms (step) remaining time: 2 days, 12:15:07 Epoch 1 | iter 6912 step 27 | loss train: 10.580, val: n/a | iter time: 352.60 ms (step) remaining time: 2 days, 12:11:12 Epoch 1 | iter 7168 step 28 | loss train: 10.554, val: n/a | iter time: 351.57 ms (step) remaining time: 2 days, 12:07:27 Epoch 1 | iter 7424 step 29 | loss train: 10.526, val: n/a | iter time: 350.36 ms (step) remaining time: 2 days, 12:03:55 Epoch 1 | iter 7680 step 30 | loss train: 10.496, val: n/a | iter time: 353.19 ms (step) remaining time: 2 days, 12:00:34 Epoch 1 | iter 7936 step 31 | loss train: 10.496, val: n/a | iter time: 350.95 ms (step) remaining time: 2 days, 11:57:21 Epoch 1 | iter 8192 step 32 | loss train: 10.421, val: n/a | iter time: 352.71 ms (step) remaining time: 2 days, 11:54:18 Epoch 1 | iter 8448 step 33 | loss train: 10.379, val: n/a | iter time: 354.15 ms (step) remaining time: 2 days, 11:51:21 Epoch 1 | iter 8704 step 34 | loss train: 10.343, val: n/a | iter time: 353.95 ms (step) remaining time: 2 days, 11:48:29 Epoch 1 | iter 8960 step 35 | loss train: 10.353, val: n/a | iter time: 351.04 ms (step) remaining time: 2 days, 11:45:44 Epoch 1 | iter 9216 step 36 | loss train: 10.323, val: n/a | iter time: 354.76 ms (step) remaining time: 2 days, 11:43:05 Epoch 1 | iter 9472 step 37 | loss train: 10.258, val: n/a | iter time: 353.18 ms (step) remaining time: 2 days, 11:40:29 Epoch 1 | iter 9728 step 38 | loss train: 10.260, val: n/a | iter time: 353.86 ms (step) remaining time: 2 days, 11:37:57 Epoch 1 | iter 9984 step 39 | loss train: 10.257, val: n/a | iter time: 356.14 ms (step) remaining time: 2 days, 11:35:50 Epoch 1 | iter 10240 step 40 | loss train: 10.179, val: n/a | iter time: 353.73 ms (step) remaining time: 2 days, 11:33:23 Epoch 1 | iter 10496 step 41 | loss train: 10.163, val: n/a | iter time: 350.49 ms (step) remaining time: 2 days, 11:30:59 Epoch 1 | iter 10752 step 42 | loss train: 10.156, val: n/a | iter time: 354.15 ms (step) remaining time: 2 days, 11:28:40 Epoch 1 | iter 11008 step 43 | loss train: 10.150, val: n/a | iter time: 350.99 ms (step) remaining time: 2 days, 11:26:24 Epoch 1 | iter 11264 step 44 | loss train: 10.089, val: n/a | iter time: 354.28 ms (step) remaining time: 2 days, 11:24:09 Epoch 1 | iter 11520 step 45 | loss train: 10.096, val: n/a | iter time: 352.46 ms (step) remaining time: 2 days, 11:21:56 Epoch 1 | iter 11776 step 46 | loss train: 10.021, val: n/a | iter time: 356.80 ms (step) remaining time: 2 days, 11:19:45 Epoch 1 | iter 12032 step 47 | loss train: 10.002, val: n/a | iter time: 355.30 ms (step) remaining time: 2 days, 11:17:36 Epoch 1 | iter 12288 step 48 | loss train: 10.021, val: n/a | iter time: 355.12 ms (step) remaining time: 2 days, 11:15:32 Epoch 1 | iter 12544 step 49 | loss train: 10.017, val: n/a | iter time: 353.81 ms (step) remaining time: 2 days, 11:13:29 Epoch 1 | iter 12800 step 50 | loss train: 9.966, val: n/a | iter time: 354.70 ms (step) remaining time: 2 days, 11:11:26 # ... Epoch 1 | iter 640256 step 2501 | loss train: 3.419, val: 3.366 | iter time: 351.28 ms (step) remaining time: 0:20:06 Epoch 1 | iter 640512 step 2502 | loss train: 3.425, val: 3.366 | iter time: 351.02 ms (step) remaining time: 0:18:40 Epoch 1 | iter 640768 step 2503 | loss train: 3.396, val: 3.366 | iter time: 351.61 ms (step) remaining time: 0:17:14 Epoch 1 | iter 641024 step 2504 | loss train: 3.466, val: 3.366 | iter time: 351.42 ms (step) remaining time: 0:15:48 Epoch 1 | iter 641280 step 2505 | loss train: 3.426, val: 3.366 | iter time: 351.72 ms (step) remaining time: 0:14:23 Epoch 1 | iter 641536 step 2506 | loss train: 3.410, val: 3.366 | iter time: 351.04 ms (step) remaining time: 0:12:57 Epoch 1 | iter 641792 step 2507 | loss train: 3.523, val: 3.366 | iter time: 352.67 ms (step) remaining time: 0:11:31 Epoch 1 | iter 642048 step 2508 | loss train: 3.518, val: 3.366 | iter time: 352.04 ms (step) remaining time: 0:10:06 Epoch 1 | iter 642304 step 2509 | loss train: 3.533, val: 3.366 | iter time: 350.88 ms (step) remaining time: 0:08:40 Epoch 1 | iter 642560 step 2510 | loss train: 3.541, val: 3.366 | iter time: 351.22 ms (step) remaining time: 0:07:14 Epoch 1 | iter 642816 step 2511 | loss train: 3.564, val: 3.366 | iter time: 352.00 ms (step) remaining time: 0:05:48 Epoch 1 | iter 643072 step 2512 | loss train: 3.462, val: 3.366 | iter time: 351.88 ms (step) remaining time: 0:04:23 Epoch 1 | iter 643328 step 2513 | loss train: 3.530, val: 3.366 | iter time: 351.49 ms (step) remaining time: 0:02:57 Epoch 1 | iter 643584 step 2514 | loss train: 3.484, val: 3.366 | iter time: 351.11 ms (step) remaining time: 0:01:31 Epoch 2 | iter 643840 step 2515 | loss train: 3.375, val: 3.366 | iter time: 352.07 ms (step) remaining time: 0:00:06 Validating ... Final evaluation | val loss: 3.366 | val ppl: 28.963 Saving checkpoint to '../out/pretrain-core/final/lit_model.pth' ---------------------------------------- | Performance | - Total tokens : 5,274,484,736 | - Training Time : 215640.67 s | - Tok/sec : 16453.94 tok/s | ---------------------------------------- | Memory Usage | - Memory Used : 20.44 GB ---------------------------------------- ``` Backup `wandb`: ```bash mv wandb wandb-pretrain-core ``` Chat with model: ```bash CUDA_VISIBLE_DEVICES=0 CUDA_LAUNCH_BLOCKING=0 PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True litgpt chat ../out/pretrain-core/final ``` ```bash CUDA_VISIBLE_DEVICES=0 CUDA_LAUNCH_BLOCKING=0 PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True time litgpt evaluate --tasks 'leaderboard' --out_dir '../evaluate/pretrain-core-0/leaderboard/' --batch_size 1 --dtype 'bfloat16' '../out/pretrain-core/final' ``` ``` Tasks |Version|Filter|n-shot| Metric | |Value | |Stderr| |-----------------------------------------------------------|-------|------|-----:|-----------------------|---|-----:|---|------| |leaderboard | N/A| | | | | | | | | - leaderboard_bbh | N/A| | | | | | | | | - leaderboard_bbh_boolean_expressions | 1|none | 3|acc_norm |↑ |0.5640|± |0.0314| | - leaderboard_bbh_causal_judgement | 1|none | 3|acc_norm |↑ |0.5187|± |0.0366| | - leaderboard_bbh_date_understanding | 1|none | 3|acc_norm |↑ |0.2000|± |0.0253| | - leaderboard_bbh_disambiguation_qa | 1|none | 3|acc_norm |↑ |0.2960|± |0.0289| | - leaderboard_bbh_formal_fallacies | 1|none | 3|acc_norm |↑ |0.4680|± |0.0316| | - leaderboard_bbh_geometric_shapes | 1|none | 3|acc_norm |↑ |0.0880|± |0.0180| | - leaderboard_bbh_hyperbaton | 1|none | 3|acc_norm |↑ |0.5160|± |0.0317| | - leaderboard_bbh_logical_deduction_five_objects | 1|none | 3|acc_norm |↑ |0.1920|± |0.0250| | - leaderboard_bbh_logical_deduction_seven_objects | 1|none | 3|acc_norm |↑ |0.1320|± |0.0215| | - leaderboard_bbh_logical_deduction_three_objects | 1|none | 3|acc_norm |↑ |0.3360|± |0.0299| | - leaderboard_bbh_movie_recommendation | 1|none | 3|acc_norm |↑ |0.2520|± |0.0275| | - leaderboard_bbh_navigate | 1|none | 3|acc_norm |↑ |0.5520|± |0.0315| | - leaderboard_bbh_object_counting | 1|none | 3|acc_norm |↑ |0.0760|± |0.0168| | - leaderboard_bbh_penguins_in_a_table | 1|none | 3|acc_norm |↑ |0.1918|± |0.0327| | - leaderboard_bbh_reasoning_about_colored_objects | 1|none | 3|acc_norm |↑ |0.0680|± |0.0160| | - leaderboard_bbh_ruin_names | 1|none | 3|acc_norm |↑ |0.2080|± |0.0257| | - leaderboard_bbh_salient_translation_error_detection | 1|none | 3|acc_norm |↑ |0.1880|± |0.0248| | - leaderboard_bbh_snarks | 1|none | 3|acc_norm |↑ |0.4607|± |0.0375| | - leaderboard_bbh_sports_understanding | 1|none | 3|acc_norm |↑ |0.4600|± |0.0316| | - leaderboard_bbh_temporal_sequences | 1|none | 3|acc_norm |↑ |0.2720|± |0.0282| | - leaderboard_bbh_tracking_shuffled_objects_five_objects | 1|none | 3|acc_norm |↑ |0.2080|± |0.0257| | - leaderboard_bbh_tracking_shuffled_objects_seven_objects| 1|none | 3|acc_norm |↑ |0.1520|± |0.0228| | - leaderboard_bbh_tracking_shuffled_objects_three_objects| 1|none | 3|acc_norm |↑ |0.3320|± |0.0298| | - leaderboard_bbh_web_of_lies | 1|none | 3|acc_norm |↑ |0.4880|± |0.0317| | - leaderboard_gpqa | N/A| | | | | | | | | - leaderboard_gpqa_diamond | 1|none | 0|acc_norm |↑ |0.2020|± |0.0286| | - leaderboard_gpqa_extended | 1|none | 0|acc_norm |↑ |0.2656|± |0.0189| | - leaderboard_gpqa_main | 1|none | 0|acc_norm |↑ |0.2567|± |0.0207| | - leaderboard_ifeval | 3|none | 0|inst_level_loose_acc |↑ |0.2350|± | N/A| | | |none | 0|inst_level_strict_acc |↑ |0.2242|± | N/A| | | |none | 0|prompt_level_loose_acc |↑ |0.1109|± |0.0135| | | |none | 0|prompt_level_strict_acc|↑ |0.1054|± |0.0132| | - leaderboard_math_hard | N/A| | | | | | | | | - leaderboard_math_algebra_hard | 2|none | 4|exact_match |↑ |0.0000|± | 0| | - leaderboard_math_counting_and_prob_hard | 2|none | 4|exact_match |↑ |0.0000|± | 0| | - leaderboard_math_geometry_hard | 2|none | 4|exact_match |↑ |0.0000|± | 0| | - leaderboard_math_intermediate_algebra_hard | 2|none | 4|exact_match |↑ |0.0000|± | 0| | - leaderboard_math_num_theory_hard | 2|none | 4|exact_match |↑ |0.0019|± |0.0019| | - leaderboard_math_prealgebra_hard | 2|none | 4|exact_match |↑ |0.0011|± |0.0011| | - leaderboard_math_precalculus_hard | 2|none | 4|exact_match |↑ |0.0000|± | 0| | - leaderboard_mmlu_pro | 0.1|none | 5|acc |↑ |0.1177|± |0.0029| | - leaderboard_musr | N/A| | | | | | | | | - leaderboard_musr_murder_mysteries | 1|none | 0|acc_norm |↑ |0.4880|± |0.0317| | - leaderboard_musr_object_placements | 1|none | 0|acc_norm |↑ |0.2266|± |0.0262| | - leaderboard_musr_team_allocation | 1|none | 0|acc_norm |↑ |0.2560|± |0.0277| ```