diff --git a/.gitattributes b/.gitattributes index a6344aac8c09253b3b630fb776ae94478aa0275b..0acc4e21ca613ad39c2bdf6a691fabd2ff7d0a68 100644 --- a/.gitattributes +++ b/.gitattributes @@ -33,3 +33,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text *.zip filter=lfs diff=lfs merge=lfs -text *.zst filter=lfs diff=lfs merge=lfs -text *tfevents* filter=lfs diff=lfs merge=lfs -text +childes/EN/adult/25/00/orthographic_lstm/training.logs filter=lfs diff=lfs merge=lfs -text diff --git a/childes/EN/adult/05/00/orthographic_gpt2/config.json b/childes/EN/adult/05/00/orthographic_gpt2/config.json new file mode 100644 index 0000000000000000000000000000000000000000..493011df283c5726a4be3aa8f39c84d43991636e --- /dev/null +++ b/childes/EN/adult/05/00/orthographic_gpt2/config.json @@ -0,0 +1,32 @@ +{ + "activation_function": "gelu_new", + "architectures": [ + "GPT2LMHeadModel" + ], + "attn_pdrop": 0.1, + "bos_token_id": 50256, + "embd_pdrop": 0.1, + "eos_token_id": 50256, + "initializer_range": 0.02, + "layer_norm_epsilon": 1e-05, + "model_type": "gpt2", + "n_ctx": 1024, + "n_embd": 768, + "n_head": 12, + "n_inner": null, + "n_layer": 12, + "n_positions": 1024, + "reorder_and_upcast_attn": false, + "resid_pdrop": 0.1, + "scale_attn_by_inverse_layer_idx": false, + "scale_attn_weights": true, + "summary_activation": null, + "summary_first_dropout": 0.1, + "summary_proj_to_labels": true, + "summary_type": "cls_index", + "summary_use_proj": true, + "torch_dtype": "float32", + "transformers_version": "4.49.0", + "use_cache": true, + "vocab_size": 115 +} diff --git a/childes/EN/adult/05/00/orthographic_gpt2/generation_config.json b/childes/EN/adult/05/00/orthographic_gpt2/generation_config.json new file mode 100644 index 0000000000000000000000000000000000000000..bd968f46ce39e754a4603c69f53a2a4a08a13dbb --- /dev/null +++ b/childes/EN/adult/05/00/orthographic_gpt2/generation_config.json @@ -0,0 +1,6 @@ +{ + "_from_model_config": true, + "bos_token_id": 50256, + "eos_token_id": 50256, + "transformers_version": "4.49.0" +} diff --git a/childes/EN/adult/05/00/orthographic_gpt2/model.safetensors b/childes/EN/adult/05/00/orthographic_gpt2/model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..651285d612240ccdd79e91d00903df01104898e7 --- /dev/null +++ b/childes/EN/adult/05/00/orthographic_gpt2/model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ccf3c666e95f031a3d4a80c072bb12095b2dfeb75d7a2abbbcedd1915f89d89b +size 343737984 diff --git a/childes/EN/adult/05/00/orthographic_gpt2/training.logs b/childes/EN/adult/05/00/orthographic_gpt2/training.logs new file mode 100644 index 0000000000000000000000000000000000000000..00587aa2d000b4f724c12c152c9ba965d8825992 --- /dev/null +++ b/childes/EN/adult/05/00/orthographic_gpt2/training.logs @@ -0,0 +1,38 @@ +06/19/2025 01:08:49 - lexical_benchmark.train.run_train - INFO - Loading gpt2 trainer class. +06/19/2025 01:09:00 - datasets - INFO - PyTorch version 2.5.1 available. +06/19/2025 01:09:00 - datasets - INFO - Polars version 1.24.0 available. +06/19/2025 01:09:01 - lexical_benchmark.train.train_params - INFO - Loading params from /lustre/fsn1/projects/rech/hhb/uye44va/workspace/lm_benchmark/dev/code/src/lexical_benchmark/train/train-params.toml +06/19/2025 01:09:01 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading char tokenizer +06/19/2025 01:09:01 - lexical_benchmark.train.trainers.gpt2 - INFO - Tokenizer loaded with vocabulary size: 115 +06/19/2025 01:09:01 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading and processing datasets with joined utterances (max_length=1024) +06/19/2025 01:09:01 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/childes/by_size/EN/adult/05/00/train.tokenized +06/19/2025 01:09:01 - lexical_benchmark.train.tokenizers - INFO - Read 561041 lines from file +06/19/2025 01:09:21 - lexical_benchmark.train.tokenizers - INFO - Created 11852 joined examples from 561041 original lines +06/19/2025 01:09:23 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/childes/by_size/EN/adult/dev/dev.tokenized +06/19/2025 01:09:23 - lexical_benchmark.train.tokenizers - INFO - Read 316899 lines from file +06/19/2025 01:09:34 - lexical_benchmark.train.tokenizers - INFO - Created 7325 joined examples from 316899 original lines +06/19/2025 01:09:35 - lexical_benchmark.train.trainers.gpt2 - INFO - Training dataset size: 11852 +06/19/2025 01:09:35 - lexical_benchmark.train.trainers.gpt2 - INFO - Validation dataset size: 7325 +06/19/2025 01:09:35 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating GPT2 configuration +06/19/2025 01:09:36 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating standard Trainer +06/19/2025 01:09:37 - lexical_benchmark.train.run_train - INFO - Running training of childes_gpt2_EN_05_00 +06/19/2025 01:12:19 - lexical_benchmark.train.run_train - INFO - Loading gpt2 trainer class. +06/19/2025 01:12:30 - datasets - INFO - PyTorch version 2.5.1 available. +06/19/2025 01:12:30 - datasets - INFO - Polars version 1.24.0 available. +06/19/2025 01:12:31 - lexical_benchmark.train.train_params - INFO - Loading params from /lustre/fsn1/projects/rech/hhb/uye44va/workspace/lm_benchmark/dev/code/src/lexical_benchmark/train/train-params.toml +06/19/2025 01:12:31 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading char tokenizer +06/19/2025 01:12:31 - lexical_benchmark.train.trainers.gpt2 - INFO - Tokenizer loaded with vocabulary size: 115 +06/19/2025 01:12:31 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading and processing datasets with joined utterances (max_length=1024) +06/19/2025 01:12:31 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/childes/by_size/EN/adult/05/00/train.tokenized +06/19/2025 01:12:31 - lexical_benchmark.train.tokenizers - INFO - Read 561041 lines from file +06/19/2025 01:12:51 - lexical_benchmark.train.tokenizers - INFO - Created 11852 joined examples from 561041 original lines +06/19/2025 01:12:53 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/childes/by_size/EN/adult/dev/dev.tokenized +06/19/2025 01:12:53 - lexical_benchmark.train.tokenizers - INFO - Read 316899 lines from file +06/19/2025 01:13:05 - lexical_benchmark.train.tokenizers - INFO - Created 7325 joined examples from 316899 original lines +06/19/2025 01:13:06 - lexical_benchmark.train.trainers.gpt2 - INFO - Training dataset size: 11852 +06/19/2025 01:13:06 - lexical_benchmark.train.trainers.gpt2 - INFO - Validation dataset size: 7325 +06/19/2025 01:13:06 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating GPT2 configuration +06/19/2025 01:13:07 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating standard Trainer +06/19/2025 01:13:07 - lexical_benchmark.train.run_train - INFO - Running training of childes_gpt2_EN_05_00 +06/19/2025 02:46:25 - lexical_benchmark.train.run_train - INFO - Training of childes_gpt2_EN_05_00 has completed ! +06/19/2025 02:46:25 - lexical_benchmark.train.run_train - INFO - Model saved to /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/models/childes/EN/adult/05/00/gpt2 diff --git a/childes/EN/adult/05/00/orthographic_gpt2/training_args.bin b/childes/EN/adult/05/00/orthographic_gpt2/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..53e97a565919d161948e162a0516a8bd0d1d8b2d --- /dev/null +++ b/childes/EN/adult/05/00/orthographic_gpt2/training_args.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:fab672c263a3536143e829f27276021c3a3b97db7119200922890f4e99af56c7 +size 5496 diff --git a/childes/EN/adult/05/00/orthographic_lstm/config.json b/childes/EN/adult/05/00/orthographic_lstm/config.json new file mode 100644 index 0000000000000000000000000000000000000000..5ef629887c7fc87d64278c9788e493ac34c60370 --- /dev/null +++ b/childes/EN/adult/05/00/orthographic_lstm/config.json @@ -0,0 +1,13 @@ +{ + "architectures": [ + "LSTMForLanguageModeling" + ], + "dropout": 0.1, + "embedding_dim": 200, + "hidden_size": 1024, + "model_type": "lstm", + "num_layers": 3, + "torch_dtype": "float32", + "transformers_version": "4.49.0", + "vocab_size": 115 +} diff --git a/childes/EN/adult/05/00/orthographic_lstm/generation_config.json b/childes/EN/adult/05/00/orthographic_lstm/generation_config.json new file mode 100644 index 0000000000000000000000000000000000000000..c1818a95e07bdccb83305ce8f9d878755e3c6119 --- /dev/null +++ b/childes/EN/adult/05/00/orthographic_lstm/generation_config.json @@ -0,0 +1,4 @@ +{ + "_from_model_config": true, + "transformers_version": "4.49.0" +} diff --git a/childes/EN/adult/05/00/orthographic_lstm/model.safetensors b/childes/EN/adult/05/00/orthographic_lstm/model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..09642cfae331445966ad87201c504054d1365fd1 --- /dev/null +++ b/childes/EN/adult/05/00/orthographic_lstm/model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:986070fc467d6b2b4caef72683b4f0a9e41cdcfc5fe2d6eba7768394613778e2 +size 87825996 diff --git a/childes/EN/adult/05/00/orthographic_lstm/training.logs b/childes/EN/adult/05/00/orthographic_lstm/training.logs new file mode 100644 index 0000000000000000000000000000000000000000..29c278f20df5a06f01c2705bde4c7c437cc25348 --- /dev/null +++ b/childes/EN/adult/05/00/orthographic_lstm/training.logs @@ -0,0 +1,53 @@ +06/19/2025 00:46:24 - lexical_benchmark.train.run_train - INFO - Loading lstm trainer class. +06/19/2025 00:46:28 - datasets - INFO - PyTorch version 2.5.1 available. +06/19/2025 00:46:28 - datasets - INFO - Polars version 1.24.0 available. +06/19/2025 00:46:29 - lexical_benchmark.train.train_params - INFO - Loading params from /lustre/fsn1/projects/rech/hhb/uye44va/workspace/lm_benchmark/dev/code/src/lexical_benchmark/train/train-params.toml +06/19/2025 00:46:29 - lexical_benchmark.train.trainers.lstm - INFO - Loading char-tokenizer +06/19/2025 00:46:29 - lexical_benchmark.train.trainers.lstm - INFO - Tokenizer loaded with vocabulary size: 115 +06/19/2025 00:46:29 - lexical_benchmark.train.trainers.lstm - INFO - Loading and processing datasets with joined utterances (max_length=1024) +06/19/2025 00:46:29 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/childes/by_size/EN/adult/05/00/train.tokenized +06/19/2025 00:46:29 - lexical_benchmark.train.tokenizers - INFO - Read 561041 lines from file +06/19/2025 00:46:49 - lexical_benchmark.train.tokenizers - INFO - Created 11852 joined examples from 561041 original lines +06/19/2025 00:46:51 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/childes/by_size/EN/adult/dev/dev.tokenized +06/19/2025 00:46:51 - lexical_benchmark.train.tokenizers - INFO - Read 316899 lines from file +06/19/2025 00:47:02 - lexical_benchmark.train.tokenizers - INFO - Created 7325 joined examples from 316899 original lines +06/19/2025 00:47:03 - lexical_benchmark.train.trainers.lstm - INFO - Training dataset size: 11852 +06/19/2025 00:47:03 - lexical_benchmark.train.trainers.lstm - INFO - Validation dataset size: 7325 +06/19/2025 00:47:03 - lexical_benchmark.train.trainers.lstm - INFO - Loading configurations & initialising LSTM model trainer +06/19/2025 00:47:04 - lexical_benchmark.train.run_train - INFO - Running training of childes_lstm_EN_05_00 +06/19/2025 01:08:44 - lexical_benchmark.train.run_train - INFO - Loading lstm trainer class. +06/19/2025 01:08:48 - datasets - INFO - PyTorch version 2.5.1 available. +06/19/2025 01:08:48 - datasets - INFO - Polars version 1.24.0 available. +06/19/2025 01:08:48 - lexical_benchmark.train.train_params - INFO - Loading params from /lustre/fsn1/projects/rech/hhb/uye44va/workspace/lm_benchmark/dev/code/src/lexical_benchmark/train/train-params.toml +06/19/2025 01:08:48 - lexical_benchmark.train.trainers.lstm - INFO - Loading char-tokenizer +06/19/2025 01:08:48 - lexical_benchmark.train.trainers.lstm - INFO - Tokenizer loaded with vocabulary size: 115 +06/19/2025 01:08:48 - lexical_benchmark.train.trainers.lstm - INFO - Loading and processing datasets with joined utterances (max_length=1024) +06/19/2025 01:08:48 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/childes/by_size/EN/adult/05/00/train.tokenized +06/19/2025 01:08:48 - lexical_benchmark.train.tokenizers - INFO - Read 561041 lines from file +06/19/2025 01:09:09 - lexical_benchmark.train.tokenizers - INFO - Created 11852 joined examples from 561041 original lines +06/19/2025 01:09:10 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/childes/by_size/EN/adult/dev/dev.tokenized +06/19/2025 01:09:10 - lexical_benchmark.train.tokenizers - INFO - Read 316899 lines from file +06/19/2025 01:09:22 - lexical_benchmark.train.tokenizers - INFO - Created 7325 joined examples from 316899 original lines +06/19/2025 01:09:23 - lexical_benchmark.train.trainers.lstm - INFO - Training dataset size: 11852 +06/19/2025 01:09:23 - lexical_benchmark.train.trainers.lstm - INFO - Validation dataset size: 7325 +06/19/2025 01:09:23 - lexical_benchmark.train.trainers.lstm - INFO - Loading configurations & initialising LSTM model trainer +06/19/2025 01:09:24 - lexical_benchmark.train.run_train - INFO - Running training of childes_lstm_EN_05_00 +06/19/2025 01:12:15 - lexical_benchmark.train.run_train - INFO - Loading lstm trainer class. +06/19/2025 01:12:20 - datasets - INFO - PyTorch version 2.5.1 available. +06/19/2025 01:12:20 - datasets - INFO - Polars version 1.24.0 available. +06/19/2025 01:12:21 - lexical_benchmark.train.train_params - INFO - Loading params from /lustre/fsn1/projects/rech/hhb/uye44va/workspace/lm_benchmark/dev/code/src/lexical_benchmark/train/train-params.toml +06/19/2025 01:12:21 - lexical_benchmark.train.trainers.lstm - INFO - Loading char-tokenizer +06/19/2025 01:12:21 - lexical_benchmark.train.trainers.lstm - INFO - Tokenizer loaded with vocabulary size: 115 +06/19/2025 01:12:21 - lexical_benchmark.train.trainers.lstm - INFO - Loading and processing datasets with joined utterances (max_length=1024) +06/19/2025 01:12:21 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/childes/by_size/EN/adult/05/00/train.tokenized +06/19/2025 01:12:21 - lexical_benchmark.train.tokenizers - INFO - Read 561041 lines from file +06/19/2025 01:12:40 - lexical_benchmark.train.tokenizers - INFO - Created 11852 joined examples from 561041 original lines +06/19/2025 01:12:42 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/childes/by_size/EN/adult/dev/dev.tokenized +06/19/2025 01:12:42 - lexical_benchmark.train.tokenizers - INFO - Read 316899 lines from file +06/19/2025 01:12:54 - lexical_benchmark.train.tokenizers - INFO - Created 7325 joined examples from 316899 original lines +06/19/2025 01:12:55 - lexical_benchmark.train.trainers.lstm - INFO - Training dataset size: 11852 +06/19/2025 01:12:55 - lexical_benchmark.train.trainers.lstm - INFO - Validation dataset size: 7325 +06/19/2025 01:12:55 - lexical_benchmark.train.trainers.lstm - INFO - Loading configurations & initialising LSTM model trainer +06/19/2025 01:12:56 - lexical_benchmark.train.run_train - INFO - Running training of childes_lstm_EN_05_00 +06/19/2025 07:24:19 - lexical_benchmark.train.run_train - INFO - Training of childes_lstm_EN_05_00 has completed ! +06/19/2025 07:24:19 - lexical_benchmark.train.run_train - INFO - Model saved to /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/models/childes/EN/adult/05/00/lstm diff --git a/childes/EN/adult/05/00/orthographic_lstm/training_args.bin b/childes/EN/adult/05/00/orthographic_lstm/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..ffa3ebcb58caf14ebcf0d75953faf3a7852d94b3 --- /dev/null +++ b/childes/EN/adult/05/00/orthographic_lstm/training_args.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:988f15f68d9ca2b3ba55756b7cbc7d948011e56f06eab4f5a5758fdd9dc7241e +size 5496 diff --git a/childes/EN/adult/15/00/orthographic_gpt2/config.json b/childes/EN/adult/15/00/orthographic_gpt2/config.json new file mode 100644 index 0000000000000000000000000000000000000000..493011df283c5726a4be3aa8f39c84d43991636e --- /dev/null +++ b/childes/EN/adult/15/00/orthographic_gpt2/config.json @@ -0,0 +1,32 @@ +{ + "activation_function": "gelu_new", + "architectures": [ + "GPT2LMHeadModel" + ], + "attn_pdrop": 0.1, + "bos_token_id": 50256, + "embd_pdrop": 0.1, + "eos_token_id": 50256, + "initializer_range": 0.02, + "layer_norm_epsilon": 1e-05, + "model_type": "gpt2", + "n_ctx": 1024, + "n_embd": 768, + "n_head": 12, + "n_inner": null, + "n_layer": 12, + "n_positions": 1024, + "reorder_and_upcast_attn": false, + "resid_pdrop": 0.1, + "scale_attn_by_inverse_layer_idx": false, + "scale_attn_weights": true, + "summary_activation": null, + "summary_first_dropout": 0.1, + "summary_proj_to_labels": true, + "summary_type": "cls_index", + "summary_use_proj": true, + "torch_dtype": "float32", + "transformers_version": "4.49.0", + "use_cache": true, + "vocab_size": 115 +} diff --git a/childes/EN/adult/15/00/orthographic_gpt2/generation_config.json b/childes/EN/adult/15/00/orthographic_gpt2/generation_config.json new file mode 100644 index 0000000000000000000000000000000000000000..bd968f46ce39e754a4603c69f53a2a4a08a13dbb --- /dev/null +++ b/childes/EN/adult/15/00/orthographic_gpt2/generation_config.json @@ -0,0 +1,6 @@ +{ + "_from_model_config": true, + "bos_token_id": 50256, + "eos_token_id": 50256, + "transformers_version": "4.49.0" +} diff --git a/childes/EN/adult/15/00/orthographic_gpt2/model.safetensors b/childes/EN/adult/15/00/orthographic_gpt2/model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..7ba716f3db9a5ab1914077c430f905e52fd55d28 --- /dev/null +++ b/childes/EN/adult/15/00/orthographic_gpt2/model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c78fa75bbb5b8c628ccc71720929b7b54094bdfd2b9c7744c620437a3e85b30a +size 343737984 diff --git a/childes/EN/adult/15/00/orthographic_gpt2/training.logs b/childes/EN/adult/15/00/orthographic_gpt2/training.logs new file mode 100644 index 0000000000000000000000000000000000000000..db7d7ae75f2c19f774d9754522bfcfc2a7416015 --- /dev/null +++ b/childes/EN/adult/15/00/orthographic_gpt2/training.logs @@ -0,0 +1,38 @@ +06/19/2025 01:08:49 - lexical_benchmark.train.run_train - INFO - Loading gpt2 trainer class. +06/19/2025 01:09:00 - datasets - INFO - PyTorch version 2.5.1 available. +06/19/2025 01:09:00 - datasets - INFO - Polars version 1.24.0 available. +06/19/2025 01:09:01 - lexical_benchmark.train.train_params - INFO - Loading params from /lustre/fsn1/projects/rech/hhb/uye44va/workspace/lm_benchmark/dev/code/src/lexical_benchmark/train/train-params.toml +06/19/2025 01:09:01 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading char tokenizer +06/19/2025 01:09:01 - lexical_benchmark.train.trainers.gpt2 - INFO - Tokenizer loaded with vocabulary size: 115 +06/19/2025 01:09:01 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading and processing datasets with joined utterances (max_length=1024) +06/19/2025 01:09:03 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/childes/by_size/EN/adult/15/00/train.tokenized +06/19/2025 01:09:03 - lexical_benchmark.train.tokenizers - INFO - Read 1633721 lines from file +06/19/2025 01:10:01 - lexical_benchmark.train.tokenizers - INFO - Created 35893 joined examples from 1633721 original lines +06/19/2025 01:10:07 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/childes/by_size/EN/adult/dev/dev.tokenized +06/19/2025 01:10:07 - lexical_benchmark.train.tokenizers - INFO - Read 316899 lines from file +06/19/2025 01:10:19 - lexical_benchmark.train.tokenizers - INFO - Created 7325 joined examples from 316899 original lines +06/19/2025 01:10:20 - lexical_benchmark.train.trainers.gpt2 - INFO - Training dataset size: 35893 +06/19/2025 01:10:20 - lexical_benchmark.train.trainers.gpt2 - INFO - Validation dataset size: 7325 +06/19/2025 01:10:20 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating GPT2 configuration +06/19/2025 01:10:20 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating standard Trainer +06/19/2025 01:10:21 - lexical_benchmark.train.run_train - INFO - Running training of childes_gpt2_EN_15_00 +06/19/2025 01:12:19 - lexical_benchmark.train.run_train - INFO - Loading gpt2 trainer class. +06/19/2025 01:12:30 - datasets - INFO - PyTorch version 2.5.1 available. +06/19/2025 01:12:30 - datasets - INFO - Polars version 1.24.0 available. +06/19/2025 01:12:31 - lexical_benchmark.train.train_params - INFO - Loading params from /lustre/fsn1/projects/rech/hhb/uye44va/workspace/lm_benchmark/dev/code/src/lexical_benchmark/train/train-params.toml +06/19/2025 01:12:31 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading char tokenizer +06/19/2025 01:12:31 - lexical_benchmark.train.trainers.gpt2 - INFO - Tokenizer loaded with vocabulary size: 115 +06/19/2025 01:12:31 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading and processing datasets with joined utterances (max_length=1024) +06/19/2025 01:12:31 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/childes/by_size/EN/adult/15/00/train.tokenized +06/19/2025 01:12:31 - lexical_benchmark.train.tokenizers - INFO - Read 1633721 lines from file +06/19/2025 01:13:30 - lexical_benchmark.train.tokenizers - INFO - Created 35893 joined examples from 1633721 original lines +06/19/2025 01:13:36 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/childes/by_size/EN/adult/dev/dev.tokenized +06/19/2025 01:13:36 - lexical_benchmark.train.tokenizers - INFO - Read 316899 lines from file +06/19/2025 01:13:47 - lexical_benchmark.train.tokenizers - INFO - Created 7325 joined examples from 316899 original lines +06/19/2025 01:13:49 - lexical_benchmark.train.trainers.gpt2 - INFO - Training dataset size: 35893 +06/19/2025 01:13:49 - lexical_benchmark.train.trainers.gpt2 - INFO - Validation dataset size: 7325 +06/19/2025 01:13:49 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating GPT2 configuration +06/19/2025 01:13:49 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating standard Trainer +06/19/2025 01:13:51 - lexical_benchmark.train.run_train - INFO - Running training of childes_gpt2_EN_15_00 +06/19/2025 03:49:45 - lexical_benchmark.train.run_train - INFO - Training of childes_gpt2_EN_15_00 has completed ! +06/19/2025 03:49:46 - lexical_benchmark.train.run_train - INFO - Model saved to /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/models/childes/EN/adult/15/00/gpt2 diff --git a/childes/EN/adult/15/00/orthographic_gpt2/training_args.bin b/childes/EN/adult/15/00/orthographic_gpt2/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..036764a8a24f4fb7b1b0bbaeb921f21ab4581a6a --- /dev/null +++ b/childes/EN/adult/15/00/orthographic_gpt2/training_args.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b001e0beba0c81b208a01ee4d6b4f4765d25b360565fe569ed7c1ae870ece854 +size 5496 diff --git a/childes/EN/adult/15/00/orthographic_lstm/config.json b/childes/EN/adult/15/00/orthographic_lstm/config.json new file mode 100644 index 0000000000000000000000000000000000000000..5ef629887c7fc87d64278c9788e493ac34c60370 --- /dev/null +++ b/childes/EN/adult/15/00/orthographic_lstm/config.json @@ -0,0 +1,13 @@ +{ + "architectures": [ + "LSTMForLanguageModeling" + ], + "dropout": 0.1, + "embedding_dim": 200, + "hidden_size": 1024, + "model_type": "lstm", + "num_layers": 3, + "torch_dtype": "float32", + "transformers_version": "4.49.0", + "vocab_size": 115 +} diff --git a/childes/EN/adult/15/00/orthographic_lstm/generation_config.json b/childes/EN/adult/15/00/orthographic_lstm/generation_config.json new file mode 100644 index 0000000000000000000000000000000000000000..c1818a95e07bdccb83305ce8f9d878755e3c6119 --- /dev/null +++ b/childes/EN/adult/15/00/orthographic_lstm/generation_config.json @@ -0,0 +1,4 @@ +{ + "_from_model_config": true, + "transformers_version": "4.49.0" +} diff --git a/childes/EN/adult/15/00/orthographic_lstm/model.safetensors b/childes/EN/adult/15/00/orthographic_lstm/model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..5e3dd69d2bee351e5689d7665b084efd0453bc19 --- /dev/null +++ b/childes/EN/adult/15/00/orthographic_lstm/model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:fd91a88444c8a4050355e4928d7c1d1d808698d1a3e252ec99c4908276239362 +size 87825996 diff --git a/childes/EN/adult/15/00/orthographic_lstm/training.logs b/childes/EN/adult/15/00/orthographic_lstm/training.logs new file mode 100644 index 0000000000000000000000000000000000000000..89b195dcb650ae365e3f48be9c80d9562a544645 --- /dev/null +++ b/childes/EN/adult/15/00/orthographic_lstm/training.logs @@ -0,0 +1,36 @@ +06/19/2025 01:08:49 - lexical_benchmark.train.run_train - INFO - Loading lstm trainer class. +06/19/2025 01:09:00 - datasets - INFO - PyTorch version 2.5.1 available. +06/19/2025 01:09:00 - datasets - INFO - Polars version 1.24.0 available. +06/19/2025 01:09:01 - lexical_benchmark.train.train_params - INFO - Loading params from /lustre/fsn1/projects/rech/hhb/uye44va/workspace/lm_benchmark/dev/code/src/lexical_benchmark/train/train-params.toml +06/19/2025 01:09:01 - lexical_benchmark.train.trainers.lstm - INFO - Loading char-tokenizer +06/19/2025 01:09:01 - lexical_benchmark.train.trainers.lstm - INFO - Tokenizer loaded with vocabulary size: 115 +06/19/2025 01:09:01 - lexical_benchmark.train.trainers.lstm - INFO - Loading and processing datasets with joined utterances (max_length=1024) +06/19/2025 01:09:03 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/childes/by_size/EN/adult/15/00/train.tokenized +06/19/2025 01:09:03 - lexical_benchmark.train.tokenizers - INFO - Read 1633721 lines from file +06/19/2025 01:10:01 - lexical_benchmark.train.tokenizers - INFO - Created 35893 joined examples from 1633721 original lines +06/19/2025 01:10:07 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/childes/by_size/EN/adult/dev/dev.tokenized +06/19/2025 01:10:07 - lexical_benchmark.train.tokenizers - INFO - Read 316899 lines from file +06/19/2025 01:10:19 - lexical_benchmark.train.tokenizers - INFO - Created 7325 joined examples from 316899 original lines +06/19/2025 01:10:20 - lexical_benchmark.train.trainers.lstm - INFO - Training dataset size: 35893 +06/19/2025 01:10:20 - lexical_benchmark.train.trainers.lstm - INFO - Validation dataset size: 7325 +06/19/2025 01:10:20 - lexical_benchmark.train.trainers.lstm - INFO - Loading configurations & initialising LSTM model trainer +06/19/2025 01:10:20 - lexical_benchmark.train.run_train - INFO - Running training of childes_lstm_EN_15_00 +06/19/2025 01:12:19 - lexical_benchmark.train.run_train - INFO - Loading lstm trainer class. +06/19/2025 01:12:30 - datasets - INFO - PyTorch version 2.5.1 available. +06/19/2025 01:12:30 - datasets - INFO - Polars version 1.24.0 available. +06/19/2025 01:12:31 - lexical_benchmark.train.train_params - INFO - Loading params from /lustre/fsn1/projects/rech/hhb/uye44va/workspace/lm_benchmark/dev/code/src/lexical_benchmark/train/train-params.toml +06/19/2025 01:12:31 - lexical_benchmark.train.trainers.lstm - INFO - Loading char-tokenizer +06/19/2025 01:12:31 - lexical_benchmark.train.trainers.lstm - INFO - Tokenizer loaded with vocabulary size: 115 +06/19/2025 01:12:31 - lexical_benchmark.train.trainers.lstm - INFO - Loading and processing datasets with joined utterances (max_length=1024) +06/19/2025 01:12:31 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/childes/by_size/EN/adult/15/00/train.tokenized +06/19/2025 01:12:31 - lexical_benchmark.train.tokenizers - INFO - Read 1633721 lines from file +06/19/2025 01:13:30 - lexical_benchmark.train.tokenizers - INFO - Created 35893 joined examples from 1633721 original lines +06/19/2025 01:13:36 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/childes/by_size/EN/adult/dev/dev.tokenized +06/19/2025 01:13:36 - lexical_benchmark.train.tokenizers - INFO - Read 316899 lines from file +06/19/2025 01:13:48 - lexical_benchmark.train.tokenizers - INFO - Created 7325 joined examples from 316899 original lines +06/19/2025 01:13:49 - lexical_benchmark.train.trainers.lstm - INFO - Training dataset size: 35893 +06/19/2025 01:13:49 - lexical_benchmark.train.trainers.lstm - INFO - Validation dataset size: 7325 +06/19/2025 01:13:49 - lexical_benchmark.train.trainers.lstm - INFO - Loading configurations & initialising LSTM model trainer +06/19/2025 01:13:51 - lexical_benchmark.train.run_train - INFO - Running training of childes_lstm_EN_15_00 +06/19/2025 12:29:35 - lexical_benchmark.train.run_train - INFO - Training of childes_lstm_EN_15_00 has completed ! +06/19/2025 12:29:35 - lexical_benchmark.train.run_train - INFO - Model saved to /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/models/childes/EN/adult/15/00/lstm diff --git a/childes/EN/adult/15/00/orthographic_lstm/training_args.bin b/childes/EN/adult/15/00/orthographic_lstm/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..643ef0347300ddefd9f7fd60b0679b35f8c8e688 --- /dev/null +++ b/childes/EN/adult/15/00/orthographic_lstm/training_args.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:317eef37ac361a92a1b321eb000eaafa336c3e81c0898947746270ae8cc5905c +size 5496 diff --git a/childes/EN/adult/20/00/orthographic_gpt2/config.json b/childes/EN/adult/20/00/orthographic_gpt2/config.json new file mode 100644 index 0000000000000000000000000000000000000000..493011df283c5726a4be3aa8f39c84d43991636e --- /dev/null +++ b/childes/EN/adult/20/00/orthographic_gpt2/config.json @@ -0,0 +1,32 @@ +{ + "activation_function": "gelu_new", + "architectures": [ + "GPT2LMHeadModel" + ], + "attn_pdrop": 0.1, + "bos_token_id": 50256, + "embd_pdrop": 0.1, + "eos_token_id": 50256, + "initializer_range": 0.02, + "layer_norm_epsilon": 1e-05, + "model_type": "gpt2", + "n_ctx": 1024, + "n_embd": 768, + "n_head": 12, + "n_inner": null, + "n_layer": 12, + "n_positions": 1024, + "reorder_and_upcast_attn": false, + "resid_pdrop": 0.1, + "scale_attn_by_inverse_layer_idx": false, + "scale_attn_weights": true, + "summary_activation": null, + "summary_first_dropout": 0.1, + "summary_proj_to_labels": true, + "summary_type": "cls_index", + "summary_use_proj": true, + "torch_dtype": "float32", + "transformers_version": "4.49.0", + "use_cache": true, + "vocab_size": 115 +} diff --git a/childes/EN/adult/20/00/orthographic_gpt2/generation_config.json b/childes/EN/adult/20/00/orthographic_gpt2/generation_config.json new file mode 100644 index 0000000000000000000000000000000000000000..bd968f46ce39e754a4603c69f53a2a4a08a13dbb --- /dev/null +++ b/childes/EN/adult/20/00/orthographic_gpt2/generation_config.json @@ -0,0 +1,6 @@ +{ + "_from_model_config": true, + "bos_token_id": 50256, + "eos_token_id": 50256, + "transformers_version": "4.49.0" +} diff --git a/childes/EN/adult/20/00/orthographic_gpt2/model.safetensors b/childes/EN/adult/20/00/orthographic_gpt2/model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..f09477898fa12242354507d1eb445da1ad3a2dca --- /dev/null +++ b/childes/EN/adult/20/00/orthographic_gpt2/model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e3e50cf3b71e8b75e0fa39345f424aa9fcbcbe85e9af12861bc31d36d3faae05 +size 343737984 diff --git a/childes/EN/adult/20/00/orthographic_gpt2/training.logs b/childes/EN/adult/20/00/orthographic_gpt2/training.logs new file mode 100644 index 0000000000000000000000000000000000000000..9c2d8b597034434bd6df4ec23754b591c0815c3c --- /dev/null +++ b/childes/EN/adult/20/00/orthographic_gpt2/training.logs @@ -0,0 +1,38 @@ +06/19/2025 01:08:48 - lexical_benchmark.train.run_train - INFO - Loading gpt2 trainer class. +06/19/2025 01:08:59 - datasets - INFO - PyTorch version 2.5.1 available. +06/19/2025 01:08:59 - datasets - INFO - Polars version 1.24.0 available. +06/19/2025 01:09:00 - lexical_benchmark.train.train_params - INFO - Loading params from /lustre/fsn1/projects/rech/hhb/uye44va/workspace/lm_benchmark/dev/code/src/lexical_benchmark/train/train-params.toml +06/19/2025 01:09:00 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading char tokenizer +06/19/2025 01:09:00 - lexical_benchmark.train.trainers.gpt2 - INFO - Tokenizer loaded with vocabulary size: 115 +06/19/2025 01:09:00 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading and processing datasets with joined utterances (max_length=1024) +06/19/2025 01:09:02 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/childes/by_size/EN/adult/20/00/train.tokenized +06/19/2025 01:09:03 - lexical_benchmark.train.tokenizers - INFO - Read 2254077 lines from file +06/19/2025 01:10:24 - lexical_benchmark.train.tokenizers - INFO - Created 47509 joined examples from 2254077 original lines +06/19/2025 01:10:31 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/childes/by_size/EN/adult/dev/dev.tokenized +06/19/2025 01:10:31 - lexical_benchmark.train.tokenizers - INFO - Read 316899 lines from file +06/19/2025 01:10:43 - lexical_benchmark.train.tokenizers - INFO - Created 7325 joined examples from 316899 original lines +06/19/2025 01:10:44 - lexical_benchmark.train.trainers.gpt2 - INFO - Training dataset size: 47509 +06/19/2025 01:10:44 - lexical_benchmark.train.trainers.gpt2 - INFO - Validation dataset size: 7325 +06/19/2025 01:10:44 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating GPT2 configuration +06/19/2025 01:10:45 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating standard Trainer +06/19/2025 01:10:46 - lexical_benchmark.train.run_train - INFO - Running training of childes_gpt2_EN_20_00 +06/19/2025 01:12:19 - lexical_benchmark.train.run_train - INFO - Loading gpt2 trainer class. +06/19/2025 01:12:29 - datasets - INFO - PyTorch version 2.5.1 available. +06/19/2025 01:12:29 - datasets - INFO - Polars version 1.24.0 available. +06/19/2025 01:12:30 - lexical_benchmark.train.train_params - INFO - Loading params from /lustre/fsn1/projects/rech/hhb/uye44va/workspace/lm_benchmark/dev/code/src/lexical_benchmark/train/train-params.toml +06/19/2025 01:12:30 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading char tokenizer +06/19/2025 01:12:30 - lexical_benchmark.train.trainers.gpt2 - INFO - Tokenizer loaded with vocabulary size: 115 +06/19/2025 01:12:30 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading and processing datasets with joined utterances (max_length=1024) +06/19/2025 01:12:30 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/childes/by_size/EN/adult/20/00/train.tokenized +06/19/2025 01:12:30 - lexical_benchmark.train.tokenizers - INFO - Read 2254077 lines from file +06/19/2025 01:13:53 - lexical_benchmark.train.tokenizers - INFO - Created 47509 joined examples from 2254077 original lines +06/19/2025 01:14:01 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/childes/by_size/EN/adult/dev/dev.tokenized +06/19/2025 01:14:01 - lexical_benchmark.train.tokenizers - INFO - Read 316899 lines from file +06/19/2025 01:14:13 - lexical_benchmark.train.tokenizers - INFO - Created 7325 joined examples from 316899 original lines +06/19/2025 01:14:14 - lexical_benchmark.train.trainers.gpt2 - INFO - Training dataset size: 47509 +06/19/2025 01:14:14 - lexical_benchmark.train.trainers.gpt2 - INFO - Validation dataset size: 7325 +06/19/2025 01:14:14 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating GPT2 configuration +06/19/2025 01:14:15 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating standard Trainer +06/19/2025 01:14:16 - lexical_benchmark.train.run_train - INFO - Running training of childes_gpt2_EN_20_00 +06/19/2025 04:20:49 - lexical_benchmark.train.run_train - INFO - Training of childes_gpt2_EN_20_00 has completed ! +06/19/2025 04:20:49 - lexical_benchmark.train.run_train - INFO - Model saved to /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/models/childes/EN/adult/20/00/gpt2 diff --git a/childes/EN/adult/20/00/orthographic_gpt2/training_args.bin b/childes/EN/adult/20/00/orthographic_gpt2/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..a21848edef6193955baf8825f1fb42a7682d5338 --- /dev/null +++ b/childes/EN/adult/20/00/orthographic_gpt2/training_args.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f747f444d9b2de50670f2fe4a1fa44e86de494e73b3110a36cbba1e060316592 +size 5496 diff --git a/childes/EN/adult/20/00/orthographic_lstm/config.json b/childes/EN/adult/20/00/orthographic_lstm/config.json new file mode 100644 index 0000000000000000000000000000000000000000..5ef629887c7fc87d64278c9788e493ac34c60370 --- /dev/null +++ b/childes/EN/adult/20/00/orthographic_lstm/config.json @@ -0,0 +1,13 @@ +{ + "architectures": [ + "LSTMForLanguageModeling" + ], + "dropout": 0.1, + "embedding_dim": 200, + "hidden_size": 1024, + "model_type": "lstm", + "num_layers": 3, + "torch_dtype": "float32", + "transformers_version": "4.49.0", + "vocab_size": 115 +} diff --git a/childes/EN/adult/20/00/orthographic_lstm/generation_config.json b/childes/EN/adult/20/00/orthographic_lstm/generation_config.json new file mode 100644 index 0000000000000000000000000000000000000000..c1818a95e07bdccb83305ce8f9d878755e3c6119 --- /dev/null +++ b/childes/EN/adult/20/00/orthographic_lstm/generation_config.json @@ -0,0 +1,4 @@ +{ + "_from_model_config": true, + "transformers_version": "4.49.0" +} diff --git a/childes/EN/adult/20/00/orthographic_lstm/model.safetensors b/childes/EN/adult/20/00/orthographic_lstm/model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..37db99e82d3ff43a6fb853010a26c3b4e36914a1 --- /dev/null +++ b/childes/EN/adult/20/00/orthographic_lstm/model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ac0bf5262860476e6789d19db0f6fc9547fd778d2cc3ba0010a8fadf6118b9fc +size 87825996 diff --git a/childes/EN/adult/20/00/orthographic_lstm/training.logs b/childes/EN/adult/20/00/orthographic_lstm/training.logs new file mode 100644 index 0000000000000000000000000000000000000000..e8245ef10493bec26871a92287588acde84af761 --- /dev/null +++ b/childes/EN/adult/20/00/orthographic_lstm/training.logs @@ -0,0 +1,36 @@ +06/19/2025 01:08:49 - lexical_benchmark.train.run_train - INFO - Loading lstm trainer class. +06/19/2025 01:09:00 - datasets - INFO - PyTorch version 2.5.1 available. +06/19/2025 01:09:00 - datasets - INFO - Polars version 1.24.0 available. +06/19/2025 01:09:01 - lexical_benchmark.train.train_params - INFO - Loading params from /lustre/fsn1/projects/rech/hhb/uye44va/workspace/lm_benchmark/dev/code/src/lexical_benchmark/train/train-params.toml +06/19/2025 01:09:01 - lexical_benchmark.train.trainers.lstm - INFO - Loading char-tokenizer +06/19/2025 01:09:01 - lexical_benchmark.train.trainers.lstm - INFO - Tokenizer loaded with vocabulary size: 115 +06/19/2025 01:09:01 - lexical_benchmark.train.trainers.lstm - INFO - Loading and processing datasets with joined utterances (max_length=1024) +06/19/2025 01:09:03 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/childes/by_size/EN/adult/20/00/train.tokenized +06/19/2025 01:09:04 - lexical_benchmark.train.tokenizers - INFO - Read 2254077 lines from file +06/19/2025 01:10:23 - lexical_benchmark.train.tokenizers - INFO - Created 47509 joined examples from 2254077 original lines +06/19/2025 01:10:31 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/childes/by_size/EN/adult/dev/dev.tokenized +06/19/2025 01:10:31 - lexical_benchmark.train.tokenizers - INFO - Read 316899 lines from file +06/19/2025 01:10:42 - lexical_benchmark.train.tokenizers - INFO - Created 7325 joined examples from 316899 original lines +06/19/2025 01:10:44 - lexical_benchmark.train.trainers.lstm - INFO - Training dataset size: 47509 +06/19/2025 01:10:44 - lexical_benchmark.train.trainers.lstm - INFO - Validation dataset size: 7325 +06/19/2025 01:10:44 - lexical_benchmark.train.trainers.lstm - INFO - Loading configurations & initialising LSTM model trainer +06/19/2025 01:10:44 - lexical_benchmark.train.run_train - INFO - Running training of childes_lstm_EN_20_00 +06/19/2025 01:12:19 - lexical_benchmark.train.run_train - INFO - Loading lstm trainer class. +06/19/2025 01:12:29 - datasets - INFO - PyTorch version 2.5.1 available. +06/19/2025 01:12:29 - datasets - INFO - Polars version 1.24.0 available. +06/19/2025 01:12:30 - lexical_benchmark.train.train_params - INFO - Loading params from /lustre/fsn1/projects/rech/hhb/uye44va/workspace/lm_benchmark/dev/code/src/lexical_benchmark/train/train-params.toml +06/19/2025 01:12:30 - lexical_benchmark.train.trainers.lstm - INFO - Loading char-tokenizer +06/19/2025 01:12:30 - lexical_benchmark.train.trainers.lstm - INFO - Tokenizer loaded with vocabulary size: 115 +06/19/2025 01:12:30 - lexical_benchmark.train.trainers.lstm - INFO - Loading and processing datasets with joined utterances (max_length=1024) +06/19/2025 01:12:30 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/childes/by_size/EN/adult/20/00/train.tokenized +06/19/2025 01:12:30 - lexical_benchmark.train.tokenizers - INFO - Read 2254077 lines from file +06/19/2025 01:13:52 - lexical_benchmark.train.tokenizers - INFO - Created 47509 joined examples from 2254077 original lines +06/19/2025 01:14:00 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/childes/by_size/EN/adult/dev/dev.tokenized +06/19/2025 01:14:00 - lexical_benchmark.train.tokenizers - INFO - Read 316899 lines from file +06/19/2025 01:14:12 - lexical_benchmark.train.tokenizers - INFO - Created 7325 joined examples from 316899 original lines +06/19/2025 01:14:13 - lexical_benchmark.train.trainers.lstm - INFO - Training dataset size: 47509 +06/19/2025 01:14:13 - lexical_benchmark.train.trainers.lstm - INFO - Validation dataset size: 7325 +06/19/2025 01:14:13 - lexical_benchmark.train.trainers.lstm - INFO - Loading configurations & initialising LSTM model trainer +06/19/2025 01:14:14 - lexical_benchmark.train.run_train - INFO - Running training of childes_lstm_EN_20_00 +06/19/2025 12:43:16 - lexical_benchmark.train.run_train - INFO - Training of childes_lstm_EN_20_00 has completed ! +06/19/2025 12:43:16 - lexical_benchmark.train.run_train - INFO - Model saved to /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/models/childes/EN/adult/20/00/lstm diff --git a/childes/EN/adult/20/00/orthographic_lstm/training_args.bin b/childes/EN/adult/20/00/orthographic_lstm/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..d3fefbaefae7078d70f6debc84d84f538a19e072 --- /dev/null +++ b/childes/EN/adult/20/00/orthographic_lstm/training_args.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:dfd3dc6bcf775b415729eb17cd094c0d97edd266cfdd9b8c72342fce910bfe1d +size 5496 diff --git a/childes/EN/adult/25/00/orthographic_gpt2/config.json b/childes/EN/adult/25/00/orthographic_gpt2/config.json new file mode 100644 index 0000000000000000000000000000000000000000..493011df283c5726a4be3aa8f39c84d43991636e --- /dev/null +++ b/childes/EN/adult/25/00/orthographic_gpt2/config.json @@ -0,0 +1,32 @@ +{ + "activation_function": "gelu_new", + "architectures": [ + "GPT2LMHeadModel" + ], + "attn_pdrop": 0.1, + "bos_token_id": 50256, + "embd_pdrop": 0.1, + "eos_token_id": 50256, + "initializer_range": 0.02, + "layer_norm_epsilon": 1e-05, + "model_type": "gpt2", + "n_ctx": 1024, + "n_embd": 768, + "n_head": 12, + "n_inner": null, + "n_layer": 12, + "n_positions": 1024, + "reorder_and_upcast_attn": false, + "resid_pdrop": 0.1, + "scale_attn_by_inverse_layer_idx": false, + "scale_attn_weights": true, + "summary_activation": null, + "summary_first_dropout": 0.1, + "summary_proj_to_labels": true, + "summary_type": "cls_index", + "summary_use_proj": true, + "torch_dtype": "float32", + "transformers_version": "4.49.0", + "use_cache": true, + "vocab_size": 115 +} diff --git a/childes/EN/adult/25/00/orthographic_gpt2/generation_config.json b/childes/EN/adult/25/00/orthographic_gpt2/generation_config.json new file mode 100644 index 0000000000000000000000000000000000000000..bd968f46ce39e754a4603c69f53a2a4a08a13dbb --- /dev/null +++ b/childes/EN/adult/25/00/orthographic_gpt2/generation_config.json @@ -0,0 +1,6 @@ +{ + "_from_model_config": true, + "bos_token_id": 50256, + "eos_token_id": 50256, + "transformers_version": "4.49.0" +} diff --git a/childes/EN/adult/25/00/orthographic_gpt2/model.safetensors b/childes/EN/adult/25/00/orthographic_gpt2/model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..f8c30e7577302bb9d125f53c424d752adc65995f --- /dev/null +++ b/childes/EN/adult/25/00/orthographic_gpt2/model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d23de09ee8d83dc39a0ee9f3d3d36e7ad26fc3da057613aca7ad6e319864463b +size 343737984 diff --git a/childes/EN/adult/25/00/orthographic_gpt2/training.logs b/childes/EN/adult/25/00/orthographic_gpt2/training.logs new file mode 100644 index 0000000000000000000000000000000000000000..b1154ecad1a37f72d645250d3c2ed535384dec9d --- /dev/null +++ b/childes/EN/adult/25/00/orthographic_gpt2/training.logs @@ -0,0 +1,38 @@ +06/16/2025 19:08:05 - lexical_benchmark.train.run_train - INFO - Loading gpt2 trainer class. +06/16/2025 19:08:09 - datasets - INFO - PyTorch version 2.5.1 available. +06/16/2025 19:08:09 - datasets - INFO - Polars version 1.24.0 available. +06/16/2025 19:08:09 - lexical_benchmark.train.train_params - INFO - Loading params from /lustre/fsn1/projects/rech/hhb/uye44va/workspace/lm_benchmark/dev/code/src/lexical_benchmark/train/train-params.toml +06/16/2025 19:08:09 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading char tokenizer +06/16/2025 19:08:09 - lexical_benchmark.train.trainers.gpt2 - INFO - Tokenizer loaded with vocabulary size: 115 +06/16/2025 19:08:09 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading and processing datasets with joined utterances (max_length=1024) +06/16/2025 19:08:09 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/childes/txt/adult/EN.tokenized.train.txt +06/16/2025 19:08:10 - lexical_benchmark.train.tokenizers - INFO - Read 3381273 lines from file +06/16/2025 19:10:19 - lexical_benchmark.train.tokenizers - INFO - Created 71947 joined examples from 3381273 original lines +06/16/2025 19:10:31 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/childes/txt/adult/EN.tokenized.dev.txt +06/16/2025 19:10:31 - lexical_benchmark.train.tokenizers - INFO - Read 353438 lines from file +06/16/2025 19:10:44 - lexical_benchmark.train.tokenizers - INFO - Created 6552 joined examples from 353438 original lines +06/16/2025 19:10:45 - lexical_benchmark.train.trainers.gpt2 - INFO - Training dataset size: 71947 +06/16/2025 19:10:45 - lexical_benchmark.train.trainers.gpt2 - INFO - Validation dataset size: 6552 +06/16/2025 19:10:45 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating GPT2 configuration +06/16/2025 19:10:46 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating standard Trainer +06/16/2025 19:10:47 - lexical_benchmark.train.run_train - INFO - Running training of childes_gpt2_childes_EN_adult +06/16/2025 19:24:45 - lexical_benchmark.train.run_train - INFO - Loading gpt2 trainer class. +06/16/2025 19:24:55 - datasets - INFO - PyTorch version 2.5.1 available. +06/16/2025 19:24:55 - datasets - INFO - Polars version 1.24.0 available. +06/16/2025 19:24:56 - lexical_benchmark.train.train_params - INFO - Loading params from /lustre/fsn1/projects/rech/hhb/uye44va/workspace/lm_benchmark/dev/code/src/lexical_benchmark/train/train-params.toml +06/16/2025 19:24:56 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading char tokenizer +06/16/2025 19:24:56 - lexical_benchmark.train.trainers.gpt2 - INFO - Tokenizer loaded with vocabulary size: 115 +06/16/2025 19:24:56 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading and processing datasets with joined utterances (max_length=1024) +06/16/2025 19:24:56 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/childes/txt/adult/EN.tokenized.train.txt +06/16/2025 19:24:56 - lexical_benchmark.train.tokenizers - INFO - Read 3381273 lines from file +06/16/2025 19:26:57 - lexical_benchmark.train.tokenizers - INFO - Created 71947 joined examples from 3381273 original lines +06/16/2025 19:27:09 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/childes/txt/adult/EN.tokenized.dev.txt +06/16/2025 19:27:09 - lexical_benchmark.train.tokenizers - INFO - Read 353438 lines from file +06/16/2025 19:27:20 - lexical_benchmark.train.tokenizers - INFO - Created 6552 joined examples from 353438 original lines +06/16/2025 19:27:21 - lexical_benchmark.train.trainers.gpt2 - INFO - Training dataset size: 71947 +06/16/2025 19:27:21 - lexical_benchmark.train.trainers.gpt2 - INFO - Validation dataset size: 6552 +06/16/2025 19:27:21 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating GPT2 configuration +06/16/2025 19:27:22 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating standard Trainer +06/16/2025 19:27:25 - lexical_benchmark.train.run_train - INFO - Running training of childes_gpt2_childes_EN_adult +06/17/2025 04:05:49 - lexical_benchmark.train.run_train - INFO - Training of childes_gpt2_childes_EN_adult has completed ! +06/17/2025 04:05:50 - lexical_benchmark.train.run_train - INFO - Model saved to /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/models/childes/adult/gpt2 diff --git a/childes/EN/adult/25/00/orthographic_gpt2/training_args.bin b/childes/EN/adult/25/00/orthographic_gpt2/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..f4704253516bc4447ed46190a706c1443e9f2b4d --- /dev/null +++ b/childes/EN/adult/25/00/orthographic_gpt2/training_args.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:43d538a1608c1c6aed0a01654ac05e1a8424269b12596a7eb79f5a9d41bd8759 +size 5496 diff --git a/childes/EN/adult/25/00/orthographic_lstm/config.json b/childes/EN/adult/25/00/orthographic_lstm/config.json new file mode 100644 index 0000000000000000000000000000000000000000..5ef629887c7fc87d64278c9788e493ac34c60370 --- /dev/null +++ b/childes/EN/adult/25/00/orthographic_lstm/config.json @@ -0,0 +1,13 @@ +{ + "architectures": [ + "LSTMForLanguageModeling" + ], + "dropout": 0.1, + "embedding_dim": 200, + "hidden_size": 1024, + "model_type": "lstm", + "num_layers": 3, + "torch_dtype": "float32", + "transformers_version": "4.49.0", + "vocab_size": 115 +} diff --git a/childes/EN/adult/25/00/orthographic_lstm/generation_config.json b/childes/EN/adult/25/00/orthographic_lstm/generation_config.json new file mode 100644 index 0000000000000000000000000000000000000000..c1818a95e07bdccb83305ce8f9d878755e3c6119 --- /dev/null +++ b/childes/EN/adult/25/00/orthographic_lstm/generation_config.json @@ -0,0 +1,4 @@ +{ + "_from_model_config": true, + "transformers_version": "4.49.0" +} diff --git a/childes/EN/adult/25/00/orthographic_lstm/model.safetensors b/childes/EN/adult/25/00/orthographic_lstm/model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..d711cbe9eca270abe1d6e90d8a2575b4e13436c7 --- /dev/null +++ b/childes/EN/adult/25/00/orthographic_lstm/model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3dce5ea2d975b1e64f78caae9f8223780c45dce083af4d339a88986cf270724c +size 87825996 diff --git a/childes/EN/adult/25/00/orthographic_lstm/training.logs b/childes/EN/adult/25/00/orthographic_lstm/training.logs new file mode 100644 index 0000000000000000000000000000000000000000..d387282195abe0dfaacdeba2405761b4f77ce898 --- /dev/null +++ b/childes/EN/adult/25/00/orthographic_lstm/training.logs @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6e44294d22cfd922d67a7c1a2ccd9742c3d829f949b8aac6000d668a2d612073 +size 85801225 diff --git a/childes/EN/adult/25/00/orthographic_lstm/training_args.bin b/childes/EN/adult/25/00/orthographic_lstm/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..d1fb2c74338ce876c02df2447d8c505806368eaf --- /dev/null +++ b/childes/EN/adult/25/00/orthographic_lstm/training_args.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6b7b8fcef6a2739d8b335014bff58c5a40c48adb2f6e99e549420794f7f3bccf +size 5496 diff --git a/childes/to_train.toml b/childes/to_train.toml new file mode 100644 index 0000000000000000000000000000000000000000..6b31f924c2d8a8d86e3fd0a93ea8d040a165d45b --- /dev/null +++ b/childes/to_train.toml @@ -0,0 +1,29 @@ +[index.0] +model_type = 'lstm' +lang = 'EN' +split = '10' +chunk = '00' +dataset_name = 'childes' +completed_training = false +generation_root = '/lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/generations/checkpoints.clean/childes/adult/EN/01/00/lstm' +model_root = '/lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/models/childes/adult/EN/01/00/lstm' +last_checkpoint = 'null' +resume = true +override = false +resume_id = 'null' +batch_size = 128 + +[index.1] +model_type = 'gpt2' +lang = 'EN' +split = '10' +chunk = '00' +dataset_name = 'childes' +completed_training = false +generation_root = '/lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/generations/checkpoints.clean/childes/adult/EN/01/00/gpt2' +model_root = '/lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/models/childes/adult/EN/01/00/gpt2' +last_checkpoint = 'null' +resume = true +override = false +resume_id = 'null' +batch_size = 32 diff --git a/stela/EN/01/00/orthographic_gpt2/config.json b/stela/EN/01/00/orthographic_gpt2/config.json new file mode 100644 index 0000000000000000000000000000000000000000..493011df283c5726a4be3aa8f39c84d43991636e --- /dev/null +++ b/stela/EN/01/00/orthographic_gpt2/config.json @@ -0,0 +1,32 @@ +{ + "activation_function": "gelu_new", + "architectures": [ + "GPT2LMHeadModel" + ], + "attn_pdrop": 0.1, + "bos_token_id": 50256, + "embd_pdrop": 0.1, + "eos_token_id": 50256, + "initializer_range": 0.02, + "layer_norm_epsilon": 1e-05, + "model_type": "gpt2", + "n_ctx": 1024, + "n_embd": 768, + "n_head": 12, + "n_inner": null, + "n_layer": 12, + "n_positions": 1024, + "reorder_and_upcast_attn": false, + "resid_pdrop": 0.1, + "scale_attn_by_inverse_layer_idx": false, + "scale_attn_weights": true, + "summary_activation": null, + "summary_first_dropout": 0.1, + "summary_proj_to_labels": true, + "summary_type": "cls_index", + "summary_use_proj": true, + "torch_dtype": "float32", + "transformers_version": "4.49.0", + "use_cache": true, + "vocab_size": 115 +} diff --git a/stela/EN/01/00/orthographic_gpt2/generation_config.json b/stela/EN/01/00/orthographic_gpt2/generation_config.json new file mode 100644 index 0000000000000000000000000000000000000000..bd968f46ce39e754a4603c69f53a2a4a08a13dbb --- /dev/null +++ b/stela/EN/01/00/orthographic_gpt2/generation_config.json @@ -0,0 +1,6 @@ +{ + "_from_model_config": true, + "bos_token_id": 50256, + "eos_token_id": 50256, + "transformers_version": "4.49.0" +} diff --git a/stela/EN/01/00/orthographic_gpt2/model.safetensors b/stela/EN/01/00/orthographic_gpt2/model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..af974abfc7fc18da25eb75c538092e30e8bdf37d --- /dev/null +++ b/stela/EN/01/00/orthographic_gpt2/model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c375a480783b7e3472c71b82bcd2c2236605a1903e5ba93c359adccb6c6e88e5 +size 343737984 diff --git a/stela/EN/01/00/orthographic_gpt2/training.logs b/stela/EN/01/00/orthographic_gpt2/training.logs new file mode 100644 index 0000000000000000000000000000000000000000..735e96a8d2f632f529388b15dbf28840b9a1eaa2 --- /dev/null +++ b/stela/EN/01/00/orthographic_gpt2/training.logs @@ -0,0 +1,315 @@ +04/30/2025 14:46:58 - lexical_benchmark.train.run_train - INFO - Loading gpt2 trainer class. +04/30/2025 14:47:03 - datasets - INFO - PyTorch version 2.5.1 available. +04/30/2025 14:47:03 - datasets - INFO - Polars version 1.26.0 available. +04/30/2025 14:47:03 - lexical_benchmark.train.train_params - INFO - Loading params from /lustre/fsn1/projects/rech/hhb/uye44va/workspace/lm_benchmark/prod/code/src/lexical_benchmark/train/train-params.toml +04/30/2025 14:47:03 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading char tokenizer +04/30/2025 14:47:03 - lexical_benchmark.train.trainers.gpt2 - INFO - Tokenizer loaded with vocabulary size: 115 +04/30/2025 14:47:03 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading and processing datasets with joined utterances (max_length=1024) +04/30/2025 14:47:03 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/01/00/train.tokenized +04/30/2025 14:47:03 - lexical_benchmark.train.tokenizers - INFO - Read 33094 lines from file +04/30/2025 14:47:06 - lexical_benchmark.train.tokenizers - INFO - Created 3703 joined examples from 33094 original lines +04/30/2025 14:47:07 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/dev/dev.tokenized +04/30/2025 14:47:07 - lexical_benchmark.train.tokenizers - INFO - Read 179806 lines from file +04/30/2025 14:47:25 - lexical_benchmark.train.tokenizers - INFO - Created 18373 joined examples from 179806 original lines +04/30/2025 14:47:28 - lexical_benchmark.train.trainers.gpt2 - INFO - Training dataset size: 3703 +04/30/2025 14:47:28 - lexical_benchmark.train.trainers.gpt2 - INFO - Validation dataset size: 18373 +04/30/2025 14:47:28 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating GPT2 configuration +04/30/2025 14:47:28 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating standard Trainer +04/30/2025 14:47:29 - lexical_benchmark.train.run_train - INFO - Running training of stela_gpt2_EN_01_00 +04/30/2025 14:51:59 - lexical_benchmark.train.run_train - INFO - Loading gpt2 trainer class. +04/30/2025 14:52:04 - datasets - INFO - PyTorch version 2.5.1 available. +04/30/2025 14:52:04 - datasets - INFO - Polars version 1.26.0 available. +04/30/2025 14:52:04 - lexical_benchmark.train.train_params - INFO - Loading params from /lustre/fsn1/projects/rech/hhb/uye44va/workspace/lm_benchmark/prod/code/src/lexical_benchmark/train/train-params.toml +04/30/2025 14:52:04 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading char tokenizer +04/30/2025 14:52:04 - lexical_benchmark.train.trainers.gpt2 - INFO - Tokenizer loaded with vocabulary size: 115 +04/30/2025 14:52:04 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading and processing datasets with joined utterances (max_length=1024) +04/30/2025 14:52:04 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/01/00/train.tokenized +04/30/2025 14:52:04 - lexical_benchmark.train.tokenizers - INFO - Read 33094 lines from file +04/30/2025 14:52:07 - lexical_benchmark.train.tokenizers - INFO - Created 3703 joined examples from 33094 original lines +04/30/2025 14:52:08 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/dev/dev.tokenized +04/30/2025 14:52:08 - lexical_benchmark.train.tokenizers - INFO - Read 179806 lines from file +04/30/2025 14:52:26 - lexical_benchmark.train.tokenizers - INFO - Created 18373 joined examples from 179806 original lines +04/30/2025 14:52:29 - lexical_benchmark.train.trainers.gpt2 - INFO - Training dataset size: 3703 +04/30/2025 14:52:29 - lexical_benchmark.train.trainers.gpt2 - INFO - Validation dataset size: 18373 +04/30/2025 14:52:29 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating GPT2 configuration +04/30/2025 14:52:30 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating standard Trainer +04/30/2025 14:52:30 - lexical_benchmark.train.run_train - INFO - Running training of stela_gpt2_EN_01_00 +... +05/23/2025 09:42:37 - urllib3.connectionpool - WARNING - Retrying (Retry(total=2, connect=None, read=None, redirect=None, status=None)) after connection broken by 'NewConnectionError(': Failed to establish a new connection: [Errno 101] Network is unreachable')': /api/4504800232407040/envelope/ +05/23/2025 09:42:37 - urllib3.connectionpool - WARNING - Retrying (Retry(total=1, connect=None, read=None, redirect=None, status=None)) after connection broken by 'NewConnectionError(': Failed to establish a new connection: [Errno 101] Network is unreachable')': /api/4504800232407040/envelope/ +05/23/2025 09:42:37 - urllib3.connectionpool - WARNING - Retrying (Retry(total=0, connect=None, read=None, redirect=None, status=None)) after connection broken by 'NewConnectionError(': Failed to establish a new connection: [Errno 101] Network is unreachable')': /api/4504800232407040/envelope/ +05/23/2025 09:42:37 - urllib3.connectionpool - WARNING - Retrying (Retry(total=2, connect=None, read=None, redirect=None, status=None)) after connection broken by 'NewConnectionError(': Failed to establish a new connection: [Errno 101] Network is unreachable')': /api/4504800232407040/envelope/ +05/23/2025 09:42:37 - urllib3.connectionpool - WARNING - Retrying (Retry(total=1, connect=None, read=None, redirect=None, status=None)) after connection broken by 'NewConnectionError(': Failed to establish a new connection: [Errno 101] Network is unreachable')': /api/4504800232407040/envelope/ +05/23/2025 09:42:37 - urllib3.connectionpool - WARNING - Retrying (Retry(total=0, connect=None, read=None, redirect=None, status=None)) after connection broken by 'NewConnectionError(': Failed to establish a new connection: [Errno 101] Network is unreachable')': /api/4504800232407040/envelope/ +05/23/2025 10:41:44 - urllib3.connectionpool - WARNING - Retrying (Retry(total=2, connect=None, read=None, redirect=None, status=None)) after connection broken by 'NewConnectionError(': Failed to establish a new connection: [Errno 101] Network is unreachable')': /api/4504800232407040/envelope/ +05/23/2025 10:41:44 - urllib3.connectionpool - WARNING - Retrying (Retry(total=1, connect=None, read=None, redirect=None, status=None)) after connection broken by 'NewConnectionError(': Failed to establish a new connection: [Errno 101] Network is unreachable')': /api/4504800232407040/envelope/ +05/23/2025 10:41:44 - urllib3.connectionpool - WARNING - Retrying (Retry(total=0, connect=None, read=None, redirect=None, status=None)) after connection broken by 'NewConnectionError(': Failed to establish a new connection: [Errno 101] Network is unreachable')': /api/4504800232407040/envelope/ +05/23/2025 10:41:44 - urllib3.connectionpool - WARNING - Retrying (Retry(total=2, connect=None, read=None, redirect=None, status=None)) after connection broken by 'NewConnectionError(': Failed to establish a new connection: [Errno 101] Network is unreachable')': /api/4504800232407040/envelope/ +05/23/2025 10:51:13 - urllib3.connectionpool - WARNING - Retrying (Retry(total=2, connect=None, read=None, redirect=None, status=None)) after connection broken by 'NewConnectionError(': Failed to establish a new connection: [Errno 101] Network is unreachable')': /api/4504800232407040/envelope/ +05/23/2025 10:51:13 - urllib3.connectionpool - WARNING - Retrying (Retry(total=1, connect=None, read=None, redirect=None, status=None)) after connection broken by 'NewConnectionError(': Failed to establish a new connection: [Errno 101] Network is unreachable')': /api/4504800232407040/envelope/ +05/23/2025 10:51:13 - urllib3.connectionpool - WARNING - Retrying (Retry(total=0, connect=None, read=None, redirect=None, status=None)) after connection broken by 'NewConnectionError(': Failed to establish a new connection: [Errno 101] Network is unreachable')': /api/4504800232407040/envelope/ +05/23/2025 10:51:13 - urllib3.connectionpool - WARNING - Retrying (Retry(total=2, connect=None, read=None, redirect=None, status=None)) after connection broken by 'NewConnectionError(': Failed to establish a new connection: [Errno 101] Network is unreachable')': /api/4504800232407040/envelope/ +05/23/2025 10:51:13 - urllib3.connectionpool - WARNING - Retrying (Retry(total=1, connect=None, read=None, redirect=None, status=None)) after connection broken by 'NewConnectionError(': Failed to establish a new connection: [Errno 101] Network is unreachable')': /api/4504800232407040/envelope/ +05/23/2025 10:51:13 - urllib3.connectionpool - WARNING - Retrying (Retry(total=0, connect=None, read=None, redirect=None, status=None)) after connection broken by 'NewConnectionError(': Failed to establish a new connection: [Errno 101] Network is unreachable')': /api/4504800232407040/envelope/ +05/23/2025 11:12:51 - urllib3.connectionpool - WARNING - Retrying (Retry(total=2, connect=None, read=None, redirect=None, status=None)) after connection broken by 'NewConnectionError(': Failed to establish a new connection: [Errno 101] Network is unreachable')': /api/4504800232407040/envelope/ +05/23/2025 11:12:51 - urllib3.connectionpool - WARNING - Retrying (Retry(total=1, connect=None, read=None, redirect=None, status=None)) after connection broken by 'NewConnectionError(': Failed to establish a new connection: [Errno 101] Network is unreachable')': /api/4504800232407040/envelope/ +05/23/2025 11:12:51 - urllib3.connectionpool - WARNING - Retrying (Retry(total=0, connect=None, read=None, redirect=None, status=None)) after connection broken by 'NewConnectionError(': Failed to establish a new connection: [Errno 101] Network is unreachable')': /api/4504800232407040/envelope/ +05/23/2025 11:12:51 - urllib3.connectionpool - WARNING - Retrying (Retry(total=2, connect=None, read=None, redirect=None, status=None)) after connection broken by 'NewConnectionError(': Failed to establish a new connection: [Errno 101] Network is unreachable')': /api/4504800232407040/envelope/ +05/23/2025 11:12:51 - urllib3.connectionpool - WARNING - Retrying (Retry(total=1, connect=None, read=None, redirect=None, status=None)) after connection broken by 'NewConnectionError(': Failed to establish a new connection: [Errno 101] Network is unreachable')': /api/4504800232407040/envelope/ +05/23/2025 11:12:51 - urllib3.connectionpool - WARNING - Retrying (Retry(total=0, connect=None, read=None, redirect=None, status=None)) after connection broken by 'NewConnectionError(': Failed to establish a new connection: [Errno 101] Network is unreachable')': /api/4504800232407040/envelope/ +05/23/2025 11:16:30 - lexical_benchmark.train.run_train - INFO - Loading gpt2 trainer class. +05/23/2025 11:16:45 - datasets - INFO - PyTorch version 2.5.1 available. +05/23/2025 11:16:45 - datasets - INFO - Polars version 1.26.0 available. +05/23/2025 11:16:46 - lexical_benchmark.train.train_params - INFO - Loading params from /lustre/fsn1/projects/rech/hhb/uye44va/workspace/lm_benchmark/prod/code/src/lexical_benchmark/train/train-params.toml +05/23/2025 11:16:46 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading char tokenizer +05/23/2025 11:16:46 - lexical_benchmark.train.trainers.gpt2 - INFO - Tokenizer loaded with vocabulary size: 115 +05/23/2025 11:16:46 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading and processing datasets with joined utterances (max_length=1024) +05/23/2025 11:16:46 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/01/00/train.tokenized +05/23/2025 11:16:46 - lexical_benchmark.train.tokenizers - INFO - Read 33094 lines from file +05/23/2025 11:16:50 - lexical_benchmark.train.tokenizers - INFO - Created 3703 joined examples from 33094 original lines +05/23/2025 11:16:50 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/dev/dev.tokenized +05/23/2025 11:16:50 - lexical_benchmark.train.tokenizers - INFO - Read 179806 lines from file +05/23/2025 11:17:08 - lexical_benchmark.train.tokenizers - INFO - Created 18373 joined examples from 179806 original lines +05/23/2025 11:17:11 - lexical_benchmark.train.trainers.gpt2 - INFO - Training dataset size: 3703 +05/23/2025 11:17:11 - lexical_benchmark.train.trainers.gpt2 - INFO - Validation dataset size: 18373 +05/23/2025 11:17:11 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating GPT2 configuration +05/23/2025 11:17:12 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating standard Trainer +05/23/2025 11:17:12 - lexical_benchmark.train.run_train - INFO - Running training of stela_gpt2_EN_01_00 +05/23/2025 11:29:47 - lexical_benchmark.train.run_train - INFO - Loading gpt2 trainer class. +05/23/2025 11:30:00 - datasets - INFO - PyTorch version 2.5.1 available. +05/23/2025 11:30:00 - datasets - INFO - Polars version 1.26.0 available. +05/23/2025 11:30:00 - lexical_benchmark.train.train_params - INFO - Loading params from /lustre/fsn1/projects/rech/hhb/uye44va/workspace/lm_benchmark/prod/code/src/lexical_benchmark/train/train-params.toml +05/23/2025 11:30:00 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading char tokenizer +05/23/2025 11:30:00 - lexical_benchmark.train.trainers.gpt2 - INFO - Tokenizer loaded with vocabulary size: 115 +05/23/2025 11:30:00 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading and processing datasets with joined utterances (max_length=1024) +05/23/2025 11:30:00 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/01/00/train.tokenized +05/23/2025 11:30:00 - lexical_benchmark.train.tokenizers - INFO - Read 33094 lines from file +05/23/2025 11:30:04 - lexical_benchmark.train.tokenizers - INFO - Created 3703 joined examples from 33094 original lines +05/23/2025 11:30:05 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/dev/dev.tokenized +05/23/2025 11:30:05 - lexical_benchmark.train.tokenizers - INFO - Read 179806 lines from file +05/23/2025 11:30:24 - lexical_benchmark.train.tokenizers - INFO - Created 18373 joined examples from 179806 original lines +05/23/2025 11:30:28 - lexical_benchmark.train.trainers.gpt2 - INFO - Training dataset size: 3703 +05/23/2025 11:30:28 - lexical_benchmark.train.trainers.gpt2 - INFO - Validation dataset size: 18373 +05/23/2025 11:30:28 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating GPT2 configuration +05/23/2025 11:30:29 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating standard Trainer +05/23/2025 11:30:29 - lexical_benchmark.train.run_train - INFO - Running training of stela_gpt2_EN_01_00 +05/23/2025 16:13:20 - lexical_benchmark.train.run_train - INFO - Loading gpt2 trainer class. +05/23/2025 16:13:33 - datasets - INFO - PyTorch version 2.5.1 available. +05/23/2025 16:13:33 - datasets - INFO - Polars version 1.26.0 available. +05/23/2025 16:13:34 - lexical_benchmark.train.train_params - INFO - Loading params from /lustre/fsn1/projects/rech/hhb/uye44va/workspace/lm_benchmark/prod/code/src/lexical_benchmark/train/train-params.toml +05/23/2025 16:13:34 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading char tokenizer +05/23/2025 16:13:34 - lexical_benchmark.train.trainers.gpt2 - INFO - Tokenizer loaded with vocabulary size: 115 +05/23/2025 16:13:34 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading and processing datasets with joined utterances (max_length=1024) +05/23/2025 16:13:34 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/01/00/train.tokenized +05/23/2025 16:13:34 - lexical_benchmark.train.tokenizers - INFO - Read 33094 lines from file +05/23/2025 16:13:37 - lexical_benchmark.train.tokenizers - INFO - Created 3703 joined examples from 33094 original lines +05/23/2025 16:13:38 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/dev/dev.tokenized +05/23/2025 16:13:38 - lexical_benchmark.train.tokenizers - INFO - Read 179806 lines from file +05/23/2025 16:13:55 - lexical_benchmark.train.tokenizers - INFO - Created 18373 joined examples from 179806 original lines +05/23/2025 16:13:58 - lexical_benchmark.train.trainers.gpt2 - INFO - Training dataset size: 3703 +05/23/2025 16:13:58 - lexical_benchmark.train.trainers.gpt2 - INFO - Validation dataset size: 18373 +05/23/2025 16:13:58 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating GPT2 configuration +05/23/2025 16:13:59 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating standard Trainer +05/23/2025 16:14:00 - lexical_benchmark.train.run_train - INFO - Running training of stela_gpt2_EN_01_00 +05/23/2025 16:24:47 - lexical_benchmark.train.run_train - INFO - Loading gpt2 trainer class. +05/23/2025 16:24:58 - datasets - INFO - PyTorch version 2.5.1 available. +05/23/2025 16:24:58 - datasets - INFO - Polars version 1.26.0 available. +05/23/2025 16:24:59 - lexical_benchmark.train.train_params - INFO - Loading params from /lustre/fsn1/projects/rech/hhb/uye44va/workspace/lm_benchmark/prod/code/src/lexical_benchmark/train/train-params.toml +05/23/2025 16:24:59 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading char tokenizer +05/23/2025 16:24:59 - lexical_benchmark.train.trainers.gpt2 - INFO - Tokenizer loaded with vocabulary size: 115 +05/23/2025 16:24:59 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading and processing datasets with joined utterances (max_length=1024) +05/23/2025 16:24:59 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/01/00/train.tokenized +05/23/2025 16:24:59 - lexical_benchmark.train.tokenizers - INFO - Read 33094 lines from file +05/23/2025 16:25:02 - lexical_benchmark.train.tokenizers - INFO - Created 3703 joined examples from 33094 original lines +05/23/2025 16:25:03 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/dev/dev.tokenized +05/23/2025 16:25:03 - lexical_benchmark.train.tokenizers - INFO - Read 179806 lines from file +05/23/2025 16:25:19 - lexical_benchmark.train.tokenizers - INFO - Created 18373 joined examples from 179806 original lines +05/23/2025 16:25:22 - lexical_benchmark.train.trainers.gpt2 - INFO - Training dataset size: 3703 +05/23/2025 16:25:22 - lexical_benchmark.train.trainers.gpt2 - INFO - Validation dataset size: 18373 +05/23/2025 16:25:22 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating GPT2 configuration +05/23/2025 16:25:23 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating standard Trainer +05/23/2025 16:25:23 - lexical_benchmark.train.run_train - INFO - Running training of stela_gpt2_EN_01_00 +05/23/2025 16:35:44 - lexical_benchmark.train.run_train - INFO - Loading gpt2 trainer class. +05/23/2025 16:35:48 - datasets - INFO - PyTorch version 2.5.1 available. +05/23/2025 16:35:48 - datasets - INFO - Polars version 1.26.0 available. +05/23/2025 16:35:48 - lexical_benchmark.train.train_params - INFO - Loading params from /lustre/fsn1/projects/rech/hhb/uye44va/workspace/lm_benchmark/prod/code/src/lexical_benchmark/train/train-params.toml +05/23/2025 16:35:48 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading char tokenizer +05/23/2025 16:35:48 - lexical_benchmark.train.trainers.gpt2 - INFO - Tokenizer loaded with vocabulary size: 115 +05/23/2025 16:35:48 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading and processing datasets with joined utterances (max_length=1024) +05/23/2025 16:35:48 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/01/00/train.tokenized +05/23/2025 16:35:48 - lexical_benchmark.train.tokenizers - INFO - Read 33094 lines from file +05/23/2025 16:35:51 - lexical_benchmark.train.tokenizers - INFO - Created 3703 joined examples from 33094 original lines +05/23/2025 16:35:52 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/dev/dev.tokenized +05/23/2025 16:35:52 - lexical_benchmark.train.tokenizers - INFO - Read 179806 lines from file +05/23/2025 16:36:09 - lexical_benchmark.train.tokenizers - INFO - Created 18373 joined examples from 179806 original lines +05/23/2025 16:36:12 - lexical_benchmark.train.trainers.gpt2 - INFO - Training dataset size: 3703 +05/23/2025 16:36:12 - lexical_benchmark.train.trainers.gpt2 - INFO - Validation dataset size: 18373 +05/23/2025 16:36:12 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating GPT2 configuration +05/23/2025 16:36:12 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating standard Trainer +05/23/2025 16:36:13 - lexical_benchmark.train.run_train - INFO - Running training of stela_gpt2_EN_01_00 +05/23/2025 16:37:14 - lexical_benchmark.train.run_train - INFO - Loading gpt2 trainer class. +05/23/2025 16:37:18 - datasets - INFO - PyTorch version 2.5.1 available. +05/23/2025 16:37:18 - datasets - INFO - Polars version 1.26.0 available. +05/23/2025 16:37:18 - lexical_benchmark.train.train_params - INFO - Loading params from /lustre/fsn1/projects/rech/hhb/uye44va/workspace/lm_benchmark/prod/code/src/lexical_benchmark/train/train-params.toml +05/23/2025 16:37:18 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading char tokenizer +05/23/2025 16:37:18 - lexical_benchmark.train.trainers.gpt2 - INFO - Tokenizer loaded with vocabulary size: 115 +05/23/2025 16:37:18 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading and processing datasets with joined utterances (max_length=1024) +05/23/2025 16:37:18 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/01/00/train.tokenized +05/23/2025 16:37:18 - lexical_benchmark.train.tokenizers - INFO - Read 33094 lines from file +05/23/2025 16:37:21 - lexical_benchmark.train.tokenizers - INFO - Created 3703 joined examples from 33094 original lines +05/23/2025 16:37:22 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/dev/dev.tokenized +05/23/2025 16:37:22 - lexical_benchmark.train.tokenizers - INFO - Read 179806 lines from file +05/23/2025 16:37:38 - lexical_benchmark.train.tokenizers - INFO - Created 18373 joined examples from 179806 original lines +05/23/2025 16:37:41 - lexical_benchmark.train.trainers.gpt2 - INFO - Training dataset size: 3703 +05/23/2025 16:37:41 - lexical_benchmark.train.trainers.gpt2 - INFO - Validation dataset size: 18373 +05/23/2025 16:37:41 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating GPT2 configuration +05/23/2025 16:37:42 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating standard Trainer +05/23/2025 16:37:42 - lexical_benchmark.train.run_train - INFO - Running training of stela_gpt2_EN_01_00 +05/23/2025 16:37:52 - lexical_benchmark.train.run_train - INFO - Loading gpt2 trainer class. +05/23/2025 16:37:56 - datasets - INFO - PyTorch version 2.5.1 available. +05/23/2025 16:37:56 - datasets - INFO - Polars version 1.26.0 available. +05/23/2025 16:37:56 - lexical_benchmark.train.train_params - INFO - Loading params from /lustre/fsn1/projects/rech/hhb/uye44va/workspace/lm_benchmark/prod/code/src/lexical_benchmark/train/train-params.toml +05/23/2025 16:37:56 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading char tokenizer +05/23/2025 16:37:56 - lexical_benchmark.train.trainers.gpt2 - INFO - Tokenizer loaded with vocabulary size: 115 +05/23/2025 16:37:56 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading and processing datasets with joined utterances (max_length=1024) +05/23/2025 16:37:56 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/01/00/train.tokenized +05/23/2025 16:37:56 - lexical_benchmark.train.tokenizers - INFO - Read 33094 lines from file +05/23/2025 16:37:59 - lexical_benchmark.train.tokenizers - INFO - Created 3703 joined examples from 33094 original lines +05/23/2025 16:38:00 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/dev/dev.tokenized +05/23/2025 16:38:00 - lexical_benchmark.train.tokenizers - INFO - Read 179806 lines from file +05/23/2025 16:38:17 - lexical_benchmark.train.tokenizers - INFO - Created 18373 joined examples from 179806 original lines +05/23/2025 16:38:20 - lexical_benchmark.train.trainers.gpt2 - INFO - Training dataset size: 3703 +05/23/2025 16:38:20 - lexical_benchmark.train.trainers.gpt2 - INFO - Validation dataset size: 18373 +05/23/2025 16:38:20 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating GPT2 configuration +05/23/2025 16:38:21 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating standard Trainer +05/23/2025 16:38:21 - lexical_benchmark.train.run_train - INFO - Running training of stela_gpt2_EN_01_00 +05/23/2025 16:39:06 - lexical_benchmark.train.run_train - INFO - Loading gpt2 trainer class. +05/23/2025 16:39:10 - datasets - INFO - PyTorch version 2.5.1 available. +05/23/2025 16:39:10 - datasets - INFO - Polars version 1.26.0 available. +05/23/2025 16:39:10 - lexical_benchmark.train.train_params - INFO - Loading params from /lustre/fsn1/projects/rech/hhb/uye44va/workspace/lm_benchmark/prod/code/src/lexical_benchmark/train/train-params.toml +05/23/2025 16:39:10 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading char tokenizer +05/23/2025 16:39:10 - lexical_benchmark.train.trainers.gpt2 - INFO - Tokenizer loaded with vocabulary size: 115 +05/23/2025 16:39:10 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading and processing datasets with joined utterances (max_length=1024) +05/23/2025 16:39:10 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/01/00/train.tokenized +05/23/2025 16:39:10 - lexical_benchmark.train.tokenizers - INFO - Read 33094 lines from file +05/23/2025 16:39:14 - lexical_benchmark.train.tokenizers - INFO - Created 3703 joined examples from 33094 original lines +05/23/2025 16:39:14 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/dev/dev.tokenized +05/23/2025 16:39:14 - lexical_benchmark.train.tokenizers - INFO - Read 179806 lines from file +05/23/2025 16:39:31 - lexical_benchmark.train.tokenizers - INFO - Created 18373 joined examples from 179806 original lines +05/23/2025 16:39:34 - lexical_benchmark.train.trainers.gpt2 - INFO - Training dataset size: 3703 +05/23/2025 16:39:34 - lexical_benchmark.train.trainers.gpt2 - INFO - Validation dataset size: 18373 +05/23/2025 16:39:34 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating GPT2 configuration +05/23/2025 16:39:35 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating standard Trainer +05/23/2025 16:39:35 - lexical_benchmark.train.run_train - INFO - Running training of stela_gpt2_EN_01_00 +05/23/2025 17:19:00 - lexical_benchmark.train.run_train - INFO - Loading gpt2 trainer class. +05/23/2025 17:19:11 - datasets - INFO - PyTorch version 2.5.1 available. +05/23/2025 17:19:11 - datasets - INFO - Polars version 1.26.0 available. +05/23/2025 17:19:12 - lexical_benchmark.train.train_params - INFO - Loading params from /lustre/fsn1/projects/rech/hhb/uye44va/workspace/lm_benchmark/prod/code/src/lexical_benchmark/train/train-params.toml +05/23/2025 17:19:12 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading char tokenizer +05/23/2025 17:19:12 - lexical_benchmark.train.trainers.gpt2 - INFO - Tokenizer loaded with vocabulary size: 115 +05/23/2025 17:19:12 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading and processing datasets with joined utterances (max_length=1024) +05/23/2025 17:19:12 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/01/00/train.tokenized +05/23/2025 17:19:12 - lexical_benchmark.train.tokenizers - INFO - Read 33094 lines from file +05/23/2025 17:19:15 - lexical_benchmark.train.tokenizers - INFO - Created 3703 joined examples from 33094 original lines +05/23/2025 17:19:16 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/dev/dev.tokenized +05/23/2025 17:19:16 - lexical_benchmark.train.tokenizers - INFO - Read 179806 lines from file +05/23/2025 17:19:32 - lexical_benchmark.train.tokenizers - INFO - Created 18373 joined examples from 179806 original lines +05/23/2025 17:19:35 - lexical_benchmark.train.trainers.gpt2 - INFO - Training dataset size: 3703 +05/23/2025 17:19:35 - lexical_benchmark.train.trainers.gpt2 - INFO - Validation dataset size: 18373 +05/23/2025 17:19:35 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating GPT2 configuration +05/23/2025 17:19:36 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating standard Trainer +05/23/2025 17:19:36 - lexical_benchmark.train.run_train - INFO - Running training of stela_gpt2_EN_01_00 +05/23/2025 17:26:19 - lexical_benchmark.train.run_train - INFO - Loading gpt2 trainer class. +05/23/2025 17:26:23 - datasets - INFO - PyTorch version 2.5.1 available. +05/23/2025 17:26:23 - datasets - INFO - Polars version 1.26.0 available. +05/23/2025 17:26:24 - lexical_benchmark.train.train_params - INFO - Loading params from /lustre/fsn1/projects/rech/hhb/uye44va/workspace/lm_benchmark/prod/code/src/lexical_benchmark/train/train-params.toml +05/23/2025 17:26:24 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading char tokenizer +05/23/2025 17:26:24 - lexical_benchmark.train.trainers.gpt2 - INFO - Tokenizer loaded with vocabulary size: 115 +05/23/2025 17:26:24 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading and processing datasets with joined utterances (max_length=1024) +05/23/2025 17:26:24 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/01/00/train.tokenized +05/23/2025 17:26:24 - lexical_benchmark.train.tokenizers - INFO - Read 33094 lines from file +05/23/2025 17:26:27 - lexical_benchmark.train.tokenizers - INFO - Created 3703 joined examples from 33094 original lines +05/23/2025 17:26:27 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/dev/dev.tokenized +05/23/2025 17:26:27 - lexical_benchmark.train.tokenizers - INFO - Read 179806 lines from file +05/23/2025 17:26:44 - lexical_benchmark.train.tokenizers - INFO - Created 18373 joined examples from 179806 original lines +05/23/2025 17:26:47 - lexical_benchmark.train.trainers.gpt2 - INFO - Training dataset size: 3703 +05/23/2025 17:26:47 - lexical_benchmark.train.trainers.gpt2 - INFO - Validation dataset size: 18373 +05/23/2025 17:26:47 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating GPT2 configuration +05/23/2025 17:26:48 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating standard Trainer +05/23/2025 17:26:48 - lexical_benchmark.train.run_train - INFO - Running training of stela_gpt2_EN_01_00 +05/23/2025 17:30:12 - lexical_benchmark.train.run_train - INFO - Loading gpt2 trainer class. +05/23/2025 17:30:16 - datasets - INFO - PyTorch version 2.5.1 available. +05/23/2025 17:30:16 - datasets - INFO - Polars version 1.26.0 available. +05/23/2025 17:30:16 - lexical_benchmark.train.train_params - INFO - Loading params from /lustre/fsn1/projects/rech/hhb/uye44va/workspace/lm_benchmark/prod/code/src/lexical_benchmark/train/train-params.toml +05/23/2025 17:30:16 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading char tokenizer +05/23/2025 17:30:16 - lexical_benchmark.train.trainers.gpt2 - INFO - Tokenizer loaded with vocabulary size: 115 +05/23/2025 17:30:16 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading and processing datasets with joined utterances (max_length=1024) +05/23/2025 17:30:16 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/01/00/train.tokenized +05/23/2025 17:30:16 - lexical_benchmark.train.tokenizers - INFO - Read 33094 lines from file +05/23/2025 17:30:19 - lexical_benchmark.train.tokenizers - INFO - Created 3703 joined examples from 33094 original lines +05/23/2025 17:30:20 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/dev/dev.tokenized +05/23/2025 17:30:20 - lexical_benchmark.train.tokenizers - INFO - Read 179806 lines from file +05/23/2025 17:30:36 - lexical_benchmark.train.tokenizers - INFO - Created 18373 joined examples from 179806 original lines +05/23/2025 17:30:39 - lexical_benchmark.train.trainers.gpt2 - INFO - Training dataset size: 3703 +05/23/2025 17:30:39 - lexical_benchmark.train.trainers.gpt2 - INFO - Validation dataset size: 18373 +05/23/2025 17:30:39 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating GPT2 configuration +05/23/2025 17:30:40 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating standard Trainer +05/23/2025 17:30:40 - lexical_benchmark.train.run_train - INFO - Running training of stela_gpt2_EN_01_00 +05/23/2025 17:39:06 - lexical_benchmark.train.run_train - INFO - Loading gpt2 trainer class. +05/23/2025 17:39:18 - datasets - INFO - PyTorch version 2.5.1 available. +05/23/2025 17:39:18 - datasets - INFO - Polars version 1.26.0 available. +05/23/2025 17:39:18 - lexical_benchmark.train.train_params - INFO - Loading params from /lustre/fsn1/projects/rech/hhb/uye44va/workspace/lm_benchmark/prod/code/src/lexical_benchmark/train/train-params.toml +05/23/2025 17:39:18 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading char tokenizer +05/23/2025 17:39:18 - lexical_benchmark.train.trainers.gpt2 - INFO - Tokenizer loaded with vocabulary size: 115 +05/23/2025 17:39:18 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading and processing datasets with joined utterances (max_length=1024) +05/23/2025 17:39:18 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/01/00/train.tokenized +05/23/2025 17:39:19 - lexical_benchmark.train.tokenizers - INFO - Read 33094 lines from file +05/23/2025 17:39:22 - lexical_benchmark.train.tokenizers - INFO - Created 3703 joined examples from 33094 original lines +05/23/2025 17:39:23 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/dev/dev.tokenized +05/23/2025 17:39:23 - lexical_benchmark.train.tokenizers - INFO - Read 179806 lines from file +05/23/2025 17:39:41 - lexical_benchmark.train.tokenizers - INFO - Created 18373 joined examples from 179806 original lines +05/23/2025 17:39:44 - lexical_benchmark.train.trainers.gpt2 - INFO - Training dataset size: 3703 +05/23/2025 17:39:44 - lexical_benchmark.train.trainers.gpt2 - INFO - Validation dataset size: 18373 +05/23/2025 17:39:44 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating GPT2 configuration +05/23/2025 17:39:45 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating standard Trainer +05/23/2025 17:39:45 - lexical_benchmark.train.run_train - INFO - Running training of stela_gpt2_EN_01_00 +05/23/2025 18:26:14 - lexical_benchmark.train.run_train - INFO - Loading gpt2 trainer class. +05/23/2025 18:26:18 - datasets - INFO - PyTorch version 2.5.1 available. +05/23/2025 18:26:18 - datasets - INFO - Polars version 1.26.0 available. +05/23/2025 18:26:19 - lexical_benchmark.train.train_params - INFO - Loading params from /lustre/fsn1/projects/rech/hhb/uye44va/workspace/lm_benchmark/prod/code/src/lexical_benchmark/train/train-params.toml +05/23/2025 18:26:19 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading char tokenizer +05/23/2025 18:26:19 - lexical_benchmark.train.trainers.gpt2 - INFO - Tokenizer loaded with vocabulary size: 115 +05/23/2025 18:26:19 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading and processing datasets with joined utterances (max_length=1024) +05/23/2025 18:26:19 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/01/00/train.tokenized +05/23/2025 18:26:19 - lexical_benchmark.train.tokenizers - INFO - Read 33094 lines from file +05/23/2025 18:26:22 - lexical_benchmark.train.tokenizers - INFO - Created 3703 joined examples from 33094 original lines +05/23/2025 18:26:23 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/dev/dev.tokenized +05/23/2025 18:26:23 - lexical_benchmark.train.tokenizers - INFO - Read 179806 lines from file +05/23/2025 18:26:40 - lexical_benchmark.train.tokenizers - INFO - Created 18373 joined examples from 179806 original lines +05/23/2025 18:26:43 - lexical_benchmark.train.trainers.gpt2 - INFO - Training dataset size: 3703 +05/23/2025 18:26:43 - lexical_benchmark.train.trainers.gpt2 - INFO - Validation dataset size: 18373 +05/23/2025 18:26:43 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating GPT2 configuration +05/23/2025 18:26:44 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating standard Trainer +05/23/2025 18:26:44 - lexical_benchmark.train.run_train - INFO - Resuming previous training using /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/models/stela/EN/01/00/gpt2/checkpoint-2000 +05/23/2025 18:26:44 - lexical_benchmark.train.run_train - INFO - Running training of stela_gpt2_EN_01_00 +06/03/2025 15:20:47 - lexical_benchmark.train.run_train - INFO - Loading gpt2 trainer class. +06/03/2025 15:20:59 - datasets - INFO - PyTorch version 2.5.1 available. +06/03/2025 15:20:59 - datasets - INFO - Polars version 1.26.0 available. +06/03/2025 15:21:00 - lexical_benchmark.train.train_params - INFO - Loading params from /lustre/fsn1/projects/rech/hhb/uye44va/workspace/lm_benchmark/prod/code/src/lexical_benchmark/train/train-params.toml +06/03/2025 15:21:00 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading char tokenizer +06/03/2025 15:21:00 - lexical_benchmark.train.trainers.gpt2 - INFO - Tokenizer loaded with vocabulary size: 115 +06/03/2025 15:21:00 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading and processing datasets with joined utterances (max_length=1024) +06/03/2025 15:21:00 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/01/00/train.tokenized +06/03/2025 15:21:00 - lexical_benchmark.train.tokenizers - INFO - Read 33094 lines from file +06/03/2025 15:21:03 - lexical_benchmark.train.tokenizers - INFO - Created 3703 joined examples from 33094 original lines +06/03/2025 15:21:04 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/dev/dev.tokenized +06/03/2025 15:21:04 - lexical_benchmark.train.tokenizers - INFO - Read 179806 lines from file +06/03/2025 15:21:22 - lexical_benchmark.train.tokenizers - INFO - Created 18373 joined examples from 179806 original lines +06/03/2025 15:21:25 - lexical_benchmark.train.trainers.gpt2 - INFO - Training dataset size: 3703 +06/03/2025 15:21:25 - lexical_benchmark.train.trainers.gpt2 - INFO - Validation dataset size: 18373 +06/03/2025 15:21:25 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating GPT2 configuration +06/03/2025 15:21:26 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating standard Trainer +06/03/2025 15:21:26 - lexical_benchmark.train.run_train - INFO - Resuming previous training using /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/models/stela/EN/01/00/gpt2/checkpoint-3000 +06/03/2025 15:21:26 - lexical_benchmark.train.run_train - INFO - Running training of stela_gpt2_EN_01_00 +06/03/2025 16:09:19 - lexical_benchmark.train.run_train - INFO - Training of stela_gpt2_EN_01_00 has completed ! +06/03/2025 16:09:19 - lexical_benchmark.train.run_train - INFO - Model saved to /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/models/stela/EN/01/00/gpt2 diff --git a/stela/EN/01/00/orthographic_gpt2/training_args.bin b/stela/EN/01/00/orthographic_gpt2/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..cc0fa8d4a9b0ceb4bf65554376807f2d001a419d --- /dev/null +++ b/stela/EN/01/00/orthographic_gpt2/training_args.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ba039a45c1b986f5a68e1de38ae9d1fc13b8e9f87931dc16fbf0fc66f97b1dd2 +size 5496 diff --git a/stela/EN/01/00/orthographic_lstm/config.json b/stela/EN/01/00/orthographic_lstm/config.json new file mode 100644 index 0000000000000000000000000000000000000000..5ef629887c7fc87d64278c9788e493ac34c60370 --- /dev/null +++ b/stela/EN/01/00/orthographic_lstm/config.json @@ -0,0 +1,13 @@ +{ + "architectures": [ + "LSTMForLanguageModeling" + ], + "dropout": 0.1, + "embedding_dim": 200, + "hidden_size": 1024, + "model_type": "lstm", + "num_layers": 3, + "torch_dtype": "float32", + "transformers_version": "4.49.0", + "vocab_size": 115 +} diff --git a/stela/EN/01/00/orthographic_lstm/model.safetensors b/stela/EN/01/00/orthographic_lstm/model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..46869ed9fbb2857600659c121ca282bb861d6a92 --- /dev/null +++ b/stela/EN/01/00/orthographic_lstm/model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a51c41b77dec2037ae7b525fbf0e4fb119fe7afd596af5399b5c2d23fed19aa6 +size 87825996 diff --git a/stela/EN/01/00/orthographic_lstm/training_args.bin b/stela/EN/01/00/orthographic_lstm/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..3ca2976a7cec42d7b91c9151db2eeb3c72332487 --- /dev/null +++ b/stela/EN/01/00/orthographic_lstm/training_args.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f6334b0907ceada821514896ee444343648d866acb2aca043f45904a5f2640c0 +size 5432 diff --git a/stela/EN/01/01/orthographic_gpt2/config.json b/stela/EN/01/01/orthographic_gpt2/config.json new file mode 100644 index 0000000000000000000000000000000000000000..493011df283c5726a4be3aa8f39c84d43991636e --- /dev/null +++ b/stela/EN/01/01/orthographic_gpt2/config.json @@ -0,0 +1,32 @@ +{ + "activation_function": "gelu_new", + "architectures": [ + "GPT2LMHeadModel" + ], + "attn_pdrop": 0.1, + "bos_token_id": 50256, + "embd_pdrop": 0.1, + "eos_token_id": 50256, + "initializer_range": 0.02, + "layer_norm_epsilon": 1e-05, + "model_type": "gpt2", + "n_ctx": 1024, + "n_embd": 768, + "n_head": 12, + "n_inner": null, + "n_layer": 12, + "n_positions": 1024, + "reorder_and_upcast_attn": false, + "resid_pdrop": 0.1, + "scale_attn_by_inverse_layer_idx": false, + "scale_attn_weights": true, + "summary_activation": null, + "summary_first_dropout": 0.1, + "summary_proj_to_labels": true, + "summary_type": "cls_index", + "summary_use_proj": true, + "torch_dtype": "float32", + "transformers_version": "4.49.0", + "use_cache": true, + "vocab_size": 115 +} diff --git a/stela/EN/01/01/orthographic_gpt2/generation_config.json b/stela/EN/01/01/orthographic_gpt2/generation_config.json new file mode 100644 index 0000000000000000000000000000000000000000..bd968f46ce39e754a4603c69f53a2a4a08a13dbb --- /dev/null +++ b/stela/EN/01/01/orthographic_gpt2/generation_config.json @@ -0,0 +1,6 @@ +{ + "_from_model_config": true, + "bos_token_id": 50256, + "eos_token_id": 50256, + "transformers_version": "4.49.0" +} diff --git a/stela/EN/01/01/orthographic_gpt2/model.safetensors b/stela/EN/01/01/orthographic_gpt2/model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..0c46722deeef554dfadf6a568e5bafaef6538b4e --- /dev/null +++ b/stela/EN/01/01/orthographic_gpt2/model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:28e145ce227b7d74234018bf1e812f99efe0b7977deed8e05ec7f335a0f9d617 +size 343737984 diff --git a/stela/EN/01/01/orthographic_gpt2/training.logs b/stela/EN/01/01/orthographic_gpt2/training.logs new file mode 100644 index 0000000000000000000000000000000000000000..973758efac755fead98e7ccd7e90f5743e27adab --- /dev/null +++ b/stela/EN/01/01/orthographic_gpt2/training.logs @@ -0,0 +1,20 @@ +06/03/2025 16:47:28 - lexical_benchmark.train.run_train - INFO - Loading gpt2 trainer class. +06/03/2025 16:47:40 - datasets - INFO - PyTorch version 2.5.1 available. +06/03/2025 16:47:40 - datasets - INFO - Polars version 1.26.0 available. +06/03/2025 16:47:40 - lexical_benchmark.train.train_params - INFO - Loading params from /lustre/fsn1/projects/rech/hhb/uye44va/workspace/lm_benchmark/prod/code/src/lexical_benchmark/train/train-params.toml +06/03/2025 16:47:40 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading char tokenizer +06/03/2025 16:47:40 - lexical_benchmark.train.trainers.gpt2 - INFO - Tokenizer loaded with vocabulary size: 115 +06/03/2025 16:47:40 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading and processing datasets with joined utterances (max_length=1024) +06/03/2025 16:47:40 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/01/01/train.tokenized +06/03/2025 16:47:40 - lexical_benchmark.train.tokenizers - INFO - Read 43857 lines from file +06/03/2025 16:47:44 - lexical_benchmark.train.tokenizers - INFO - Created 3630 joined examples from 43857 original lines +06/03/2025 16:47:44 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/dev/dev.tokenized +06/03/2025 16:47:44 - lexical_benchmark.train.tokenizers - INFO - Read 179806 lines from file +06/03/2025 16:48:00 - lexical_benchmark.train.tokenizers - INFO - Created 18373 joined examples from 179806 original lines +06/03/2025 16:48:03 - lexical_benchmark.train.trainers.gpt2 - INFO - Training dataset size: 3630 +06/03/2025 16:48:03 - lexical_benchmark.train.trainers.gpt2 - INFO - Validation dataset size: 18373 +06/03/2025 16:48:03 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating GPT2 configuration +06/03/2025 16:48:04 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating standard Trainer +06/03/2025 16:48:05 - lexical_benchmark.train.run_train - INFO - Running training of stela_gpt2_EN_01_01 +06/03/2025 18:06:53 - lexical_benchmark.train.run_train - INFO - Training of stela_gpt2_EN_01_01 has completed ! +06/03/2025 18:06:54 - lexical_benchmark.train.run_train - INFO - Model saved to /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/models/stela/EN/01/01/gpt2 diff --git a/stela/EN/01/01/orthographic_gpt2/training_args.bin b/stela/EN/01/01/orthographic_gpt2/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..539a046da386e4a28bfa63637617f241d00e64df --- /dev/null +++ b/stela/EN/01/01/orthographic_gpt2/training_args.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3fe5b6aae0e549c6c63c06ba3bd99bbb4016ebd2bc864505310ba3b7b34ebf31 +size 5496 diff --git a/stela/EN/01/01/orthographic_lstm/config.json b/stela/EN/01/01/orthographic_lstm/config.json new file mode 100644 index 0000000000000000000000000000000000000000..5ef629887c7fc87d64278c9788e493ac34c60370 --- /dev/null +++ b/stela/EN/01/01/orthographic_lstm/config.json @@ -0,0 +1,13 @@ +{ + "architectures": [ + "LSTMForLanguageModeling" + ], + "dropout": 0.1, + "embedding_dim": 200, + "hidden_size": 1024, + "model_type": "lstm", + "num_layers": 3, + "torch_dtype": "float32", + "transformers_version": "4.49.0", + "vocab_size": 115 +} diff --git a/stela/EN/01/01/orthographic_lstm/generation_config.json b/stela/EN/01/01/orthographic_lstm/generation_config.json new file mode 100644 index 0000000000000000000000000000000000000000..c1818a95e07bdccb83305ce8f9d878755e3c6119 --- /dev/null +++ b/stela/EN/01/01/orthographic_lstm/generation_config.json @@ -0,0 +1,4 @@ +{ + "_from_model_config": true, + "transformers_version": "4.49.0" +} diff --git a/stela/EN/01/01/orthographic_lstm/model.safetensors b/stela/EN/01/01/orthographic_lstm/model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..52da5b9babd434485f96d7d76d6395f0445bc333 --- /dev/null +++ b/stela/EN/01/01/orthographic_lstm/model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:85c38b21b8f1f164d88c8c41968134b0205c82c4df2e37b4bbe4cdae6eca0618 +size 87825996 diff --git a/stela/EN/01/01/orthographic_lstm/training.logs b/stela/EN/01/01/orthographic_lstm/training.logs new file mode 100644 index 0000000000000000000000000000000000000000..526038ce74c39163933ec74c24b88e2005adef19 --- /dev/null +++ b/stela/EN/01/01/orthographic_lstm/training.logs @@ -0,0 +1,19 @@ +06/03/2025 16:47:28 - lexical_benchmark.train.run_train - INFO - Loading lstm trainer class. +06/03/2025 16:47:39 - datasets - INFO - PyTorch version 2.5.1 available. +06/03/2025 16:47:39 - datasets - INFO - Polars version 1.26.0 available. +06/03/2025 16:47:40 - lexical_benchmark.train.train_params - INFO - Loading params from /lustre/fsn1/projects/rech/hhb/uye44va/workspace/lm_benchmark/prod/code/src/lexical_benchmark/train/train-params.toml +06/03/2025 16:47:40 - lexical_benchmark.train.trainers.lstm - INFO - Loading char-tokenizer +06/03/2025 16:47:40 - lexical_benchmark.train.trainers.lstm - INFO - Tokenizer loaded with vocabulary size: 115 +06/03/2025 16:47:40 - lexical_benchmark.train.trainers.lstm - INFO - Loading and processing datasets with joined utterances (max_length=1024) +06/03/2025 16:47:40 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/01/01/train.tokenized +06/03/2025 16:47:40 - lexical_benchmark.train.tokenizers - INFO - Read 43857 lines from file +06/03/2025 16:47:43 - lexical_benchmark.train.tokenizers - INFO - Created 3630 joined examples from 43857 original lines +06/03/2025 16:47:44 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/dev/dev.tokenized +06/03/2025 16:47:44 - lexical_benchmark.train.tokenizers - INFO - Read 179806 lines from file +06/03/2025 16:48:00 - lexical_benchmark.train.tokenizers - INFO - Created 18373 joined examples from 179806 original lines +06/03/2025 16:48:02 - lexical_benchmark.train.trainers.lstm - INFO - Training dataset size: 3630 +06/03/2025 16:48:02 - lexical_benchmark.train.trainers.lstm - INFO - Validation dataset size: 18373 +06/03/2025 16:48:02 - lexical_benchmark.train.trainers.lstm - INFO - Loading configurations & initialising LSTM model trainer +06/03/2025 16:48:03 - lexical_benchmark.train.run_train - INFO - Running training of stela_lstm_EN_01_01 +06/03/2025 17:49:06 - lexical_benchmark.train.run_train - INFO - Training of stela_lstm_EN_01_01 has completed ! +06/03/2025 17:49:06 - lexical_benchmark.train.run_train - INFO - Model saved to /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/models/stela/EN/01/01/lstm diff --git a/stela/EN/01/01/orthographic_lstm/training_args.bin b/stela/EN/01/01/orthographic_lstm/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..1e2f8210fbd437e92daddfd172696f2c7da3decd --- /dev/null +++ b/stela/EN/01/01/orthographic_lstm/training_args.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:35870905125b2128355d05a934f456fbbb5b5d18e63ba7a7d040b49408b0bc6d +size 5496 diff --git a/stela/EN/02/00/orthographic_gpt2/config.json b/stela/EN/02/00/orthographic_gpt2/config.json new file mode 100644 index 0000000000000000000000000000000000000000..493011df283c5726a4be3aa8f39c84d43991636e --- /dev/null +++ b/stela/EN/02/00/orthographic_gpt2/config.json @@ -0,0 +1,32 @@ +{ + "activation_function": "gelu_new", + "architectures": [ + "GPT2LMHeadModel" + ], + "attn_pdrop": 0.1, + "bos_token_id": 50256, + "embd_pdrop": 0.1, + "eos_token_id": 50256, + "initializer_range": 0.02, + "layer_norm_epsilon": 1e-05, + "model_type": "gpt2", + "n_ctx": 1024, + "n_embd": 768, + "n_head": 12, + "n_inner": null, + "n_layer": 12, + "n_positions": 1024, + "reorder_and_upcast_attn": false, + "resid_pdrop": 0.1, + "scale_attn_by_inverse_layer_idx": false, + "scale_attn_weights": true, + "summary_activation": null, + "summary_first_dropout": 0.1, + "summary_proj_to_labels": true, + "summary_type": "cls_index", + "summary_use_proj": true, + "torch_dtype": "float32", + "transformers_version": "4.49.0", + "use_cache": true, + "vocab_size": 115 +} diff --git a/stela/EN/02/00/orthographic_gpt2/generation_config.json b/stela/EN/02/00/orthographic_gpt2/generation_config.json new file mode 100644 index 0000000000000000000000000000000000000000..bd968f46ce39e754a4603c69f53a2a4a08a13dbb --- /dev/null +++ b/stela/EN/02/00/orthographic_gpt2/generation_config.json @@ -0,0 +1,6 @@ +{ + "_from_model_config": true, + "bos_token_id": 50256, + "eos_token_id": 50256, + "transformers_version": "4.49.0" +} diff --git a/stela/EN/02/00/orthographic_gpt2/model.safetensors b/stela/EN/02/00/orthographic_gpt2/model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..2278a71ad7797eeb0cb819da879211367a54ccf4 --- /dev/null +++ b/stela/EN/02/00/orthographic_gpt2/model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7ce9c80a5e35d99613d4cbef2b4a25ce19db227faceb3152d59312b74cc2a9e3 +size 343737984 diff --git a/stela/EN/02/00/orthographic_gpt2/training.logs b/stela/EN/02/00/orthographic_gpt2/training.logs new file mode 100644 index 0000000000000000000000000000000000000000..5c4de541ba65770819846a784f503937fed06109 --- /dev/null +++ b/stela/EN/02/00/orthographic_gpt2/training.logs @@ -0,0 +1,170 @@ +04/30/2025 18:19:34 - lexical_benchmark.train.run_train - INFO - Loading gpt2 trainer class. +04/30/2025 18:19:46 - datasets - INFO - PyTorch version 2.5.1 available. +04/30/2025 18:19:46 - datasets - INFO - Polars version 1.26.0 available. +04/30/2025 18:19:47 - lexical_benchmark.train.train_params - INFO - Loading params from /lustre/fsn1/projects/rech/hhb/uye44va/workspace/lm_benchmark/prod/code/src/lexical_benchmark/train/train-params.toml +04/30/2025 18:19:47 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading char tokenizer +04/30/2025 18:19:47 - lexical_benchmark.train.trainers.gpt2 - INFO - Tokenizer loaded with vocabulary size: 115 +04/30/2025 18:19:47 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading and processing datasets with joined utterances (max_length=1024) +04/30/2025 18:19:47 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/02/02/train.tokenized +04/30/2025 18:19:47 - lexical_benchmark.train.tokenizers - INFO - Read 75727 lines from file +04/30/2025 18:19:54 - lexical_benchmark.train.tokenizers - INFO - Created 6868 joined examples from 75727 original lines +04/30/2025 18:19:55 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/dev/dev.tokenized +04/30/2025 18:19:55 - lexical_benchmark.train.tokenizers - INFO - Read 179806 lines from file +04/30/2025 18:20:11 - lexical_benchmark.train.tokenizers - INFO - Created 18373 joined examples from 179806 original lines +04/30/2025 18:20:14 - lexical_benchmark.train.trainers.gpt2 - INFO - Training dataset size: 6868 +04/30/2025 18:20:14 - lexical_benchmark.train.trainers.gpt2 - INFO - Validation dataset size: 18373 +04/30/2025 18:20:14 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating GPT2 configuration +04/30/2025 18:20:15 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating standard Trainer +04/30/2025 18:20:15 - lexical_benchmark.train.run_train - INFO - Running training of stela_gpt2_EN_02_02 +04/30/2025 19:27:09 - urllib3.connectionpool - WARNING - Retrying (Retry(total=2, connect=None, read=None, redirect=None, status=None)) after connection broken by 'NewConnectionError(': Failed to establish a new connection: [Errno 101] Network is unreachable')': /api/4504800232407040/envelope/ +04/30/2025 19:27:09 - urllib3.connectionpool - WARNING - Retrying (Retry(total=1, connect=None, read=None, redirect=None, status=None)) after connection broken by 'NewConnectionError(': Failed to establish a new connection: [Errno 101] Network is unreachable')': /api/4504800232407040/envelope/ +04/30/2025 19:27:09 - urllib3.connectionpool - WARNING - Retrying (Retry(total=0, connect=None, read=None, redirect=None, status=None)) after connection broken by 'NewConnectionError(': Failed to establish a new connection: [Errno 101] Network is unreachable')': /api/4504800232407040/envelope/ +04/30/2025 19:27:09 - urllib3.connectionpool - WARNING - Retrying (Retry(total=2, connect=None, read=None, redirect=None, status=None)) after connection broken by 'NewConnectionError(': Failed to establish a new connection: [Errno 101] Network is unreachable')': /api/4504800232407040/envelope/ +04/30/2025 19:27:09 - urllib3.connectionpool - WARNING - Retrying (Retry(total=1, connect=None, read=None, redirect=None, status=None)) after connection broken by 'NewConnectionError(': Failed to establish a new connection: [Errno 101] Network is unreachable')': /api/4504800232407040/envelope/ +04/30/2025 19:27:09 - urllib3.connectionpool - WARNING - Retrying (Retry(total=0, connect=None, read=None, redirect=None, status=None)) after connection broken by 'NewConnectionError(': Failed to establish a new connection: [Errno 101] Network is unreachable')': /api/4504800232407040/envelope/ +04/30/2025 19:49:30 - lexical_benchmark.train.run_train - INFO - Loading gpt2 trainer class. +04/30/2025 19:49:42 - datasets - INFO - PyTorch version 2.5.1 available. +04/30/2025 19:49:42 - datasets - INFO - Polars version 1.26.0 available. +04/30/2025 19:49:43 - lexical_benchmark.train.train_params - INFO - Loading params from /lustre/fsn1/projects/rech/hhb/ucx81cx/workspace/lm_benchmark/dev/code/src/lexical_benchmark/train/train-params.toml +04/30/2025 19:49:43 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading char tokenizer +04/30/2025 19:49:43 - lexical_benchmark.train.trainers.gpt2 - INFO - Tokenizer loaded with vocabulary size: 115 +04/30/2025 19:49:43 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading and processing datasets with joined utterances (max_length=1024) +04/30/2025 19:49:43 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/02/02/train.tokenized +04/30/2025 19:49:43 - lexical_benchmark.train.tokenizers - INFO - Read 75727 lines from file +04/30/2025 19:49:49 - lexical_benchmark.train.tokenizers - INFO - Created 6868 joined examples from 75727 original lines +04/30/2025 19:49:50 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/dev/dev.tokenized +04/30/2025 19:49:51 - lexical_benchmark.train.tokenizers - INFO - Read 179806 lines from file +04/30/2025 19:50:07 - lexical_benchmark.train.tokenizers - INFO - Created 18373 joined examples from 179806 original lines +04/30/2025 19:50:09 - lexical_benchmark.train.trainers.gpt2 - INFO - Training dataset size: 6868 +04/30/2025 19:50:09 - lexical_benchmark.train.trainers.gpt2 - INFO - Validation dataset size: 18373 +04/30/2025 19:50:09 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating GPT2 configuration +04/30/2025 19:50:10 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating standard Trainer +04/30/2025 19:50:11 - lexical_benchmark.train.run_train - INFO - Running training of stela_gpt2_EN_02_02 +04/30/2025 20:08:34 - lexical_benchmark.train.run_train - INFO - Loading gpt2 trainer class. +04/30/2025 20:08:47 - datasets - INFO - PyTorch version 2.5.1 available. +04/30/2025 20:08:47 - datasets - INFO - Polars version 1.26.0 available. +04/30/2025 20:08:47 - lexical_benchmark.train.train_params - INFO - Loading params from /lustre/fsn1/projects/rech/hhb/ucx81cx/workspace/lm_benchmark/dev/code/src/lexical_benchmark/train/train-params.toml +04/30/2025 20:08:47 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading char tokenizer +04/30/2025 20:08:47 - lexical_benchmark.train.trainers.gpt2 - INFO - Tokenizer loaded with vocabulary size: 115 +04/30/2025 20:08:47 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading and processing datasets with joined utterances (max_length=1024) +04/30/2025 20:08:47 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/02/02/train.tokenized +04/30/2025 20:08:47 - lexical_benchmark.train.tokenizers - INFO - Read 75727 lines from file +04/30/2025 20:08:54 - lexical_benchmark.train.tokenizers - INFO - Created 6868 joined examples from 75727 original lines +04/30/2025 20:08:55 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/dev/dev.tokenized +04/30/2025 20:08:55 - lexical_benchmark.train.tokenizers - INFO - Read 179806 lines from file +04/30/2025 20:09:11 - lexical_benchmark.train.tokenizers - INFO - Created 18373 joined examples from 179806 original lines +04/30/2025 20:09:14 - lexical_benchmark.train.trainers.gpt2 - INFO - Training dataset size: 6868 +04/30/2025 20:09:14 - lexical_benchmark.train.trainers.gpt2 - INFO - Validation dataset size: 18373 +04/30/2025 20:09:14 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating GPT2 configuration +04/30/2025 20:09:15 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating standard Trainer +04/30/2025 20:09:15 - lexical_benchmark.train.run_train - INFO - Running training of stela_gpt2_EN_02_02 +05/23/2025 11:45:14 - lexical_benchmark.train.run_train - INFO - Loading gpt2 trainer class. +05/23/2025 11:45:26 - datasets - INFO - PyTorch version 2.5.1 available. +05/23/2025 11:45:26 - datasets - INFO - Polars version 1.26.0 available. +05/23/2025 11:45:26 - lexical_benchmark.train.train_params - INFO - Loading params from /lustre/fsn1/projects/rech/hhb/uye44va/workspace/lm_benchmark/prod/code/src/lexical_benchmark/train/train-params.toml +05/23/2025 11:45:26 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading char tokenizer +05/23/2025 11:45:26 - lexical_benchmark.train.trainers.gpt2 - INFO - Tokenizer loaded with vocabulary size: 115 +05/23/2025 11:45:26 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading and processing datasets with joined utterances (max_length=1024) +05/23/2025 11:45:27 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/02/00/train.tokenized +05/23/2025 11:45:27 - lexical_benchmark.train.tokenizers - INFO - Read 70732 lines from file +05/23/2025 11:45:33 - lexical_benchmark.train.tokenizers - INFO - Created 7355 joined examples from 70732 original lines +05/23/2025 11:45:34 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/dev/dev.tokenized +05/23/2025 11:45:34 - lexical_benchmark.train.tokenizers - INFO - Read 179806 lines from file +05/23/2025 11:45:52 - lexical_benchmark.train.tokenizers - INFO - Created 18373 joined examples from 179806 original lines +05/23/2025 11:45:54 - lexical_benchmark.train.trainers.gpt2 - INFO - Training dataset size: 7355 +05/23/2025 11:45:54 - lexical_benchmark.train.trainers.gpt2 - INFO - Validation dataset size: 18373 +05/23/2025 11:45:54 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating GPT2 configuration +05/23/2025 11:45:55 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating standard Trainer +05/23/2025 11:45:56 - lexical_benchmark.train.run_train - INFO - Running training of stela_gpt2_EN_02_00 +05/23/2025 17:33:48 - lexical_benchmark.train.run_train - INFO - Loading gpt2 trainer class. +05/23/2025 17:33:53 - datasets - INFO - PyTorch version 2.5.1 available. +05/23/2025 17:33:53 - datasets - INFO - Polars version 1.26.0 available. +05/23/2025 17:33:53 - lexical_benchmark.train.train_params - INFO - Loading params from /lustre/fsn1/projects/rech/hhb/uye44va/workspace/lm_benchmark/prod/code/src/lexical_benchmark/train/train-params.toml +05/23/2025 17:33:53 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading char tokenizer +05/23/2025 17:33:53 - lexical_benchmark.train.trainers.gpt2 - INFO - Tokenizer loaded with vocabulary size: 115 +05/23/2025 17:33:53 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading and processing datasets with joined utterances (max_length=1024) +05/23/2025 17:33:53 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/02/00/train.tokenized +05/23/2025 17:33:53 - lexical_benchmark.train.tokenizers - INFO - Read 70732 lines from file +05/23/2025 17:33:59 - lexical_benchmark.train.tokenizers - INFO - Created 7355 joined examples from 70732 original lines +05/23/2025 17:34:00 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/dev/dev.tokenized +05/23/2025 17:34:00 - lexical_benchmark.train.tokenizers - INFO - Read 179806 lines from file +05/23/2025 17:34:17 - lexical_benchmark.train.tokenizers - INFO - Created 18373 joined examples from 179806 original lines +05/23/2025 17:34:20 - lexical_benchmark.train.trainers.gpt2 - INFO - Training dataset size: 7355 +05/23/2025 17:34:20 - lexical_benchmark.train.trainers.gpt2 - INFO - Validation dataset size: 18373 +05/23/2025 17:34:20 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating GPT2 configuration +05/23/2025 17:34:20 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating standard Trainer +05/23/2025 17:34:21 - lexical_benchmark.train.run_train - INFO - Running training of stela_gpt2_EN_02_00 +05/23/2025 18:48:00 - lexical_benchmark.train.run_train - INFO - Loading gpt2 trainer class. +05/23/2025 18:48:13 - datasets - INFO - PyTorch version 2.5.1 available. +05/23/2025 18:48:13 - datasets - INFO - Polars version 1.26.0 available. +05/23/2025 18:48:13 - lexical_benchmark.train.train_params - INFO - Loading params from /lustre/fsn1/projects/rech/hhb/uye44va/workspace/lm_benchmark/prod/code/src/lexical_benchmark/train/train-params.toml +05/23/2025 18:48:13 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading char tokenizer +05/23/2025 18:48:14 - lexical_benchmark.train.trainers.gpt2 - INFO - Tokenizer loaded with vocabulary size: 115 +05/23/2025 18:48:14 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading and processing datasets with joined utterances (max_length=1024) +05/23/2025 18:48:14 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/02/00/train.tokenized +05/23/2025 18:48:14 - lexical_benchmark.train.tokenizers - INFO - Read 70732 lines from file +05/23/2025 18:48:21 - lexical_benchmark.train.tokenizers - INFO - Created 7355 joined examples from 70732 original lines +05/23/2025 18:48:22 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/dev/dev.tokenized +05/23/2025 18:48:22 - lexical_benchmark.train.tokenizers - INFO - Read 179806 lines from file +05/23/2025 18:48:43 - lexical_benchmark.train.tokenizers - INFO - Created 18373 joined examples from 179806 original lines +05/23/2025 18:48:46 - lexical_benchmark.train.trainers.gpt2 - INFO - Training dataset size: 7355 +05/23/2025 18:48:46 - lexical_benchmark.train.trainers.gpt2 - INFO - Validation dataset size: 18373 +05/23/2025 18:48:46 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating GPT2 configuration +05/23/2025 18:48:47 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating standard Trainer +05/23/2025 18:48:48 - lexical_benchmark.train.run_train - INFO - Running training of stela_gpt2_EN_02_00 +05/25/2025 22:46:20 - lexical_benchmark.train.run_train - INFO - Loading gpt2 trainer class. +05/25/2025 22:46:32 - datasets - INFO - PyTorch version 2.5.1 available. +05/25/2025 22:46:32 - datasets - INFO - Polars version 1.26.0 available. +05/25/2025 22:46:32 - lexical_benchmark.train.train_params - INFO - Loading params from /lustre/fsn1/projects/rech/hhb/uye44va/workspace/lm_benchmark/prod/code/src/lexical_benchmark/train/train-params.toml +05/25/2025 22:46:32 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading char tokenizer +05/25/2025 22:46:32 - lexical_benchmark.train.trainers.gpt2 - INFO - Tokenizer loaded with vocabulary size: 115 +05/25/2025 22:46:32 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading and processing datasets with joined utterances (max_length=1024) +05/25/2025 22:46:33 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/02/00/train.tokenized +05/25/2025 22:46:33 - lexical_benchmark.train.tokenizers - INFO - Read 70732 lines from file +05/25/2025 22:46:39 - lexical_benchmark.train.tokenizers - INFO - Created 7355 joined examples from 70732 original lines +05/25/2025 22:46:40 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/dev/dev.tokenized +05/25/2025 22:46:40 - lexical_benchmark.train.tokenizers - INFO - Read 179806 lines from file +05/25/2025 22:46:56 - lexical_benchmark.train.tokenizers - INFO - Created 18373 joined examples from 179806 original lines +05/25/2025 22:46:59 - lexical_benchmark.train.trainers.gpt2 - INFO - Training dataset size: 7355 +05/25/2025 22:46:59 - lexical_benchmark.train.trainers.gpt2 - INFO - Validation dataset size: 18373 +05/25/2025 22:46:59 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating GPT2 configuration +05/25/2025 22:47:00 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating standard Trainer +05/25/2025 22:47:00 - lexical_benchmark.train.run_train - INFO - Running training of stela_gpt2_EN_02_00 +05/26/2025 00:32:20 - lexical_benchmark.train.run_train - INFO - Loading gpt2 trainer class. +05/26/2025 00:32:32 - datasets - INFO - PyTorch version 2.5.1 available. +05/26/2025 00:32:32 - datasets - INFO - Polars version 1.26.0 available. +05/26/2025 00:32:32 - lexical_benchmark.train.train_params - INFO - Loading params from /lustre/fsn1/projects/rech/hhb/uye44va/workspace/lm_benchmark/prod/code/src/lexical_benchmark/train/train-params.toml +05/26/2025 00:32:32 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading char tokenizer +05/26/2025 00:32:32 - lexical_benchmark.train.trainers.gpt2 - INFO - Tokenizer loaded with vocabulary size: 115 +05/26/2025 00:32:32 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading and processing datasets with joined utterances (max_length=1024) +05/26/2025 00:32:32 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/02/00/train.tokenized +05/26/2025 00:32:32 - lexical_benchmark.train.tokenizers - INFO - Read 70732 lines from file +05/26/2025 00:32:38 - lexical_benchmark.train.tokenizers - INFO - Created 7355 joined examples from 70732 original lines +05/26/2025 00:32:39 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/dev/dev.tokenized +05/26/2025 00:32:40 - lexical_benchmark.train.tokenizers - INFO - Read 179806 lines from file +05/26/2025 00:32:56 - lexical_benchmark.train.tokenizers - INFO - Created 18373 joined examples from 179806 original lines +05/26/2025 00:32:59 - lexical_benchmark.train.trainers.gpt2 - INFO - Training dataset size: 7355 +05/26/2025 00:32:59 - lexical_benchmark.train.trainers.gpt2 - INFO - Validation dataset size: 18373 +05/26/2025 00:32:59 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating GPT2 configuration +05/26/2025 00:32:59 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating standard Trainer +05/26/2025 00:33:00 - lexical_benchmark.train.run_train - INFO - Running training of stela_gpt2_EN_02_00 +05/26/2025 00:38:26 - lexical_benchmark.train.run_train - INFO - Loading gpt2 trainer class. +05/26/2025 00:38:31 - datasets - INFO - PyTorch version 2.5.1 available. +05/26/2025 00:38:31 - datasets - INFO - Polars version 1.26.0 available. +05/26/2025 00:38:32 - lexical_benchmark.train.train_params - INFO - Loading params from /lustre/fsn1/projects/rech/hhb/uye44va/workspace/lm_benchmark/prod/code/src/lexical_benchmark/train/train-params.toml +05/26/2025 00:38:32 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading char tokenizer +05/26/2025 00:38:32 - lexical_benchmark.train.trainers.gpt2 - INFO - Tokenizer loaded with vocabulary size: 115 +05/26/2025 00:38:32 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading and processing datasets with joined utterances (max_length=1024) +05/26/2025 00:38:32 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/02/00/train.tokenized +05/26/2025 00:38:32 - lexical_benchmark.train.tokenizers - INFO - Read 70732 lines from file +05/26/2025 00:38:39 - lexical_benchmark.train.tokenizers - INFO - Created 7355 joined examples from 70732 original lines +05/26/2025 00:38:40 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/dev/dev.tokenized +05/26/2025 00:38:40 - lexical_benchmark.train.tokenizers - INFO - Read 179806 lines from file +05/26/2025 00:39:01 - lexical_benchmark.train.tokenizers - INFO - Created 18373 joined examples from 179806 original lines +05/26/2025 00:39:04 - lexical_benchmark.train.trainers.gpt2 - INFO - Training dataset size: 7355 +05/26/2025 00:39:04 - lexical_benchmark.train.trainers.gpt2 - INFO - Validation dataset size: 18373 +05/26/2025 00:39:04 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating GPT2 configuration +05/26/2025 00:39:05 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating standard Trainer +05/26/2025 00:39:05 - lexical_benchmark.train.run_train - INFO - Running training of stela_gpt2_EN_02_00 +05/26/2025 01:58:18 - lexical_benchmark.train.run_train - INFO - Training of stela_gpt2_EN_02_00 has completed ! +05/26/2025 01:58:19 - lexical_benchmark.train.run_train - INFO - Model saved to /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/models/stela/EN/02/00/gpt2 diff --git a/stela/EN/02/00/orthographic_gpt2/training_args.bin b/stela/EN/02/00/orthographic_gpt2/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..0db84ba59cfe6602900464247649da70490bd8af --- /dev/null +++ b/stela/EN/02/00/orthographic_gpt2/training_args.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:39bb8aca29f2585ea2be15fb1589265dafc784bca85a412cb9bdf173a6914bd9 +size 5496 diff --git a/stela/EN/02/00/orthographic_lstm/config.json b/stela/EN/02/00/orthographic_lstm/config.json new file mode 100644 index 0000000000000000000000000000000000000000..5ef629887c7fc87d64278c9788e493ac34c60370 --- /dev/null +++ b/stela/EN/02/00/orthographic_lstm/config.json @@ -0,0 +1,13 @@ +{ + "architectures": [ + "LSTMForLanguageModeling" + ], + "dropout": 0.1, + "embedding_dim": 200, + "hidden_size": 1024, + "model_type": "lstm", + "num_layers": 3, + "torch_dtype": "float32", + "transformers_version": "4.49.0", + "vocab_size": 115 +} diff --git a/stela/EN/02/00/orthographic_lstm/generation_config.json b/stela/EN/02/00/orthographic_lstm/generation_config.json new file mode 100644 index 0000000000000000000000000000000000000000..c1818a95e07bdccb83305ce8f9d878755e3c6119 --- /dev/null +++ b/stela/EN/02/00/orthographic_lstm/generation_config.json @@ -0,0 +1,4 @@ +{ + "_from_model_config": true, + "transformers_version": "4.49.0" +} diff --git a/stela/EN/02/00/orthographic_lstm/model.safetensors b/stela/EN/02/00/orthographic_lstm/model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..2be0ff0ff2a68304f66377b940421c22fee44bfe --- /dev/null +++ b/stela/EN/02/00/orthographic_lstm/model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:137808183c3a52f9cb6b14cf7d658518145363e38dfcc76513ccfb477336238a +size 87825996 diff --git a/stela/EN/02/00/orthographic_lstm/training_args.bin b/stela/EN/02/00/orthographic_lstm/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..94f0998a60adbbc8f4f7bbea7e51ca479a2d65f0 --- /dev/null +++ b/stela/EN/02/00/orthographic_lstm/training_args.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:5159f3d84cc34f5414da5a34afbf737c8170bdb9f911121069ae6eccfe7aecf3 +size 5496 diff --git a/stela/EN/02/01/orthographic_gpt2/config.json b/stela/EN/02/01/orthographic_gpt2/config.json new file mode 100644 index 0000000000000000000000000000000000000000..493011df283c5726a4be3aa8f39c84d43991636e --- /dev/null +++ b/stela/EN/02/01/orthographic_gpt2/config.json @@ -0,0 +1,32 @@ +{ + "activation_function": "gelu_new", + "architectures": [ + "GPT2LMHeadModel" + ], + "attn_pdrop": 0.1, + "bos_token_id": 50256, + "embd_pdrop": 0.1, + "eos_token_id": 50256, + "initializer_range": 0.02, + "layer_norm_epsilon": 1e-05, + "model_type": "gpt2", + "n_ctx": 1024, + "n_embd": 768, + "n_head": 12, + "n_inner": null, + "n_layer": 12, + "n_positions": 1024, + "reorder_and_upcast_attn": false, + "resid_pdrop": 0.1, + "scale_attn_by_inverse_layer_idx": false, + "scale_attn_weights": true, + "summary_activation": null, + "summary_first_dropout": 0.1, + "summary_proj_to_labels": true, + "summary_type": "cls_index", + "summary_use_proj": true, + "torch_dtype": "float32", + "transformers_version": "4.49.0", + "use_cache": true, + "vocab_size": 115 +} diff --git a/stela/EN/02/01/orthographic_gpt2/generation_config.json b/stela/EN/02/01/orthographic_gpt2/generation_config.json new file mode 100644 index 0000000000000000000000000000000000000000..bd968f46ce39e754a4603c69f53a2a4a08a13dbb --- /dev/null +++ b/stela/EN/02/01/orthographic_gpt2/generation_config.json @@ -0,0 +1,6 @@ +{ + "_from_model_config": true, + "bos_token_id": 50256, + "eos_token_id": 50256, + "transformers_version": "4.49.0" +} diff --git a/stela/EN/02/01/orthographic_gpt2/model.safetensors b/stela/EN/02/01/orthographic_gpt2/model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..f1700bbd7d855e8f253a92d179ddd3765cc16bb4 --- /dev/null +++ b/stela/EN/02/01/orthographic_gpt2/model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0d7ed791a1e5a03a9991340b14175d0f5e997801465e03f1ea37b1a855a54cde +size 343737984 diff --git a/stela/EN/02/01/orthographic_gpt2/training.logs b/stela/EN/02/01/orthographic_gpt2/training.logs new file mode 100644 index 0000000000000000000000000000000000000000..3474276bc018e935185f3728b0e06f7a4fc2cccf --- /dev/null +++ b/stela/EN/02/01/orthographic_gpt2/training.logs @@ -0,0 +1,20 @@ +06/03/2025 16:47:29 - lexical_benchmark.train.run_train - INFO - Loading gpt2 trainer class. +06/03/2025 16:47:41 - datasets - INFO - PyTorch version 2.5.1 available. +06/03/2025 16:47:41 - datasets - INFO - Polars version 1.26.0 available. +06/03/2025 16:47:41 - lexical_benchmark.train.train_params - INFO - Loading params from /lustre/fsn1/projects/rech/hhb/uye44va/workspace/lm_benchmark/prod/code/src/lexical_benchmark/train/train-params.toml +06/03/2025 16:47:41 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading char tokenizer +06/03/2025 16:47:42 - lexical_benchmark.train.trainers.gpt2 - INFO - Tokenizer loaded with vocabulary size: 115 +06/03/2025 16:47:42 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading and processing datasets with joined utterances (max_length=1024) +06/03/2025 16:47:42 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/02/01/train.tokenized +06/03/2025 16:47:42 - lexical_benchmark.train.tokenizers - INFO - Read 81216 lines from file +06/03/2025 16:47:48 - lexical_benchmark.train.tokenizers - INFO - Created 7336 joined examples from 81216 original lines +06/03/2025 16:47:50 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/dev/dev.tokenized +06/03/2025 16:47:50 - lexical_benchmark.train.tokenizers - INFO - Read 179806 lines from file +06/03/2025 16:48:06 - lexical_benchmark.train.tokenizers - INFO - Created 18373 joined examples from 179806 original lines +06/03/2025 16:48:09 - lexical_benchmark.train.trainers.gpt2 - INFO - Training dataset size: 7336 +06/03/2025 16:48:09 - lexical_benchmark.train.trainers.gpt2 - INFO - Validation dataset size: 18373 +06/03/2025 16:48:09 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating GPT2 configuration +06/03/2025 16:48:09 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating standard Trainer +06/03/2025 16:48:10 - lexical_benchmark.train.run_train - INFO - Running training of stela_gpt2_EN_02_01 +06/03/2025 18:07:10 - lexical_benchmark.train.run_train - INFO - Training of stela_gpt2_EN_02_01 has completed ! +06/03/2025 18:07:10 - lexical_benchmark.train.run_train - INFO - Model saved to /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/models/stela/EN/02/01/gpt2 diff --git a/stela/EN/02/01/orthographic_gpt2/training_args.bin b/stela/EN/02/01/orthographic_gpt2/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..96766964caf610370a6f2189c269176840763c03 --- /dev/null +++ b/stela/EN/02/01/orthographic_gpt2/training_args.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1cf3b9ba3a12bcc1fa8d8b01071964314573853abf87dac3f065102e980d723a +size 5496 diff --git a/stela/EN/02/01/orthographic_lstm/config.json b/stela/EN/02/01/orthographic_lstm/config.json new file mode 100644 index 0000000000000000000000000000000000000000..5ef629887c7fc87d64278c9788e493ac34c60370 --- /dev/null +++ b/stela/EN/02/01/orthographic_lstm/config.json @@ -0,0 +1,13 @@ +{ + "architectures": [ + "LSTMForLanguageModeling" + ], + "dropout": 0.1, + "embedding_dim": 200, + "hidden_size": 1024, + "model_type": "lstm", + "num_layers": 3, + "torch_dtype": "float32", + "transformers_version": "4.49.0", + "vocab_size": 115 +} diff --git a/stela/EN/02/01/orthographic_lstm/generation_config.json b/stela/EN/02/01/orthographic_lstm/generation_config.json new file mode 100644 index 0000000000000000000000000000000000000000..c1818a95e07bdccb83305ce8f9d878755e3c6119 --- /dev/null +++ b/stela/EN/02/01/orthographic_lstm/generation_config.json @@ -0,0 +1,4 @@ +{ + "_from_model_config": true, + "transformers_version": "4.49.0" +} diff --git a/stela/EN/02/01/orthographic_lstm/model.safetensors b/stela/EN/02/01/orthographic_lstm/model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..26596c8d48b7e8a6478687c60d0d4d7dcf05d1f6 --- /dev/null +++ b/stela/EN/02/01/orthographic_lstm/model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7e8d50d6b85ae5f96496c2e4d43a79c8804ef14dd19095c4eb28e6b4e801bc99 +size 87825996 diff --git a/stela/EN/02/01/orthographic_lstm/training.logs b/stela/EN/02/01/orthographic_lstm/training.logs new file mode 100644 index 0000000000000000000000000000000000000000..c6ffb0f58f38b0f9cdf7d573fb6b3e50cf1e7b6d --- /dev/null +++ b/stela/EN/02/01/orthographic_lstm/training.logs @@ -0,0 +1,19 @@ +06/03/2025 16:47:28 - lexical_benchmark.train.run_train - INFO - Loading lstm trainer class. +06/03/2025 16:47:40 - datasets - INFO - PyTorch version 2.5.1 available. +06/03/2025 16:47:40 - datasets - INFO - Polars version 1.26.0 available. +06/03/2025 16:47:40 - lexical_benchmark.train.train_params - INFO - Loading params from /lustre/fsn1/projects/rech/hhb/uye44va/workspace/lm_benchmark/prod/code/src/lexical_benchmark/train/train-params.toml +06/03/2025 16:47:40 - lexical_benchmark.train.trainers.lstm - INFO - Loading char-tokenizer +06/03/2025 16:47:40 - lexical_benchmark.train.trainers.lstm - INFO - Tokenizer loaded with vocabulary size: 115 +06/03/2025 16:47:40 - lexical_benchmark.train.trainers.lstm - INFO - Loading and processing datasets with joined utterances (max_length=1024) +06/03/2025 16:47:40 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/02/01/train.tokenized +06/03/2025 16:47:40 - lexical_benchmark.train.tokenizers - INFO - Read 81216 lines from file +06/03/2025 16:47:47 - lexical_benchmark.train.tokenizers - INFO - Created 7336 joined examples from 81216 original lines +06/03/2025 16:47:48 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/dev/dev.tokenized +06/03/2025 16:47:48 - lexical_benchmark.train.tokenizers - INFO - Read 179806 lines from file +06/03/2025 16:48:05 - lexical_benchmark.train.tokenizers - INFO - Created 18373 joined examples from 179806 original lines +06/03/2025 16:48:08 - lexical_benchmark.train.trainers.lstm - INFO - Training dataset size: 7336 +06/03/2025 16:48:08 - lexical_benchmark.train.trainers.lstm - INFO - Validation dataset size: 18373 +06/03/2025 16:48:08 - lexical_benchmark.train.trainers.lstm - INFO - Loading configurations & initialising LSTM model trainer +06/03/2025 16:48:08 - lexical_benchmark.train.run_train - INFO - Running training of stela_lstm_EN_02_01 +06/03/2025 18:28:17 - lexical_benchmark.train.run_train - INFO - Training of stela_lstm_EN_02_01 has completed ! +06/03/2025 18:28:17 - lexical_benchmark.train.run_train - INFO - Model saved to /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/models/stela/EN/02/01/lstm diff --git a/stela/EN/02/01/orthographic_lstm/training_args.bin b/stela/EN/02/01/orthographic_lstm/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..6b1346850bc15b3edc9c9b544140d10be9327877 --- /dev/null +++ b/stela/EN/02/01/orthographic_lstm/training_args.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e848d1a8b75e427105357d5af70d3173d5282f1ce9439d5668d916e7cc7a6fcf +size 5496 diff --git a/stela/EN/02/02/orthographic_gpt2/training.logs b/stela/EN/02/02/orthographic_gpt2/training.logs new file mode 100644 index 0000000000000000000000000000000000000000..e69de29bb2d1d6434b8b29ae775ad8c2e48c5391 diff --git a/stela/EN/03/00/orthographic_gpt2/config.json b/stela/EN/03/00/orthographic_gpt2/config.json new file mode 100644 index 0000000000000000000000000000000000000000..493011df283c5726a4be3aa8f39c84d43991636e --- /dev/null +++ b/stela/EN/03/00/orthographic_gpt2/config.json @@ -0,0 +1,32 @@ +{ + "activation_function": "gelu_new", + "architectures": [ + "GPT2LMHeadModel" + ], + "attn_pdrop": 0.1, + "bos_token_id": 50256, + "embd_pdrop": 0.1, + "eos_token_id": 50256, + "initializer_range": 0.02, + "layer_norm_epsilon": 1e-05, + "model_type": "gpt2", + "n_ctx": 1024, + "n_embd": 768, + "n_head": 12, + "n_inner": null, + "n_layer": 12, + "n_positions": 1024, + "reorder_and_upcast_attn": false, + "resid_pdrop": 0.1, + "scale_attn_by_inverse_layer_idx": false, + "scale_attn_weights": true, + "summary_activation": null, + "summary_first_dropout": 0.1, + "summary_proj_to_labels": true, + "summary_type": "cls_index", + "summary_use_proj": true, + "torch_dtype": "float32", + "transformers_version": "4.49.0", + "use_cache": true, + "vocab_size": 115 +} diff --git a/stela/EN/03/00/orthographic_gpt2/generation_config.json b/stela/EN/03/00/orthographic_gpt2/generation_config.json new file mode 100644 index 0000000000000000000000000000000000000000..bd968f46ce39e754a4603c69f53a2a4a08a13dbb --- /dev/null +++ b/stela/EN/03/00/orthographic_gpt2/generation_config.json @@ -0,0 +1,6 @@ +{ + "_from_model_config": true, + "bos_token_id": 50256, + "eos_token_id": 50256, + "transformers_version": "4.49.0" +} diff --git a/stela/EN/03/00/orthographic_gpt2/model.safetensors b/stela/EN/03/00/orthographic_gpt2/model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..8163341d562fd9198cf30e9fc214f62459f7235c --- /dev/null +++ b/stela/EN/03/00/orthographic_gpt2/model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7355a6f808256130c7d5938465ce3fe27ac9c45d84ad114a87a5fa6afc55fe23 +size 343737984 diff --git a/stela/EN/03/00/orthographic_gpt2/training.logs b/stela/EN/03/00/orthographic_gpt2/training.logs new file mode 100644 index 0000000000000000000000000000000000000000..8ed2e18069fca3a02f0102f55cb82eb863a5912c --- /dev/null +++ b/stela/EN/03/00/orthographic_gpt2/training.logs @@ -0,0 +1,224 @@ +04/30/2025 18:19:34 - lexical_benchmark.train.run_train - INFO - Loading gpt2 trainer class. +04/30/2025 18:19:46 - datasets - INFO - PyTorch version 2.5.1 available. +04/30/2025 18:19:46 - datasets - INFO - Polars version 1.26.0 available. +04/30/2025 18:19:47 - lexical_benchmark.train.train_params - INFO - Loading params from /lustre/fsn1/projects/rech/hhb/uye44va/workspace/lm_benchmark/prod/code/src/lexical_benchmark/train/train-params.toml +04/30/2025 18:19:47 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading char tokenizer +04/30/2025 18:19:47 - lexical_benchmark.train.trainers.gpt2 - INFO - Tokenizer loaded with vocabulary size: 115 +04/30/2025 18:19:47 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading and processing datasets with joined utterances (max_length=1024) +04/30/2025 18:19:47 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/03/03/train.tokenized +04/30/2025 18:19:47 - lexical_benchmark.train.tokenizers - INFO - Read 111140 lines from file +04/30/2025 18:19:57 - lexical_benchmark.train.tokenizers - INFO - Created 11046 joined examples from 111140 original lines +04/30/2025 18:19:59 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/dev/dev.tokenized +04/30/2025 18:19:59 - lexical_benchmark.train.tokenizers - INFO - Read 179806 lines from file +04/30/2025 18:20:15 - lexical_benchmark.train.tokenizers - INFO - Created 18373 joined examples from 179806 original lines +04/30/2025 18:20:18 - lexical_benchmark.train.trainers.gpt2 - INFO - Training dataset size: 11046 +04/30/2025 18:20:18 - lexical_benchmark.train.trainers.gpt2 - INFO - Validation dataset size: 18373 +04/30/2025 18:20:18 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating GPT2 configuration +04/30/2025 18:20:19 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating standard Trainer +04/30/2025 18:20:19 - lexical_benchmark.train.run_train - INFO - Running training of stela_gpt2_EN_03_03 +04/30/2025 19:27:14 - urllib3.connectionpool - WARNING - Retrying (Retry(total=2, connect=None, read=None, redirect=None, status=None)) after connection broken by 'NewConnectionError(': Failed to establish a new connection: [Errno 101] Network is unreachable')': /api/4504800232407040/envelope/ +04/30/2025 19:27:14 - urllib3.connectionpool - WARNING - Retrying (Retry(total=1, connect=None, read=None, redirect=None, status=None)) after connection broken by 'NewConnectionError(': Failed to establish a new connection: [Errno 101] Network is unreachable')': /api/4504800232407040/envelope/ +04/30/2025 19:27:14 - urllib3.connectionpool - WARNING - Retrying (Retry(total=0, connect=None, read=None, redirect=None, status=None)) after connection broken by 'NewConnectionError(': Failed to establish a new connection: [Errno 101] Network is unreachable')': /api/4504800232407040/envelope/ +04/30/2025 19:27:14 - urllib3.connectionpool - WARNING - Retrying (Retry(total=2, connect=None, read=None, redirect=None, status=None)) after connection broken by 'NewConnectionError(': Failed to establish a new connection: [Errno 101] Network is unreachable')': /api/4504800232407040/envelope/ +04/30/2025 19:27:14 - urllib3.connectionpool - WARNING - Retrying (Retry(total=1, connect=None, read=None, redirect=None, status=None)) after connection broken by 'NewConnectionError(': Failed to establish a new connection: [Errno 101] Network is unreachable')': /api/4504800232407040/envelope/ +04/30/2025 19:27:14 - urllib3.connectionpool - WARNING - Retrying (Retry(total=0, connect=None, read=None, redirect=None, status=None)) after connection broken by 'NewConnectionError(': Failed to establish a new connection: [Errno 101] Network is unreachable')': /api/4504800232407040/envelope/ +04/30/2025 19:51:33 - lexical_benchmark.train.run_train - INFO - Loading gpt2 trainer class. +04/30/2025 19:51:46 - datasets - INFO - PyTorch version 2.5.1 available. +04/30/2025 19:51:46 - datasets - INFO - Polars version 1.26.0 available. +04/30/2025 19:51:46 - lexical_benchmark.train.train_params - INFO - Loading params from /lustre/fsn1/projects/rech/hhb/ucx81cx/workspace/lm_benchmark/dev/code/src/lexical_benchmark/train/train-params.toml +04/30/2025 19:51:46 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading char tokenizer +04/30/2025 19:51:46 - lexical_benchmark.train.trainers.gpt2 - INFO - Tokenizer loaded with vocabulary size: 115 +04/30/2025 19:51:46 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading and processing datasets with joined utterances (max_length=1024) +04/30/2025 19:51:46 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/03/03/train.tokenized +04/30/2025 19:51:46 - lexical_benchmark.train.tokenizers - INFO - Read 111140 lines from file +04/30/2025 19:51:57 - lexical_benchmark.train.tokenizers - INFO - Created 11046 joined examples from 111140 original lines +04/30/2025 19:51:59 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/dev/dev.tokenized +04/30/2025 19:51:59 - lexical_benchmark.train.tokenizers - INFO - Read 179806 lines from file +04/30/2025 19:52:16 - lexical_benchmark.train.tokenizers - INFO - Created 18373 joined examples from 179806 original lines +04/30/2025 19:52:18 - lexical_benchmark.train.trainers.gpt2 - INFO - Training dataset size: 11046 +04/30/2025 19:52:18 - lexical_benchmark.train.trainers.gpt2 - INFO - Validation dataset size: 18373 +04/30/2025 19:52:18 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating GPT2 configuration +04/30/2025 19:52:19 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating standard Trainer +04/30/2025 19:52:20 - lexical_benchmark.train.run_train - INFO - Running training of stela_gpt2_EN_03_03 +04/30/2025 20:08:34 - lexical_benchmark.train.run_train - INFO - Loading gpt2 trainer class. +04/30/2025 20:08:47 - datasets - INFO - PyTorch version 2.5.1 available. +04/30/2025 20:08:47 - datasets - INFO - Polars version 1.26.0 available. +04/30/2025 20:08:47 - lexical_benchmark.train.train_params - INFO - Loading params from /lustre/fsn1/projects/rech/hhb/ucx81cx/workspace/lm_benchmark/dev/code/src/lexical_benchmark/train/train-params.toml +04/30/2025 20:08:47 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading char tokenizer +04/30/2025 20:08:47 - lexical_benchmark.train.trainers.gpt2 - INFO - Tokenizer loaded with vocabulary size: 115 +04/30/2025 20:08:47 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading and processing datasets with joined utterances (max_length=1024) +04/30/2025 20:08:47 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/03/03/train.tokenized +04/30/2025 20:08:48 - lexical_benchmark.train.tokenizers - INFO - Read 111140 lines from file +04/30/2025 20:08:57 - lexical_benchmark.train.tokenizers - INFO - Created 11046 joined examples from 111140 original lines +04/30/2025 20:08:59 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/dev/dev.tokenized +04/30/2025 20:08:59 - lexical_benchmark.train.tokenizers - INFO - Read 179806 lines from file +04/30/2025 20:09:15 - lexical_benchmark.train.tokenizers - INFO - Created 18373 joined examples from 179806 original lines +04/30/2025 20:09:18 - lexical_benchmark.train.trainers.gpt2 - INFO - Training dataset size: 11046 +04/30/2025 20:09:18 - lexical_benchmark.train.trainers.gpt2 - INFO - Validation dataset size: 18373 +04/30/2025 20:09:18 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating GPT2 configuration +04/30/2025 20:09:19 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating standard Trainer +04/30/2025 20:09:19 - lexical_benchmark.train.run_train - INFO - Running training of stela_gpt2_EN_03_03 +05/23/2025 11:45:11 - lexical_benchmark.train.run_train - INFO - Loading gpt2 trainer class. +05/23/2025 11:45:19 - datasets - INFO - PyTorch version 2.5.1 available. +05/23/2025 11:45:19 - datasets - INFO - Polars version 1.26.0 available. +05/23/2025 11:45:19 - lexical_benchmark.train.train_params - INFO - Loading params from /lustre/fsn1/projects/rech/hhb/uye44va/workspace/lm_benchmark/prod/code/src/lexical_benchmark/train/train-params.toml +05/23/2025 11:45:19 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading char tokenizer +05/23/2025 11:45:19 - lexical_benchmark.train.trainers.gpt2 - INFO - Tokenizer loaded with vocabulary size: 115 +05/23/2025 11:45:19 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading and processing datasets with joined utterances (max_length=1024) +05/23/2025 11:45:20 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/03/00/train.tokenized +05/23/2025 11:45:20 - lexical_benchmark.train.tokenizers - INFO - Read 110192 lines from file +05/23/2025 11:45:29 - lexical_benchmark.train.tokenizers - INFO - Created 11027 joined examples from 110192 original lines +05/23/2025 11:45:31 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/dev/dev.tokenized +05/23/2025 11:45:31 - lexical_benchmark.train.tokenizers - INFO - Read 179806 lines from file +05/23/2025 11:45:47 - lexical_benchmark.train.tokenizers - INFO - Created 18373 joined examples from 179806 original lines +05/23/2025 11:45:49 - lexical_benchmark.train.trainers.gpt2 - INFO - Training dataset size: 11027 +05/23/2025 11:45:49 - lexical_benchmark.train.trainers.gpt2 - INFO - Validation dataset size: 18373 +05/23/2025 11:45:49 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating GPT2 configuration +05/23/2025 11:45:50 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating standard Trainer +05/23/2025 11:45:51 - lexical_benchmark.train.run_train - INFO - Running training of stela_gpt2_EN_03_00 +05/23/2025 17:33:52 - lexical_benchmark.train.run_train - INFO - Loading gpt2 trainer class. +05/23/2025 17:34:03 - datasets - INFO - PyTorch version 2.5.1 available. +05/23/2025 17:34:03 - datasets - INFO - Polars version 1.26.0 available. +05/23/2025 17:34:04 - lexical_benchmark.train.train_params - INFO - Loading params from /lustre/fsn1/projects/rech/hhb/uye44va/workspace/lm_benchmark/prod/code/src/lexical_benchmark/train/train-params.toml +05/23/2025 17:34:04 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading char tokenizer +05/23/2025 17:34:04 - lexical_benchmark.train.trainers.gpt2 - INFO - Tokenizer loaded with vocabulary size: 115 +05/23/2025 17:34:04 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading and processing datasets with joined utterances (max_length=1024) +05/23/2025 17:34:04 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/03/00/train.tokenized +05/23/2025 17:34:04 - lexical_benchmark.train.tokenizers - INFO - Read 110192 lines from file +05/23/2025 17:34:14 - lexical_benchmark.train.tokenizers - INFO - Created 11027 joined examples from 110192 original lines +05/23/2025 17:34:15 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/dev/dev.tokenized +05/23/2025 17:34:15 - lexical_benchmark.train.tokenizers - INFO - Read 179806 lines from file +05/23/2025 17:34:32 - lexical_benchmark.train.tokenizers - INFO - Created 18373 joined examples from 179806 original lines +05/23/2025 17:34:34 - lexical_benchmark.train.trainers.gpt2 - INFO - Training dataset size: 11027 +05/23/2025 17:34:34 - lexical_benchmark.train.trainers.gpt2 - INFO - Validation dataset size: 18373 +05/23/2025 17:34:34 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating GPT2 configuration +05/23/2025 17:34:35 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating standard Trainer +05/23/2025 17:34:35 - lexical_benchmark.train.run_train - INFO - Running training of stela_gpt2_EN_03_00 +05/23/2025 18:48:00 - lexical_benchmark.train.run_train - INFO - Loading gpt2 trainer class. +05/23/2025 18:48:12 - datasets - INFO - PyTorch version 2.5.1 available. +05/23/2025 18:48:12 - datasets - INFO - Polars version 1.26.0 available. +05/23/2025 18:48:13 - lexical_benchmark.train.train_params - INFO - Loading params from /lustre/fsn1/projects/rech/hhb/uye44va/workspace/lm_benchmark/prod/code/src/lexical_benchmark/train/train-params.toml +05/23/2025 18:48:13 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading char tokenizer +05/23/2025 18:48:13 - lexical_benchmark.train.trainers.gpt2 - INFO - Tokenizer loaded with vocabulary size: 115 +05/23/2025 18:48:13 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading and processing datasets with joined utterances (max_length=1024) +05/23/2025 18:48:13 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/03/00/train.tokenized +05/23/2025 18:48:13 - lexical_benchmark.train.tokenizers - INFO - Read 110192 lines from file +05/23/2025 18:48:24 - lexical_benchmark.train.tokenizers - INFO - Created 11027 joined examples from 110192 original lines +05/23/2025 18:48:26 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/dev/dev.tokenized +05/23/2025 18:48:26 - lexical_benchmark.train.tokenizers - INFO - Read 179806 lines from file +05/23/2025 18:48:44 - lexical_benchmark.train.tokenizers - INFO - Created 18373 joined examples from 179806 original lines +05/23/2025 18:48:47 - lexical_benchmark.train.trainers.gpt2 - INFO - Training dataset size: 11027 +05/23/2025 18:48:47 - lexical_benchmark.train.trainers.gpt2 - INFO - Validation dataset size: 18373 +05/23/2025 18:48:47 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating GPT2 configuration +05/23/2025 18:48:48 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating standard Trainer +05/23/2025 18:48:49 - lexical_benchmark.train.run_train - INFO - Running training of stela_gpt2_EN_03_00 +05/25/2025 18:09:24 - lexical_benchmark.train.run_train - INFO - Loading gpt2 trainer class. +05/25/2025 18:09:36 - datasets - INFO - PyTorch version 2.5.1 available. +05/25/2025 18:09:36 - datasets - INFO - Polars version 1.26.0 available. +05/25/2025 18:09:36 - lexical_benchmark.train.train_params - INFO - Loading params from /lustre/fsn1/projects/rech/hhb/uye44va/workspace/lm_benchmark/prod/code/src/lexical_benchmark/train/train-params.toml +05/25/2025 18:09:36 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading char tokenizer +05/25/2025 18:09:36 - lexical_benchmark.train.trainers.gpt2 - INFO - Tokenizer loaded with vocabulary size: 115 +05/25/2025 18:09:36 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading and processing datasets with joined utterances (max_length=1024) +05/25/2025 18:09:36 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/03/00/train.tokenized +05/25/2025 18:09:36 - lexical_benchmark.train.tokenizers - INFO - Read 110192 lines from file +05/25/2025 18:09:46 - lexical_benchmark.train.tokenizers - INFO - Created 11027 joined examples from 110192 original lines +05/25/2025 18:09:48 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/dev/dev.tokenized +05/25/2025 18:09:48 - lexical_benchmark.train.tokenizers - INFO - Read 179806 lines from file +05/25/2025 18:10:04 - lexical_benchmark.train.tokenizers - INFO - Created 18373 joined examples from 179806 original lines +05/25/2025 18:10:07 - lexical_benchmark.train.trainers.gpt2 - INFO - Training dataset size: 11027 +05/25/2025 18:10:07 - lexical_benchmark.train.trainers.gpt2 - INFO - Validation dataset size: 18373 +05/25/2025 18:10:07 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating GPT2 configuration +05/25/2025 18:10:08 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating standard Trainer +05/25/2025 18:10:08 - lexical_benchmark.train.run_train - INFO - Running training of stela_gpt2_EN_03_00 +05/25/2025 18:17:30 - lexical_benchmark.train.run_train - INFO - Loading gpt2 trainer class. +05/25/2025 18:17:34 - datasets - INFO - PyTorch version 2.5.1 available. +05/25/2025 18:17:34 - datasets - INFO - Polars version 1.26.0 available. +05/25/2025 18:17:34 - lexical_benchmark.train.train_params - INFO - Loading params from /lustre/fsn1/projects/rech/hhb/uye44va/workspace/lm_benchmark/prod/code/src/lexical_benchmark/train/train-params.toml +05/25/2025 18:17:34 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading char tokenizer +05/25/2025 18:17:34 - lexical_benchmark.train.trainers.gpt2 - INFO - Tokenizer loaded with vocabulary size: 115 +05/25/2025 18:17:34 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading and processing datasets with joined utterances (max_length=1024) +05/25/2025 18:17:34 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/03/00/train.tokenized +05/25/2025 18:17:34 - lexical_benchmark.train.tokenizers - INFO - Read 110192 lines from file +05/25/2025 18:17:44 - lexical_benchmark.train.tokenizers - INFO - Created 11027 joined examples from 110192 original lines +05/25/2025 18:17:46 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/dev/dev.tokenized +05/25/2025 18:17:46 - lexical_benchmark.train.tokenizers - INFO - Read 179806 lines from file +05/25/2025 18:18:02 - lexical_benchmark.train.tokenizers - INFO - Created 18373 joined examples from 179806 original lines +05/25/2025 18:18:05 - lexical_benchmark.train.trainers.gpt2 - INFO - Training dataset size: 11027 +05/25/2025 18:18:05 - lexical_benchmark.train.trainers.gpt2 - INFO - Validation dataset size: 18373 +05/25/2025 18:18:05 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating GPT2 configuration +05/25/2025 18:18:05 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating standard Trainer +05/25/2025 18:18:06 - lexical_benchmark.train.run_train - INFO - Running training of stela_gpt2_EN_03_00 +05/25/2025 18:18:42 - lexical_benchmark.train.run_train - INFO - Loading gpt2 trainer class. +05/25/2025 18:18:46 - datasets - INFO - PyTorch version 2.5.1 available. +05/25/2025 18:18:46 - datasets - INFO - Polars version 1.26.0 available. +05/25/2025 18:18:47 - lexical_benchmark.train.train_params - INFO - Loading params from /lustre/fsn1/projects/rech/hhb/uye44va/workspace/lm_benchmark/prod/code/src/lexical_benchmark/train/train-params.toml +05/25/2025 18:18:47 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading char tokenizer +05/25/2025 18:18:47 - lexical_benchmark.train.trainers.gpt2 - INFO - Tokenizer loaded with vocabulary size: 115 +05/25/2025 18:18:47 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading and processing datasets with joined utterances (max_length=1024) +05/25/2025 18:18:47 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/03/00/train.tokenized +05/25/2025 18:18:47 - lexical_benchmark.train.tokenizers - INFO - Read 110192 lines from file +05/25/2025 18:18:57 - lexical_benchmark.train.tokenizers - INFO - Created 11027 joined examples from 110192 original lines +05/25/2025 18:18:58 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/dev/dev.tokenized +05/25/2025 18:18:58 - lexical_benchmark.train.tokenizers - INFO - Read 179806 lines from file +05/25/2025 18:19:15 - lexical_benchmark.train.tokenizers - INFO - Created 18373 joined examples from 179806 original lines +05/25/2025 18:19:17 - lexical_benchmark.train.trainers.gpt2 - INFO - Training dataset size: 11027 +05/25/2025 18:19:17 - lexical_benchmark.train.trainers.gpt2 - INFO - Validation dataset size: 18373 +05/25/2025 18:19:17 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating GPT2 configuration +05/25/2025 18:19:18 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating standard Trainer +05/25/2025 18:19:18 - lexical_benchmark.train.run_train - INFO - Running training of stela_gpt2_EN_03_00 +05/25/2025 22:46:20 - lexical_benchmark.train.run_train - INFO - Loading gpt2 trainer class. +05/25/2025 22:46:32 - datasets - INFO - PyTorch version 2.5.1 available. +05/25/2025 22:46:32 - datasets - INFO - Polars version 1.26.0 available. +05/25/2025 22:46:33 - lexical_benchmark.train.train_params - INFO - Loading params from /lustre/fsn1/projects/rech/hhb/uye44va/workspace/lm_benchmark/prod/code/src/lexical_benchmark/train/train-params.toml +05/25/2025 22:46:33 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading char tokenizer +05/25/2025 22:46:33 - lexical_benchmark.train.trainers.gpt2 - INFO - Tokenizer loaded with vocabulary size: 115 +05/25/2025 22:46:33 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading and processing datasets with joined utterances (max_length=1024) +05/25/2025 22:46:33 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/03/00/train.tokenized +05/25/2025 22:46:33 - lexical_benchmark.train.tokenizers - INFO - Read 110192 lines from file +05/25/2025 22:46:43 - lexical_benchmark.train.tokenizers - INFO - Created 11027 joined examples from 110192 original lines +05/25/2025 22:46:44 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/dev/dev.tokenized +05/25/2025 22:46:44 - lexical_benchmark.train.tokenizers - INFO - Read 179806 lines from file +05/25/2025 22:47:01 - lexical_benchmark.train.tokenizers - INFO - Created 18373 joined examples from 179806 original lines +05/25/2025 22:47:03 - lexical_benchmark.train.trainers.gpt2 - INFO - Training dataset size: 11027 +05/25/2025 22:47:03 - lexical_benchmark.train.trainers.gpt2 - INFO - Validation dataset size: 18373 +05/25/2025 22:47:03 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating GPT2 configuration +05/25/2025 22:47:04 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating standard Trainer +05/25/2025 22:47:04 - lexical_benchmark.train.run_train - INFO - Running training of stela_gpt2_EN_03_00 +05/26/2025 00:32:20 - lexical_benchmark.train.run_train - INFO - Loading gpt2 trainer class. +05/26/2025 00:32:33 - datasets - INFO - PyTorch version 2.5.1 available. +05/26/2025 00:32:33 - datasets - INFO - Polars version 1.26.0 available. +05/26/2025 00:32:34 - lexical_benchmark.train.train_params - INFO - Loading params from /lustre/fsn1/projects/rech/hhb/uye44va/workspace/lm_benchmark/prod/code/src/lexical_benchmark/train/train-params.toml +05/26/2025 00:32:34 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading char tokenizer +05/26/2025 00:32:34 - lexical_benchmark.train.trainers.gpt2 - INFO - Tokenizer loaded with vocabulary size: 115 +05/26/2025 00:32:34 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading and processing datasets with joined utterances (max_length=1024) +05/26/2025 00:32:34 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/03/00/train.tokenized +05/26/2025 00:32:34 - lexical_benchmark.train.tokenizers - INFO - Read 110192 lines from file +05/26/2025 00:32:43 - lexical_benchmark.train.tokenizers - INFO - Created 11027 joined examples from 110192 original lines +05/26/2025 00:32:45 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/dev/dev.tokenized +05/26/2025 00:32:45 - lexical_benchmark.train.tokenizers - INFO - Read 179806 lines from file +05/26/2025 00:33:02 - lexical_benchmark.train.tokenizers - INFO - Created 18373 joined examples from 179806 original lines +05/26/2025 00:33:04 - lexical_benchmark.train.trainers.gpt2 - INFO - Training dataset size: 11027 +05/26/2025 00:33:04 - lexical_benchmark.train.trainers.gpt2 - INFO - Validation dataset size: 18373 +05/26/2025 00:33:04 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating GPT2 configuration +05/26/2025 00:33:05 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating standard Trainer +05/26/2025 00:33:06 - lexical_benchmark.train.run_train - INFO - Running training of stela_gpt2_EN_03_00 +05/26/2025 00:38:30 - lexical_benchmark.train.run_train - INFO - Loading gpt2 trainer class. +05/26/2025 00:38:41 - datasets - INFO - PyTorch version 2.5.1 available. +05/26/2025 00:38:41 - datasets - INFO - Polars version 1.26.0 available. +05/26/2025 00:38:42 - lexical_benchmark.train.train_params - INFO - Loading params from /lustre/fsn1/projects/rech/hhb/uye44va/workspace/lm_benchmark/prod/code/src/lexical_benchmark/train/train-params.toml +05/26/2025 00:38:42 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading char tokenizer +05/26/2025 00:38:42 - lexical_benchmark.train.trainers.gpt2 - INFO - Tokenizer loaded with vocabulary size: 115 +05/26/2025 00:38:42 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading and processing datasets with joined utterances (max_length=1024) +05/26/2025 00:38:42 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/03/00/train.tokenized +05/26/2025 00:38:42 - lexical_benchmark.train.tokenizers - INFO - Read 110192 lines from file +05/26/2025 00:38:51 - lexical_benchmark.train.tokenizers - INFO - Created 11027 joined examples from 110192 original lines +05/26/2025 00:38:53 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/dev/dev.tokenized +05/26/2025 00:38:53 - lexical_benchmark.train.tokenizers - INFO - Read 179806 lines from file +05/26/2025 00:39:09 - lexical_benchmark.train.tokenizers - INFO - Created 18373 joined examples from 179806 original lines +05/26/2025 00:39:12 - lexical_benchmark.train.trainers.gpt2 - INFO - Training dataset size: 11027 +05/26/2025 00:39:12 - lexical_benchmark.train.trainers.gpt2 - INFO - Validation dataset size: 18373 +05/26/2025 00:39:12 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating GPT2 configuration +05/26/2025 00:39:13 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating standard Trainer +05/26/2025 00:39:13 - lexical_benchmark.train.run_train - INFO - Running training of stela_gpt2_EN_03_00 +05/26/2025 02:14:01 - lexical_benchmark.train.run_train - INFO - Training of stela_gpt2_EN_03_00 has completed ! +05/26/2025 02:14:01 - lexical_benchmark.train.run_train - INFO - Model saved to /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/models/stela/EN/03/00/gpt2 diff --git a/stela/EN/03/00/orthographic_gpt2/training_args.bin b/stela/EN/03/00/orthographic_gpt2/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..52447208b02c828faf5bdb63c4db979dc8803ba2 --- /dev/null +++ b/stela/EN/03/00/orthographic_gpt2/training_args.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:cd1486da9066f02b0fecfeab382314183b925a8e36403685e5998ebcb5b21c13 +size 5496 diff --git a/stela/EN/03/00/orthographic_lstm/config.json b/stela/EN/03/00/orthographic_lstm/config.json new file mode 100644 index 0000000000000000000000000000000000000000..5ef629887c7fc87d64278c9788e493ac34c60370 --- /dev/null +++ b/stela/EN/03/00/orthographic_lstm/config.json @@ -0,0 +1,13 @@ +{ + "architectures": [ + "LSTMForLanguageModeling" + ], + "dropout": 0.1, + "embedding_dim": 200, + "hidden_size": 1024, + "model_type": "lstm", + "num_layers": 3, + "torch_dtype": "float32", + "transformers_version": "4.49.0", + "vocab_size": 115 +} diff --git a/stela/EN/03/00/orthographic_lstm/generation_config.json b/stela/EN/03/00/orthographic_lstm/generation_config.json new file mode 100644 index 0000000000000000000000000000000000000000..c1818a95e07bdccb83305ce8f9d878755e3c6119 --- /dev/null +++ b/stela/EN/03/00/orthographic_lstm/generation_config.json @@ -0,0 +1,4 @@ +{ + "_from_model_config": true, + "transformers_version": "4.49.0" +} diff --git a/stela/EN/03/00/orthographic_lstm/model.safetensors b/stela/EN/03/00/orthographic_lstm/model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..30eb806e5fef16896153c7e09f605f4afb70f04a --- /dev/null +++ b/stela/EN/03/00/orthographic_lstm/model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:be0b244a956aef7a58f9a028090c2754f8ab219bae41ba555f103db2e0ba54d2 +size 87825996 diff --git a/stela/EN/03/00/orthographic_lstm/training_args.bin b/stela/EN/03/00/orthographic_lstm/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..62b0aff907a7b5fbca3ee4bce9ea8918ebeef1f8 --- /dev/null +++ b/stela/EN/03/00/orthographic_lstm/training_args.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:cc59d301f621c5c42eb768dc6c9f48eb90cf2f2278252afc7549b1fa0abba37c +size 5496 diff --git a/stela/EN/03/01/orthographic_gpt2/config.json b/stela/EN/03/01/orthographic_gpt2/config.json new file mode 100644 index 0000000000000000000000000000000000000000..493011df283c5726a4be3aa8f39c84d43991636e --- /dev/null +++ b/stela/EN/03/01/orthographic_gpt2/config.json @@ -0,0 +1,32 @@ +{ + "activation_function": "gelu_new", + "architectures": [ + "GPT2LMHeadModel" + ], + "attn_pdrop": 0.1, + "bos_token_id": 50256, + "embd_pdrop": 0.1, + "eos_token_id": 50256, + "initializer_range": 0.02, + "layer_norm_epsilon": 1e-05, + "model_type": "gpt2", + "n_ctx": 1024, + "n_embd": 768, + "n_head": 12, + "n_inner": null, + "n_layer": 12, + "n_positions": 1024, + "reorder_and_upcast_attn": false, + "resid_pdrop": 0.1, + "scale_attn_by_inverse_layer_idx": false, + "scale_attn_weights": true, + "summary_activation": null, + "summary_first_dropout": 0.1, + "summary_proj_to_labels": true, + "summary_type": "cls_index", + "summary_use_proj": true, + "torch_dtype": "float32", + "transformers_version": "4.49.0", + "use_cache": true, + "vocab_size": 115 +} diff --git a/stela/EN/03/01/orthographic_gpt2/generation_config.json b/stela/EN/03/01/orthographic_gpt2/generation_config.json new file mode 100644 index 0000000000000000000000000000000000000000..bd968f46ce39e754a4603c69f53a2a4a08a13dbb --- /dev/null +++ b/stela/EN/03/01/orthographic_gpt2/generation_config.json @@ -0,0 +1,6 @@ +{ + "_from_model_config": true, + "bos_token_id": 50256, + "eos_token_id": 50256, + "transformers_version": "4.49.0" +} diff --git a/stela/EN/03/01/orthographic_gpt2/model.safetensors b/stela/EN/03/01/orthographic_gpt2/model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..39e353e38073f7fe3273e1a66f0656ffd97af697 --- /dev/null +++ b/stela/EN/03/01/orthographic_gpt2/model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:5b30593cbcfb57c3c5982042e2f08c17fc8a35b57e2ae8144c36be622d6fa3a7 +size 343737984 diff --git a/stela/EN/03/01/orthographic_gpt2/training.logs b/stela/EN/03/01/orthographic_gpt2/training.logs new file mode 100644 index 0000000000000000000000000000000000000000..9bf52aaa7315563d5d809d0a6072e033a2e9f726 --- /dev/null +++ b/stela/EN/03/01/orthographic_gpt2/training.logs @@ -0,0 +1,20 @@ +06/03/2025 16:47:29 - lexical_benchmark.train.run_train - INFO - Loading gpt2 trainer class. +06/03/2025 16:47:41 - datasets - INFO - PyTorch version 2.5.1 available. +06/03/2025 16:47:41 - datasets - INFO - Polars version 1.26.0 available. +06/03/2025 16:47:41 - lexical_benchmark.train.train_params - INFO - Loading params from /lustre/fsn1/projects/rech/hhb/uye44va/workspace/lm_benchmark/prod/code/src/lexical_benchmark/train/train-params.toml +06/03/2025 16:47:41 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading char tokenizer +06/03/2025 16:47:42 - lexical_benchmark.train.trainers.gpt2 - INFO - Tokenizer loaded with vocabulary size: 115 +06/03/2025 16:47:42 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading and processing datasets with joined utterances (max_length=1024) +06/03/2025 16:47:42 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/03/01/train.tokenized +06/03/2025 16:47:42 - lexical_benchmark.train.tokenizers - INFO - Read 116181 lines from file +06/03/2025 16:47:52 - lexical_benchmark.train.tokenizers - INFO - Created 11028 joined examples from 116181 original lines +06/03/2025 16:47:54 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/dev/dev.tokenized +06/03/2025 16:47:54 - lexical_benchmark.train.tokenizers - INFO - Read 179806 lines from file +06/03/2025 16:48:10 - lexical_benchmark.train.tokenizers - INFO - Created 18373 joined examples from 179806 original lines +06/03/2025 16:48:13 - lexical_benchmark.train.trainers.gpt2 - INFO - Training dataset size: 11028 +06/03/2025 16:48:13 - lexical_benchmark.train.trainers.gpt2 - INFO - Validation dataset size: 18373 +06/03/2025 16:48:13 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating GPT2 configuration +06/03/2025 16:48:13 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating standard Trainer +06/03/2025 16:48:14 - lexical_benchmark.train.run_train - INFO - Running training of stela_gpt2_EN_03_01 +06/03/2025 18:23:07 - lexical_benchmark.train.run_train - INFO - Training of stela_gpt2_EN_03_01 has completed ! +06/03/2025 18:23:07 - lexical_benchmark.train.run_train - INFO - Model saved to /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/models/stela/EN/03/01/gpt2 diff --git a/stela/EN/03/01/orthographic_gpt2/training_args.bin b/stela/EN/03/01/orthographic_gpt2/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..c2f38c18eb71e7d4398516f0f2c0ce0be5977fc7 --- /dev/null +++ b/stela/EN/03/01/orthographic_gpt2/training_args.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1e60338a81d4550caabcadc3ac9c028fc9bd7ed0659b9558b6286e6e98c2dd76 +size 5496 diff --git a/stela/EN/03/01/orthographic_lstm/config.json b/stela/EN/03/01/orthographic_lstm/config.json new file mode 100644 index 0000000000000000000000000000000000000000..5ef629887c7fc87d64278c9788e493ac34c60370 --- /dev/null +++ b/stela/EN/03/01/orthographic_lstm/config.json @@ -0,0 +1,13 @@ +{ + "architectures": [ + "LSTMForLanguageModeling" + ], + "dropout": 0.1, + "embedding_dim": 200, + "hidden_size": 1024, + "model_type": "lstm", + "num_layers": 3, + "torch_dtype": "float32", + "transformers_version": "4.49.0", + "vocab_size": 115 +} diff --git a/stela/EN/03/01/orthographic_lstm/generation_config.json b/stela/EN/03/01/orthographic_lstm/generation_config.json new file mode 100644 index 0000000000000000000000000000000000000000..c1818a95e07bdccb83305ce8f9d878755e3c6119 --- /dev/null +++ b/stela/EN/03/01/orthographic_lstm/generation_config.json @@ -0,0 +1,4 @@ +{ + "_from_model_config": true, + "transformers_version": "4.49.0" +} diff --git a/stela/EN/03/01/orthographic_lstm/model.safetensors b/stela/EN/03/01/orthographic_lstm/model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..a956667da254bf928c02dc9b2d2ddc139bcb016c --- /dev/null +++ b/stela/EN/03/01/orthographic_lstm/model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:fe80edd2af741a6708157290a0cbf892504c32a5ff8a7406b9b52044fbf7fc45 +size 87825996 diff --git a/stela/EN/03/01/orthographic_lstm/training.logs b/stela/EN/03/01/orthographic_lstm/training.logs new file mode 100644 index 0000000000000000000000000000000000000000..9e349ad9c6629c19c6561c49a5216fbdc64f9dc8 --- /dev/null +++ b/stela/EN/03/01/orthographic_lstm/training.logs @@ -0,0 +1,19 @@ +06/03/2025 16:47:29 - lexical_benchmark.train.run_train - INFO - Loading lstm trainer class. +06/03/2025 16:47:43 - datasets - INFO - PyTorch version 2.5.1 available. +06/03/2025 16:47:43 - datasets - INFO - Polars version 1.26.0 available. +06/03/2025 16:47:43 - lexical_benchmark.train.train_params - INFO - Loading params from /lustre/fsn1/projects/rech/hhb/uye44va/workspace/lm_benchmark/prod/code/src/lexical_benchmark/train/train-params.toml +06/03/2025 16:47:43 - lexical_benchmark.train.trainers.lstm - INFO - Loading char-tokenizer +06/03/2025 16:47:43 - lexical_benchmark.train.trainers.lstm - INFO - Tokenizer loaded with vocabulary size: 115 +06/03/2025 16:47:43 - lexical_benchmark.train.trainers.lstm - INFO - Loading and processing datasets with joined utterances (max_length=1024) +06/03/2025 16:47:43 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/03/01/train.tokenized +06/03/2025 16:47:43 - lexical_benchmark.train.tokenizers - INFO - Read 116181 lines from file +06/03/2025 16:47:53 - lexical_benchmark.train.tokenizers - INFO - Created 11028 joined examples from 116181 original lines +06/03/2025 16:47:54 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/dev/dev.tokenized +06/03/2025 16:47:54 - lexical_benchmark.train.tokenizers - INFO - Read 179806 lines from file +06/03/2025 16:48:11 - lexical_benchmark.train.tokenizers - INFO - Created 18373 joined examples from 179806 original lines +06/03/2025 16:48:13 - lexical_benchmark.train.trainers.lstm - INFO - Training dataset size: 11028 +06/03/2025 16:48:13 - lexical_benchmark.train.trainers.lstm - INFO - Validation dataset size: 18373 +06/03/2025 16:48:13 - lexical_benchmark.train.trainers.lstm - INFO - Loading configurations & initialising LSTM model trainer +06/03/2025 16:48:14 - lexical_benchmark.train.run_train - INFO - Running training of stela_lstm_EN_03_01 +06/03/2025 19:02:13 - lexical_benchmark.train.run_train - INFO - Training of stela_lstm_EN_03_01 has completed ! +06/03/2025 19:02:13 - lexical_benchmark.train.run_train - INFO - Model saved to /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/models/stela/EN/03/01/lstm diff --git a/stela/EN/03/01/orthographic_lstm/training_args.bin b/stela/EN/03/01/orthographic_lstm/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..61d2a93bd6e4519bbab594690313809d21161884 --- /dev/null +++ b/stela/EN/03/01/orthographic_lstm/training_args.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3aa4d29dbea494b738636d5771c5242aadb9453ce4b68bb8a883a2c17893bfda +size 5496 diff --git a/stela/EN/03/03/orthographic_gpt2/training.logs b/stela/EN/03/03/orthographic_gpt2/training.logs new file mode 100644 index 0000000000000000000000000000000000000000..61e9a216fb9c894679f04604c084d407b391f1fc --- /dev/null +++ b/stela/EN/03/03/orthographic_gpt2/training.logs @@ -0,0 +1,6 @@ +05/22/2025 20:09:40 - urllib3.connectionpool - WARNING - Retrying (Retry(total=2, connect=None, read=None, redirect=None, status=None)) after connection broken by 'NewConnectionError(': Failed to establish a new connection: [Errno 101] Network is unreachable')': /api/4504800232407040/envelope/ +05/22/2025 20:09:40 - urllib3.connectionpool - WARNING - Retrying (Retry(total=1, connect=None, read=None, redirect=None, status=None)) after connection broken by 'NewConnectionError(': Failed to establish a new connection: [Errno 101] Network is unreachable')': /api/4504800232407040/envelope/ +05/22/2025 20:09:40 - urllib3.connectionpool - WARNING - Retrying (Retry(total=0, connect=None, read=None, redirect=None, status=None)) after connection broken by 'NewConnectionError(': Failed to establish a new connection: [Errno 101] Network is unreachable')': /api/4504800232407040/envelope/ +05/22/2025 20:09:40 - urllib3.connectionpool - WARNING - Retrying (Retry(total=2, connect=None, read=None, redirect=None, status=None)) after connection broken by 'NewConnectionError(': Failed to establish a new connection: [Errno 101] Network is unreachable')': /api/4504800232407040/envelope/ +05/22/2025 20:09:40 - urllib3.connectionpool - WARNING - Retrying (Retry(total=1, connect=None, read=None, redirect=None, status=None)) after connection broken by 'NewConnectionError(': Failed to establish a new connection: [Errno 101] Network is unreachable')': /api/4504800232407040/envelope/ +05/22/2025 20:09:40 - urllib3.connectionpool - WARNING - Retrying (Retry(total=0, connect=None, read=None, redirect=None, status=None)) after connection broken by 'NewConnectionError(': Failed to establish a new connection: [Errno 101] Network is unreachable')': /api/4504800232407040/envelope/ diff --git a/stela/EN/04/00/orthographic_gpt2/config.json b/stela/EN/04/00/orthographic_gpt2/config.json new file mode 100644 index 0000000000000000000000000000000000000000..493011df283c5726a4be3aa8f39c84d43991636e --- /dev/null +++ b/stela/EN/04/00/orthographic_gpt2/config.json @@ -0,0 +1,32 @@ +{ + "activation_function": "gelu_new", + "architectures": [ + "GPT2LMHeadModel" + ], + "attn_pdrop": 0.1, + "bos_token_id": 50256, + "embd_pdrop": 0.1, + "eos_token_id": 50256, + "initializer_range": 0.02, + "layer_norm_epsilon": 1e-05, + "model_type": "gpt2", + "n_ctx": 1024, + "n_embd": 768, + "n_head": 12, + "n_inner": null, + "n_layer": 12, + "n_positions": 1024, + "reorder_and_upcast_attn": false, + "resid_pdrop": 0.1, + "scale_attn_by_inverse_layer_idx": false, + "scale_attn_weights": true, + "summary_activation": null, + "summary_first_dropout": 0.1, + "summary_proj_to_labels": true, + "summary_type": "cls_index", + "summary_use_proj": true, + "torch_dtype": "float32", + "transformers_version": "4.49.0", + "use_cache": true, + "vocab_size": 115 +} diff --git a/stela/EN/04/00/orthographic_gpt2/generation_config.json b/stela/EN/04/00/orthographic_gpt2/generation_config.json new file mode 100644 index 0000000000000000000000000000000000000000..bd968f46ce39e754a4603c69f53a2a4a08a13dbb --- /dev/null +++ b/stela/EN/04/00/orthographic_gpt2/generation_config.json @@ -0,0 +1,6 @@ +{ + "_from_model_config": true, + "bos_token_id": 50256, + "eos_token_id": 50256, + "transformers_version": "4.49.0" +} diff --git a/stela/EN/04/00/orthographic_gpt2/model.safetensors b/stela/EN/04/00/orthographic_gpt2/model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..9621ea6741ca5ffad29e4d5915c8d048767a658a --- /dev/null +++ b/stela/EN/04/00/orthographic_gpt2/model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:eea05a1b1f368a484f04752002a3609ca3ba147b783d4f637b9ded45350541af +size 343737984 diff --git a/stela/EN/04/00/orthographic_gpt2/training.logs b/stela/EN/04/00/orthographic_gpt2/training.logs new file mode 100644 index 0000000000000000000000000000000000000000..dc0a4acabe93bf71f6a9c8bef6fedac6bc7dc047 --- /dev/null +++ b/stela/EN/04/00/orthographic_gpt2/training.logs @@ -0,0 +1,171 @@ +04/30/2025 18:19:34 - lexical_benchmark.train.run_train - INFO - Loading gpt2 trainer class. +04/30/2025 18:19:47 - datasets - INFO - PyTorch version 2.5.1 available. +04/30/2025 18:19:47 - datasets - INFO - Polars version 1.26.0 available. +04/30/2025 18:19:47 - lexical_benchmark.train.train_params - INFO - Loading params from /lustre/fsn1/projects/rech/hhb/uye44va/workspace/lm_benchmark/prod/code/src/lexical_benchmark/train/train-params.toml +04/30/2025 18:19:47 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading char tokenizer +04/30/2025 18:19:47 - lexical_benchmark.train.trainers.gpt2 - INFO - Tokenizer loaded with vocabulary size: 115 +04/30/2025 18:19:47 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading and processing datasets with joined utterances (max_length=1024) +04/30/2025 18:19:47 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/04/04/train.tokenized +04/30/2025 18:19:47 - lexical_benchmark.train.tokenizers - INFO - Read 150813 lines from file +04/30/2025 18:20:00 - lexical_benchmark.train.tokenizers - INFO - Created 14763 joined examples from 150813 original lines +04/30/2025 18:20:02 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/dev/dev.tokenized +04/30/2025 18:20:03 - lexical_benchmark.train.tokenizers - INFO - Read 179806 lines from file +04/30/2025 18:20:19 - lexical_benchmark.train.tokenizers - INFO - Created 18373 joined examples from 179806 original lines +04/30/2025 18:20:22 - lexical_benchmark.train.trainers.gpt2 - INFO - Training dataset size: 14763 +04/30/2025 18:20:22 - lexical_benchmark.train.trainers.gpt2 - INFO - Validation dataset size: 18373 +04/30/2025 18:20:22 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating GPT2 configuration +04/30/2025 18:20:22 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating standard Trainer +04/30/2025 18:20:23 - lexical_benchmark.train.run_train - INFO - Running training of stela_gpt2_EN_04_04 +... +04/30/2025 19:27:14 - urllib3.connectionpool - WARNING - Retrying (Retry(total=2, connect=None, read=None, redirect=None, status=None)) after connection broken by 'NewConnectionError(': Failed to establish a new connection: [Errno 101] Network is unreachable')': /api/4504800232407040/envelope/ +04/30/2025 19:27:14 - urllib3.connectionpool - WARNING - Retrying (Retry(total=1, connect=None, read=None, redirect=None, status=None)) after connection broken by 'NewConnectionError(': Failed to establish a new connection: [Errno 101] Network is unreachable')': /api/4504800232407040/envelope/ +04/30/2025 19:27:14 - urllib3.connectionpool - WARNING - Retrying (Retry(total=0, connect=None, read=None, redirect=None, status=None)) after connection broken by 'NewConnectionError(': Failed to establish a new connection: [Errno 101] Network is unreachable')': /api/4504800232407040/envelope/ +04/30/2025 19:27:14 - urllib3.connectionpool - WARNING - Retrying (Retry(total=2, connect=None, read=None, redirect=None, status=None)) after connection broken by 'NewConnectionError(': Failed to establish a new connection: [Errno 101] Network is unreachable')': /api/4504800232407040/envelope/ +04/30/2025 19:27:14 - urllib3.connectionpool - WARNING - Retrying (Retry(total=1, connect=None, read=None, redirect=None, status=None)) after connection broken by 'NewConnectionError(': Failed to establish a new connection: [Errno 101] Network is unreachable')': /api/4504800232407040/envelope/ +04/30/2025 19:27:14 - urllib3.connectionpool - WARNING - Retrying (Retry(total=0, connect=None, read=None, redirect=None, status=None)) after connection broken by 'NewConnectionError(': Failed to establish a new connection: [Errno 101] Network is unreachable')': /api/4504800232407040/envelope/ +04/30/2025 19:52:13 - lexical_benchmark.train.run_train - INFO - Loading gpt2 trainer class. +04/30/2025 19:52:27 - datasets - INFO - PyTorch version 2.5.1 available. +04/30/2025 19:52:27 - datasets - INFO - Polars version 1.26.0 available. +04/30/2025 19:52:27 - lexical_benchmark.train.train_params - INFO - Loading params from /lustre/fsn1/projects/rech/hhb/ucx81cx/workspace/lm_benchmark/dev/code/src/lexical_benchmark/train/train-params.toml +04/30/2025 19:52:27 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading char tokenizer +04/30/2025 19:52:28 - lexical_benchmark.train.trainers.gpt2 - INFO - Tokenizer loaded with vocabulary size: 115 +04/30/2025 19:52:28 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading and processing datasets with joined utterances (max_length=1024) +04/30/2025 19:52:28 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/04/04/train.tokenized +04/30/2025 19:52:28 - lexical_benchmark.train.tokenizers - INFO - Read 150813 lines from file +04/30/2025 19:52:44 - lexical_benchmark.train.tokenizers - INFO - Created 14763 joined examples from 150813 original lines +04/30/2025 19:52:47 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/dev/dev.tokenized +04/30/2025 19:52:47 - lexical_benchmark.train.tokenizers - INFO - Read 179806 lines from file +04/30/2025 19:53:08 - lexical_benchmark.train.tokenizers - INFO - Created 18373 joined examples from 179806 original lines +04/30/2025 19:53:11 - lexical_benchmark.train.trainers.gpt2 - INFO - Training dataset size: 14763 +04/30/2025 19:53:11 - lexical_benchmark.train.trainers.gpt2 - INFO - Validation dataset size: 18373 +04/30/2025 19:53:11 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating GPT2 configuration +04/30/2025 19:53:12 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating standard Trainer +04/30/2025 19:53:12 - lexical_benchmark.train.run_train - INFO - Running training of stela_gpt2_EN_04_04 +04/30/2025 20:08:34 - lexical_benchmark.train.run_train - INFO - Loading gpt2 trainer class. +04/30/2025 20:08:47 - datasets - INFO - PyTorch version 2.5.1 available. +04/30/2025 20:08:47 - datasets - INFO - Polars version 1.26.0 available. +04/30/2025 20:08:47 - lexical_benchmark.train.train_params - INFO - Loading params from /lustre/fsn1/projects/rech/hhb/ucx81cx/workspace/lm_benchmark/dev/code/src/lexical_benchmark/train/train-params.toml +04/30/2025 20:08:47 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading char tokenizer +04/30/2025 20:08:47 - lexical_benchmark.train.trainers.gpt2 - INFO - Tokenizer loaded with vocabulary size: 115 +04/30/2025 20:08:47 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading and processing datasets with joined utterances (max_length=1024) +04/30/2025 20:08:47 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/04/04/train.tokenized +04/30/2025 20:08:48 - lexical_benchmark.train.tokenizers - INFO - Read 150813 lines from file +04/30/2025 20:09:01 - lexical_benchmark.train.tokenizers - INFO - Created 14763 joined examples from 150813 original lines +04/30/2025 20:09:03 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/dev/dev.tokenized +04/30/2025 20:09:03 - lexical_benchmark.train.tokenizers - INFO - Read 179806 lines from file +04/30/2025 20:09:19 - lexical_benchmark.train.tokenizers - INFO - Created 18373 joined examples from 179806 original lines +04/30/2025 20:09:22 - lexical_benchmark.train.trainers.gpt2 - INFO - Training dataset size: 14763 +04/30/2025 20:09:22 - lexical_benchmark.train.trainers.gpt2 - INFO - Validation dataset size: 18373 +04/30/2025 20:09:22 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating GPT2 configuration +04/30/2025 20:09:23 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating standard Trainer +04/30/2025 20:09:23 - lexical_benchmark.train.run_train - INFO - Running training of stela_gpt2_EN_04_04 +05/23/2025 11:45:14 - lexical_benchmark.train.run_train - INFO - Loading gpt2 trainer class. +05/23/2025 11:45:26 - datasets - INFO - PyTorch version 2.5.1 available. +05/23/2025 11:45:26 - datasets - INFO - Polars version 1.26.0 available. +05/23/2025 11:45:27 - lexical_benchmark.train.train_params - INFO - Loading params from /lustre/fsn1/projects/rech/hhb/uye44va/workspace/lm_benchmark/prod/code/src/lexical_benchmark/train/train-params.toml +05/23/2025 11:45:27 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading char tokenizer +05/23/2025 11:45:27 - lexical_benchmark.train.trainers.gpt2 - INFO - Tokenizer loaded with vocabulary size: 115 +05/23/2025 11:45:27 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading and processing datasets with joined utterances (max_length=1024) +05/23/2025 11:45:27 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/04/00/train.tokenized +05/23/2025 11:45:27 - lexical_benchmark.train.tokenizers - INFO - Read 146511 lines from file +05/23/2025 11:45:41 - lexical_benchmark.train.tokenizers - INFO - Created 14678 joined examples from 146511 original lines +05/23/2025 11:45:43 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/dev/dev.tokenized +05/23/2025 11:45:43 - lexical_benchmark.train.tokenizers - INFO - Read 179806 lines from file +05/23/2025 11:46:00 - lexical_benchmark.train.tokenizers - INFO - Created 18373 joined examples from 179806 original lines +05/23/2025 11:46:03 - lexical_benchmark.train.trainers.gpt2 - INFO - Training dataset size: 14678 +05/23/2025 11:46:03 - lexical_benchmark.train.trainers.gpt2 - INFO - Validation dataset size: 18373 +05/23/2025 11:46:03 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating GPT2 configuration +05/23/2025 11:46:04 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating standard Trainer +05/23/2025 11:46:04 - lexical_benchmark.train.run_train - INFO - Running training of stela_gpt2_EN_04_00 +05/23/2025 17:33:52 - lexical_benchmark.train.run_train - INFO - Loading gpt2 trainer class. +05/23/2025 17:34:05 - datasets - INFO - PyTorch version 2.5.1 available. +05/23/2025 17:34:05 - datasets - INFO - Polars version 1.26.0 available. +05/23/2025 17:34:05 - lexical_benchmark.train.train_params - INFO - Loading params from /lustre/fsn1/projects/rech/hhb/uye44va/workspace/lm_benchmark/prod/code/src/lexical_benchmark/train/train-params.toml +05/23/2025 17:34:05 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading char tokenizer +05/23/2025 17:34:05 - lexical_benchmark.train.trainers.gpt2 - INFO - Tokenizer loaded with vocabulary size: 115 +05/23/2025 17:34:05 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading and processing datasets with joined utterances (max_length=1024) +05/23/2025 17:34:05 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/04/00/train.tokenized +05/23/2025 17:34:05 - lexical_benchmark.train.tokenizers - INFO - Read 146511 lines from file +05/23/2025 17:34:21 - lexical_benchmark.train.tokenizers - INFO - Created 14678 joined examples from 146511 original lines +05/23/2025 17:34:24 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/dev/dev.tokenized +05/23/2025 17:34:24 - lexical_benchmark.train.tokenizers - INFO - Read 179806 lines from file +05/23/2025 17:34:44 - lexical_benchmark.train.tokenizers - INFO - Created 18373 joined examples from 179806 original lines +05/23/2025 17:34:47 - lexical_benchmark.train.trainers.gpt2 - INFO - Training dataset size: 14678 +05/23/2025 17:34:47 - lexical_benchmark.train.trainers.gpt2 - INFO - Validation dataset size: 18373 +05/23/2025 17:34:47 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating GPT2 configuration +05/23/2025 17:34:48 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating standard Trainer +05/23/2025 17:34:48 - lexical_benchmark.train.run_train - INFO - Running training of stela_gpt2_EN_04_00 +05/23/2025 18:48:00 - lexical_benchmark.train.run_train - INFO - Loading gpt2 trainer class. +05/23/2025 18:48:13 - datasets - INFO - PyTorch version 2.5.1 available. +05/23/2025 18:48:13 - datasets - INFO - Polars version 1.26.0 available. +05/23/2025 18:48:13 - lexical_benchmark.train.train_params - INFO - Loading params from /lustre/fsn1/projects/rech/hhb/uye44va/workspace/lm_benchmark/prod/code/src/lexical_benchmark/train/train-params.toml +05/23/2025 18:48:13 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading char tokenizer +05/23/2025 18:48:13 - lexical_benchmark.train.trainers.gpt2 - INFO - Tokenizer loaded with vocabulary size: 115 +05/23/2025 18:48:13 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading and processing datasets with joined utterances (max_length=1024) +05/23/2025 18:48:13 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/04/00/train.tokenized +05/23/2025 18:48:13 - lexical_benchmark.train.tokenizers - INFO - Read 146511 lines from file +05/23/2025 18:48:26 - lexical_benchmark.train.tokenizers - INFO - Created 14678 joined examples from 146511 original lines +05/23/2025 18:48:28 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/dev/dev.tokenized +05/23/2025 18:48:28 - lexical_benchmark.train.tokenizers - INFO - Read 179806 lines from file +05/23/2025 18:48:45 - lexical_benchmark.train.tokenizers - INFO - Created 18373 joined examples from 179806 original lines +05/23/2025 18:48:47 - lexical_benchmark.train.trainers.gpt2 - INFO - Training dataset size: 14678 +05/23/2025 18:48:47 - lexical_benchmark.train.trainers.gpt2 - INFO - Validation dataset size: 18373 +05/23/2025 18:48:47 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating GPT2 configuration +05/23/2025 18:48:48 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating standard Trainer +05/23/2025 18:48:49 - lexical_benchmark.train.run_train - INFO - Running training of stela_gpt2_EN_04_00 +05/25/2025 22:46:20 - lexical_benchmark.train.run_train - INFO - Loading gpt2 trainer class. +05/25/2025 22:46:32 - datasets - INFO - PyTorch version 2.5.1 available. +05/25/2025 22:46:32 - datasets - INFO - Polars version 1.26.0 available. +05/25/2025 22:46:32 - lexical_benchmark.train.train_params - INFO - Loading params from /lustre/fsn1/projects/rech/hhb/uye44va/workspace/lm_benchmark/prod/code/src/lexical_benchmark/train/train-params.toml +05/25/2025 22:46:32 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading char tokenizer +05/25/2025 22:46:33 - lexical_benchmark.train.trainers.gpt2 - INFO - Tokenizer loaded with vocabulary size: 115 +05/25/2025 22:46:33 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading and processing datasets with joined utterances (max_length=1024) +05/25/2025 22:46:33 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/04/00/train.tokenized +05/25/2025 22:46:33 - lexical_benchmark.train.tokenizers - INFO - Read 146511 lines from file +05/25/2025 22:46:46 - lexical_benchmark.train.tokenizers - INFO - Created 14678 joined examples from 146511 original lines +05/25/2025 22:46:48 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/dev/dev.tokenized +05/25/2025 22:46:48 - lexical_benchmark.train.tokenizers - INFO - Read 179806 lines from file +05/25/2025 22:47:05 - lexical_benchmark.train.tokenizers - INFO - Created 18373 joined examples from 179806 original lines +05/25/2025 22:47:08 - lexical_benchmark.train.trainers.gpt2 - INFO - Training dataset size: 14678 +05/25/2025 22:47:08 - lexical_benchmark.train.trainers.gpt2 - INFO - Validation dataset size: 18373 +05/25/2025 22:47:08 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating GPT2 configuration +05/25/2025 22:47:08 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating standard Trainer +05/25/2025 22:47:09 - lexical_benchmark.train.run_train - INFO - Running training of stela_gpt2_EN_04_00 +05/26/2025 00:32:20 - lexical_benchmark.train.run_train - INFO - Loading gpt2 trainer class. +05/26/2025 00:32:33 - datasets - INFO - PyTorch version 2.5.1 available. +05/26/2025 00:32:33 - datasets - INFO - Polars version 1.26.0 available. +05/26/2025 00:32:33 - lexical_benchmark.train.train_params - INFO - Loading params from /lustre/fsn1/projects/rech/hhb/uye44va/workspace/lm_benchmark/prod/code/src/lexical_benchmark/train/train-params.toml +05/26/2025 00:32:33 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading char tokenizer +05/26/2025 00:32:33 - lexical_benchmark.train.trainers.gpt2 - INFO - Tokenizer loaded with vocabulary size: 115 +05/26/2025 00:32:33 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading and processing datasets with joined utterances (max_length=1024) +05/26/2025 00:32:33 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/04/00/train.tokenized +05/26/2025 00:32:33 - lexical_benchmark.train.tokenizers - INFO - Read 146511 lines from file +05/26/2025 00:32:46 - lexical_benchmark.train.tokenizers - INFO - Created 14678 joined examples from 146511 original lines +05/26/2025 00:32:48 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/dev/dev.tokenized +05/26/2025 00:32:48 - lexical_benchmark.train.tokenizers - INFO - Read 179806 lines from file +05/26/2025 00:33:04 - lexical_benchmark.train.tokenizers - INFO - Created 18373 joined examples from 179806 original lines +05/26/2025 00:33:07 - lexical_benchmark.train.trainers.gpt2 - INFO - Training dataset size: 14678 +05/26/2025 00:33:07 - lexical_benchmark.train.trainers.gpt2 - INFO - Validation dataset size: 18373 +05/26/2025 00:33:07 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating GPT2 configuration +05/26/2025 00:33:08 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating standard Trainer +05/26/2025 00:33:08 - lexical_benchmark.train.run_train - INFO - Running training of stela_gpt2_EN_04_00 +05/26/2025 00:38:30 - lexical_benchmark.train.run_train - INFO - Loading gpt2 trainer class. +05/26/2025 00:38:42 - datasets - INFO - PyTorch version 2.5.1 available. +05/26/2025 00:38:42 - datasets - INFO - Polars version 1.26.0 available. +05/26/2025 00:38:42 - lexical_benchmark.train.train_params - INFO - Loading params from /lustre/fsn1/projects/rech/hhb/uye44va/workspace/lm_benchmark/prod/code/src/lexical_benchmark/train/train-params.toml +05/26/2025 00:38:42 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading char tokenizer +05/26/2025 00:38:42 - lexical_benchmark.train.trainers.gpt2 - INFO - Tokenizer loaded with vocabulary size: 115 +05/26/2025 00:38:42 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading and processing datasets with joined utterances (max_length=1024) +05/26/2025 00:38:42 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/04/00/train.tokenized +05/26/2025 00:38:42 - lexical_benchmark.train.tokenizers - INFO - Read 146511 lines from file +05/26/2025 00:38:55 - lexical_benchmark.train.tokenizers - INFO - Created 14678 joined examples from 146511 original lines +05/26/2025 00:38:57 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/dev/dev.tokenized +05/26/2025 00:38:57 - lexical_benchmark.train.tokenizers - INFO - Read 179806 lines from file +05/26/2025 00:39:13 - lexical_benchmark.train.tokenizers - INFO - Created 18373 joined examples from 179806 original lines +05/26/2025 00:39:16 - lexical_benchmark.train.trainers.gpt2 - INFO - Training dataset size: 14678 +05/26/2025 00:39:16 - lexical_benchmark.train.trainers.gpt2 - INFO - Validation dataset size: 18373 +05/26/2025 00:39:16 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating GPT2 configuration +05/26/2025 00:39:17 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating standard Trainer +05/26/2025 00:39:17 - lexical_benchmark.train.run_train - INFO - Running training of stela_gpt2_EN_04_00 +05/26/2025 02:15:19 - lexical_benchmark.train.run_train - INFO - Training of stela_gpt2_EN_04_00 has completed ! +05/26/2025 02:15:19 - lexical_benchmark.train.run_train - INFO - Model saved to /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/models/stela/EN/04/00/gpt2 diff --git a/stela/EN/04/00/orthographic_gpt2/training_args.bin b/stela/EN/04/00/orthographic_gpt2/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..2793ca21ad7115674e196f83a842004eebb2741d --- /dev/null +++ b/stela/EN/04/00/orthographic_gpt2/training_args.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:73da87bac060335058464dcb750758903888005889a92f8abf35819c6b0ece0a +size 5496 diff --git a/stela/EN/04/00/orthographic_lstm/config.json b/stela/EN/04/00/orthographic_lstm/config.json new file mode 100644 index 0000000000000000000000000000000000000000..5ef629887c7fc87d64278c9788e493ac34c60370 --- /dev/null +++ b/stela/EN/04/00/orthographic_lstm/config.json @@ -0,0 +1,13 @@ +{ + "architectures": [ + "LSTMForLanguageModeling" + ], + "dropout": 0.1, + "embedding_dim": 200, + "hidden_size": 1024, + "model_type": "lstm", + "num_layers": 3, + "torch_dtype": "float32", + "transformers_version": "4.49.0", + "vocab_size": 115 +} diff --git a/stela/EN/04/00/orthographic_lstm/generation_config.json b/stela/EN/04/00/orthographic_lstm/generation_config.json new file mode 100644 index 0000000000000000000000000000000000000000..c1818a95e07bdccb83305ce8f9d878755e3c6119 --- /dev/null +++ b/stela/EN/04/00/orthographic_lstm/generation_config.json @@ -0,0 +1,4 @@ +{ + "_from_model_config": true, + "transformers_version": "4.49.0" +} diff --git a/stela/EN/04/00/orthographic_lstm/model.safetensors b/stela/EN/04/00/orthographic_lstm/model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..8ccde70d220778a9041cf23a7870152b8f266327 --- /dev/null +++ b/stela/EN/04/00/orthographic_lstm/model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:27d0cb83a2d044a0d17ede844281f93123a165827153b090f243085d3c2ff3ab +size 87825996 diff --git a/stela/EN/04/00/orthographic_lstm/training_args.bin b/stela/EN/04/00/orthographic_lstm/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..a8a64adddb73efb5fec410f61d3773211cf4966a --- /dev/null +++ b/stela/EN/04/00/orthographic_lstm/training_args.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a51c5978ded6649550a7bcde44b344e3fb13d3c3c09516600f01b925cfafc6b5 +size 5496 diff --git a/stela/EN/04/01/orthographic_gpt2/config.json b/stela/EN/04/01/orthographic_gpt2/config.json new file mode 100644 index 0000000000000000000000000000000000000000..493011df283c5726a4be3aa8f39c84d43991636e --- /dev/null +++ b/stela/EN/04/01/orthographic_gpt2/config.json @@ -0,0 +1,32 @@ +{ + "activation_function": "gelu_new", + "architectures": [ + "GPT2LMHeadModel" + ], + "attn_pdrop": 0.1, + "bos_token_id": 50256, + "embd_pdrop": 0.1, + "eos_token_id": 50256, + "initializer_range": 0.02, + "layer_norm_epsilon": 1e-05, + "model_type": "gpt2", + "n_ctx": 1024, + "n_embd": 768, + "n_head": 12, + "n_inner": null, + "n_layer": 12, + "n_positions": 1024, + "reorder_and_upcast_attn": false, + "resid_pdrop": 0.1, + "scale_attn_by_inverse_layer_idx": false, + "scale_attn_weights": true, + "summary_activation": null, + "summary_first_dropout": 0.1, + "summary_proj_to_labels": true, + "summary_type": "cls_index", + "summary_use_proj": true, + "torch_dtype": "float32", + "transformers_version": "4.49.0", + "use_cache": true, + "vocab_size": 115 +} diff --git a/stela/EN/04/01/orthographic_gpt2/generation_config.json b/stela/EN/04/01/orthographic_gpt2/generation_config.json new file mode 100644 index 0000000000000000000000000000000000000000..bd968f46ce39e754a4603c69f53a2a4a08a13dbb --- /dev/null +++ b/stela/EN/04/01/orthographic_gpt2/generation_config.json @@ -0,0 +1,6 @@ +{ + "_from_model_config": true, + "bos_token_id": 50256, + "eos_token_id": 50256, + "transformers_version": "4.49.0" +} diff --git a/stela/EN/04/01/orthographic_gpt2/model.safetensors b/stela/EN/04/01/orthographic_gpt2/model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..80682f9d4974f93718b694d07d7e82f9288ca25e --- /dev/null +++ b/stela/EN/04/01/orthographic_gpt2/model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ea7694e5789dc70c1abe4d6c53e2a373bf01e40a3cd101113db1eb78f9d9185b +size 343737984 diff --git a/stela/EN/04/01/orthographic_gpt2/training.logs b/stela/EN/04/01/orthographic_gpt2/training.logs new file mode 100644 index 0000000000000000000000000000000000000000..442e40be12a15339201fa21da7ea5e49e9f3a284 --- /dev/null +++ b/stela/EN/04/01/orthographic_gpt2/training.logs @@ -0,0 +1,20 @@ +06/03/2025 16:47:28 - lexical_benchmark.train.run_train - INFO - Loading gpt2 trainer class. +06/03/2025 16:47:41 - datasets - INFO - PyTorch version 2.5.1 available. +06/03/2025 16:47:41 - datasets - INFO - Polars version 1.26.0 available. +06/03/2025 16:47:41 - lexical_benchmark.train.train_params - INFO - Loading params from /lustre/fsn1/projects/rech/hhb/uye44va/workspace/lm_benchmark/prod/code/src/lexical_benchmark/train/train-params.toml +06/03/2025 16:47:41 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading char tokenizer +06/03/2025 16:47:41 - lexical_benchmark.train.trainers.gpt2 - INFO - Tokenizer loaded with vocabulary size: 115 +06/03/2025 16:47:41 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading and processing datasets with joined utterances (max_length=1024) +06/03/2025 16:47:42 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/04/01/train.tokenized +06/03/2025 16:47:42 - lexical_benchmark.train.tokenizers - INFO - Read 153617 lines from file +06/03/2025 16:47:56 - lexical_benchmark.train.tokenizers - INFO - Created 14658 joined examples from 153617 original lines +06/03/2025 16:47:58 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/dev/dev.tokenized +06/03/2025 16:47:58 - lexical_benchmark.train.tokenizers - INFO - Read 179806 lines from file +06/03/2025 16:48:17 - lexical_benchmark.train.tokenizers - INFO - Created 18373 joined examples from 179806 original lines +06/03/2025 16:48:19 - lexical_benchmark.train.trainers.gpt2 - INFO - Training dataset size: 14658 +06/03/2025 16:48:19 - lexical_benchmark.train.trainers.gpt2 - INFO - Validation dataset size: 18373 +06/03/2025 16:48:19 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating GPT2 configuration +06/03/2025 16:48:20 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating standard Trainer +06/03/2025 16:48:21 - lexical_benchmark.train.run_train - INFO - Running training of stela_gpt2_EN_04_01 +06/03/2025 18:24:36 - lexical_benchmark.train.run_train - INFO - Training of stela_gpt2_EN_04_01 has completed ! +06/03/2025 18:24:36 - lexical_benchmark.train.run_train - INFO - Model saved to /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/models/stela/EN/04/01/gpt2 diff --git a/stela/EN/04/01/orthographic_gpt2/training_args.bin b/stela/EN/04/01/orthographic_gpt2/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..139151bb83a6a748b079316697e10cce71a1fcf2 --- /dev/null +++ b/stela/EN/04/01/orthographic_gpt2/training_args.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:97547963fc0cad987c27f79dc8c8344e71c91a2b133aa27099e5b2089071b8f4 +size 5496 diff --git a/stela/EN/04/01/orthographic_lstm/config.json b/stela/EN/04/01/orthographic_lstm/config.json new file mode 100644 index 0000000000000000000000000000000000000000..5ef629887c7fc87d64278c9788e493ac34c60370 --- /dev/null +++ b/stela/EN/04/01/orthographic_lstm/config.json @@ -0,0 +1,13 @@ +{ + "architectures": [ + "LSTMForLanguageModeling" + ], + "dropout": 0.1, + "embedding_dim": 200, + "hidden_size": 1024, + "model_type": "lstm", + "num_layers": 3, + "torch_dtype": "float32", + "transformers_version": "4.49.0", + "vocab_size": 115 +} diff --git a/stela/EN/04/01/orthographic_lstm/generation_config.json b/stela/EN/04/01/orthographic_lstm/generation_config.json new file mode 100644 index 0000000000000000000000000000000000000000..c1818a95e07bdccb83305ce8f9d878755e3c6119 --- /dev/null +++ b/stela/EN/04/01/orthographic_lstm/generation_config.json @@ -0,0 +1,4 @@ +{ + "_from_model_config": true, + "transformers_version": "4.49.0" +} diff --git a/stela/EN/04/01/orthographic_lstm/model.safetensors b/stela/EN/04/01/orthographic_lstm/model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..fc9590ec042000b16941b062f2799fc149382571 --- /dev/null +++ b/stela/EN/04/01/orthographic_lstm/model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c8818b1fbb6a83b927b0753b0737422c296313f87ebdc8332304e400e746ef1e +size 87825996 diff --git a/stela/EN/04/01/orthographic_lstm/training.logs b/stela/EN/04/01/orthographic_lstm/training.logs new file mode 100644 index 0000000000000000000000000000000000000000..f16cc13ec800bef017ad9a416b18a12b110e73ac --- /dev/null +++ b/stela/EN/04/01/orthographic_lstm/training.logs @@ -0,0 +1,19 @@ +06/03/2025 16:47:29 - lexical_benchmark.train.run_train - INFO - Loading lstm trainer class. +06/03/2025 16:47:42 - datasets - INFO - PyTorch version 2.5.1 available. +06/03/2025 16:47:42 - datasets - INFO - Polars version 1.26.0 available. +06/03/2025 16:47:42 - lexical_benchmark.train.train_params - INFO - Loading params from /lustre/fsn1/projects/rech/hhb/uye44va/workspace/lm_benchmark/prod/code/src/lexical_benchmark/train/train-params.toml +06/03/2025 16:47:42 - lexical_benchmark.train.trainers.lstm - INFO - Loading char-tokenizer +06/03/2025 16:47:42 - lexical_benchmark.train.trainers.lstm - INFO - Tokenizer loaded with vocabulary size: 115 +06/03/2025 16:47:42 - lexical_benchmark.train.trainers.lstm - INFO - Loading and processing datasets with joined utterances (max_length=1024) +06/03/2025 16:47:42 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/04/01/train.tokenized +06/03/2025 16:47:42 - lexical_benchmark.train.tokenizers - INFO - Read 153617 lines from file +06/03/2025 16:47:55 - lexical_benchmark.train.tokenizers - INFO - Created 14658 joined examples from 153617 original lines +06/03/2025 16:47:57 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/dev/dev.tokenized +06/03/2025 16:47:57 - lexical_benchmark.train.tokenizers - INFO - Read 179806 lines from file +06/03/2025 16:48:14 - lexical_benchmark.train.tokenizers - INFO - Created 18373 joined examples from 179806 original lines +06/03/2025 16:48:17 - lexical_benchmark.train.trainers.lstm - INFO - Training dataset size: 14658 +06/03/2025 16:48:17 - lexical_benchmark.train.trainers.lstm - INFO - Validation dataset size: 18373 +06/03/2025 16:48:17 - lexical_benchmark.train.trainers.lstm - INFO - Loading configurations & initialising LSTM model trainer +06/03/2025 16:48:17 - lexical_benchmark.train.run_train - INFO - Running training of stela_lstm_EN_04_01 +06/03/2025 20:17:56 - lexical_benchmark.train.run_train - INFO - Training of stela_lstm_EN_04_01 has completed ! +06/03/2025 20:17:56 - lexical_benchmark.train.run_train - INFO - Model saved to /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/models/stela/EN/04/01/lstm diff --git a/stela/EN/04/01/orthographic_lstm/training_args.bin b/stela/EN/04/01/orthographic_lstm/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..05c635e0afab8fa07a468f5d2d4346ebfbabc4c4 --- /dev/null +++ b/stela/EN/04/01/orthographic_lstm/training_args.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ef5a36e033fc921b2fb3d9c6181fe9bfc950ca87e5c1dfec907a1847c7ceccac +size 5496 diff --git a/stela/EN/04/04/orthographic_gpt2/training.logs b/stela/EN/04/04/orthographic_gpt2/training.logs new file mode 100644 index 0000000000000000000000000000000000000000..aa4d61bf707a2023eb786b5255a16fad0c393a28 --- /dev/null +++ b/stela/EN/04/04/orthographic_gpt2/training.logs @@ -0,0 +1,6 @@ +05/22/2025 20:09:40 - urllib3.connectionpool - WARNING - Retrying (Retry(total=2, connect=None, read=None, redirect=None, status=None)) after connection broken by 'NewConnectionError(': Failed to establish a new connection: [Errno 101] Network is unreachable')': /api/4504800232407040/envelope/ +05/22/2025 20:09:40 - urllib3.connectionpool - WARNING - Retrying (Retry(total=1, connect=None, read=None, redirect=None, status=None)) after connection broken by 'NewConnectionError(': Failed to establish a new connection: [Errno 101] Network is unreachable')': /api/4504800232407040/envelope/ +05/22/2025 20:09:40 - urllib3.connectionpool - WARNING - Retrying (Retry(total=0, connect=None, read=None, redirect=None, status=None)) after connection broken by 'NewConnectionError(': Failed to establish a new connection: [Errno 101] Network is unreachable')': /api/4504800232407040/envelope/ +05/22/2025 20:09:40 - urllib3.connectionpool - WARNING - Retrying (Retry(total=2, connect=None, read=None, redirect=None, status=None)) after connection broken by 'NewConnectionError(': Failed to establish a new connection: [Errno 101] Network is unreachable')': /api/4504800232407040/envelope/ +05/22/2025 20:09:40 - urllib3.connectionpool - WARNING - Retrying (Retry(total=1, connect=None, read=None, redirect=None, status=None)) after connection broken by 'NewConnectionError(': Failed to establish a new connection: [Errno 101] Network is unreachable')': /api/4504800232407040/envelope/ +05/22/2025 20:09:40 - urllib3.connectionpool - WARNING - Retrying (Retry(total=0, connect=None, read=None, redirect=None, status=None)) after connection broken by 'NewConnectionError(': Failed to establish a new connection: [Errno 101] Network is unreachable')': /api/4504800232407040/envelope/ diff --git a/stela/EN/05/00/orthographic_gpt2/config.json b/stela/EN/05/00/orthographic_gpt2/config.json new file mode 100644 index 0000000000000000000000000000000000000000..493011df283c5726a4be3aa8f39c84d43991636e --- /dev/null +++ b/stela/EN/05/00/orthographic_gpt2/config.json @@ -0,0 +1,32 @@ +{ + "activation_function": "gelu_new", + "architectures": [ + "GPT2LMHeadModel" + ], + "attn_pdrop": 0.1, + "bos_token_id": 50256, + "embd_pdrop": 0.1, + "eos_token_id": 50256, + "initializer_range": 0.02, + "layer_norm_epsilon": 1e-05, + "model_type": "gpt2", + "n_ctx": 1024, + "n_embd": 768, + "n_head": 12, + "n_inner": null, + "n_layer": 12, + "n_positions": 1024, + "reorder_and_upcast_attn": false, + "resid_pdrop": 0.1, + "scale_attn_by_inverse_layer_idx": false, + "scale_attn_weights": true, + "summary_activation": null, + "summary_first_dropout": 0.1, + "summary_proj_to_labels": true, + "summary_type": "cls_index", + "summary_use_proj": true, + "torch_dtype": "float32", + "transformers_version": "4.49.0", + "use_cache": true, + "vocab_size": 115 +} diff --git a/stela/EN/05/00/orthographic_gpt2/generation_config.json b/stela/EN/05/00/orthographic_gpt2/generation_config.json new file mode 100644 index 0000000000000000000000000000000000000000..bd968f46ce39e754a4603c69f53a2a4a08a13dbb --- /dev/null +++ b/stela/EN/05/00/orthographic_gpt2/generation_config.json @@ -0,0 +1,6 @@ +{ + "_from_model_config": true, + "bos_token_id": 50256, + "eos_token_id": 50256, + "transformers_version": "4.49.0" +} diff --git a/stela/EN/05/00/orthographic_gpt2/model.safetensors b/stela/EN/05/00/orthographic_gpt2/model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..5e02a8e8c03110bba31e84c2193cec5b22826f08 --- /dev/null +++ b/stela/EN/05/00/orthographic_gpt2/model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f5adf29e209f05b59127f499021fccf96aa731aeec263d10458af6c46f22c838 +size 343737984 diff --git a/stela/EN/05/00/orthographic_gpt2/training.logs b/stela/EN/05/00/orthographic_gpt2/training.logs new file mode 100644 index 0000000000000000000000000000000000000000..aee5b6ccd78811d03e96e40b4b050fe23301be39 --- /dev/null +++ b/stela/EN/05/00/orthographic_gpt2/training.logs @@ -0,0 +1,170 @@ +04/30/2025 18:19:34 - lexical_benchmark.train.run_train - INFO - Loading gpt2 trainer class. +04/30/2025 18:19:46 - datasets - INFO - PyTorch version 2.5.1 available. +04/30/2025 18:19:46 - datasets - INFO - Polars version 1.26.0 available. +04/30/2025 18:19:47 - lexical_benchmark.train.train_params - INFO - Loading params from /lustre/fsn1/projects/rech/hhb/uye44va/workspace/lm_benchmark/prod/code/src/lexical_benchmark/train/train-params.toml +04/30/2025 18:19:47 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading char tokenizer +04/30/2025 18:19:47 - lexical_benchmark.train.trainers.gpt2 - INFO - Tokenizer loaded with vocabulary size: 115 +04/30/2025 18:19:47 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading and processing datasets with joined utterances (max_length=1024) +04/30/2025 18:19:48 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/05/05/train.tokenized +04/30/2025 18:19:48 - lexical_benchmark.train.tokenizers - INFO - Read 191818 lines from file +04/30/2025 18:20:05 - lexical_benchmark.train.tokenizers - INFO - Created 18295 joined examples from 191818 original lines +04/30/2025 18:20:07 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/dev/dev.tokenized +04/30/2025 18:20:07 - lexical_benchmark.train.tokenizers - INFO - Read 179806 lines from file +04/30/2025 18:20:24 - lexical_benchmark.train.tokenizers - INFO - Created 18373 joined examples from 179806 original lines +04/30/2025 18:20:27 - lexical_benchmark.train.trainers.gpt2 - INFO - Training dataset size: 18295 +04/30/2025 18:20:27 - lexical_benchmark.train.trainers.gpt2 - INFO - Validation dataset size: 18373 +04/30/2025 18:20:27 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating GPT2 configuration +04/30/2025 18:20:28 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating standard Trainer +04/30/2025 18:20:28 - lexical_benchmark.train.run_train - INFO - Running training of stela_gpt2_EN_05_05 +04/30/2025 19:27:36 - urllib3.connectionpool - WARNING - Retrying (Retry(total=2, connect=None, read=None, redirect=None, status=None)) after connection broken by 'NewConnectionError(': Failed to establish a new connection: [Errno 101] Network is unreachable')': /api/4504800232407040/envelope/ +04/30/2025 19:27:36 - urllib3.connectionpool - WARNING - Retrying (Retry(total=1, connect=None, read=None, redirect=None, status=None)) after connection broken by 'NewConnectionError(': Failed to establish a new connection: [Errno 101] Network is unreachable')': /api/4504800232407040/envelope/ +04/30/2025 19:27:36 - urllib3.connectionpool - WARNING - Retrying (Retry(total=0, connect=None, read=None, redirect=None, status=None)) after connection broken by 'NewConnectionError(': Failed to establish a new connection: [Errno 101] Network is unreachable')': /api/4504800232407040/envelope/ +04/30/2025 19:27:36 - urllib3.connectionpool - WARNING - Retrying (Retry(total=2, connect=None, read=None, redirect=None, status=None)) after connection broken by 'NewConnectionError(': Failed to establish a new connection: [Errno 101] Network is unreachable')': /api/4504800232407040/envelope/ +04/30/2025 19:27:36 - urllib3.connectionpool - WARNING - Retrying (Retry(total=1, connect=None, read=None, redirect=None, status=None)) after connection broken by 'NewConnectionError(': Failed to establish a new connection: [Errno 101] Network is unreachable')': /api/4504800232407040/envelope/ +04/30/2025 19:27:36 - urllib3.connectionpool - WARNING - Retrying (Retry(total=0, connect=None, read=None, redirect=None, status=None)) after connection broken by 'NewConnectionError(': Failed to establish a new connection: [Errno 101] Network is unreachable')': /api/4504800232407040/envelope/ +04/30/2025 19:53:27 - lexical_benchmark.train.run_train - INFO - Loading gpt2 trainer class. +04/30/2025 19:53:33 - datasets - INFO - PyTorch version 2.5.1 available. +04/30/2025 19:53:33 - datasets - INFO - Polars version 1.26.0 available. +04/30/2025 19:53:34 - lexical_benchmark.train.train_params - INFO - Loading params from /lustre/fsn1/projects/rech/hhb/ucx81cx/workspace/lm_benchmark/dev/code/src/lexical_benchmark/train/train-params.toml +04/30/2025 19:53:34 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading char tokenizer +04/30/2025 19:53:34 - lexical_benchmark.train.trainers.gpt2 - INFO - Tokenizer loaded with vocabulary size: 115 +04/30/2025 19:53:34 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading and processing datasets with joined utterances (max_length=1024) +04/30/2025 19:53:34 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/05/05/train.tokenized +04/30/2025 19:53:34 - lexical_benchmark.train.tokenizers - INFO - Read 191818 lines from file +04/30/2025 19:53:54 - lexical_benchmark.train.tokenizers - INFO - Created 18295 joined examples from 191818 original lines +04/30/2025 19:53:58 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/dev/dev.tokenized +04/30/2025 19:53:58 - lexical_benchmark.train.tokenizers - INFO - Read 179806 lines from file +04/30/2025 19:54:19 - lexical_benchmark.train.tokenizers - INFO - Created 18373 joined examples from 179806 original lines +04/30/2025 19:54:22 - lexical_benchmark.train.trainers.gpt2 - INFO - Training dataset size: 18295 +04/30/2025 19:54:22 - lexical_benchmark.train.trainers.gpt2 - INFO - Validation dataset size: 18373 +04/30/2025 19:54:22 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating GPT2 configuration +04/30/2025 19:54:23 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating standard Trainer +04/30/2025 19:54:23 - lexical_benchmark.train.run_train - INFO - Running training of stela_gpt2_EN_05_05 +04/30/2025 20:08:34 - lexical_benchmark.train.run_train - INFO - Loading gpt2 trainer class. +04/30/2025 20:08:47 - datasets - INFO - PyTorch version 2.5.1 available. +04/30/2025 20:08:47 - datasets - INFO - Polars version 1.26.0 available. +04/30/2025 20:08:47 - lexical_benchmark.train.train_params - INFO - Loading params from /lustre/fsn1/projects/rech/hhb/ucx81cx/workspace/lm_benchmark/dev/code/src/lexical_benchmark/train/train-params.toml +04/30/2025 20:08:47 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading char tokenizer +04/30/2025 20:08:47 - lexical_benchmark.train.trainers.gpt2 - INFO - Tokenizer loaded with vocabulary size: 115 +04/30/2025 20:08:47 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading and processing datasets with joined utterances (max_length=1024) +04/30/2025 20:08:47 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/05/05/train.tokenized +04/30/2025 20:08:47 - lexical_benchmark.train.tokenizers - INFO - Read 191818 lines from file +04/30/2025 20:09:04 - lexical_benchmark.train.tokenizers - INFO - Created 18295 joined examples from 191818 original lines +04/30/2025 20:09:07 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/dev/dev.tokenized +04/30/2025 20:09:07 - lexical_benchmark.train.tokenizers - INFO - Read 179806 lines from file +04/30/2025 20:09:23 - lexical_benchmark.train.tokenizers - INFO - Created 18373 joined examples from 179806 original lines +04/30/2025 20:09:26 - lexical_benchmark.train.trainers.gpt2 - INFO - Training dataset size: 18295 +04/30/2025 20:09:26 - lexical_benchmark.train.trainers.gpt2 - INFO - Validation dataset size: 18373 +04/30/2025 20:09:26 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating GPT2 configuration +04/30/2025 20:09:27 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating standard Trainer +04/30/2025 20:09:27 - lexical_benchmark.train.run_train - INFO - Running training of stela_gpt2_EN_05_05 +05/23/2025 11:45:14 - lexical_benchmark.train.run_train - INFO - Loading gpt2 trainer class. +05/23/2025 11:45:28 - datasets - INFO - PyTorch version 2.5.1 available. +05/23/2025 11:45:28 - datasets - INFO - Polars version 1.26.0 available. +05/23/2025 11:45:28 - lexical_benchmark.train.train_params - INFO - Loading params from /lustre/fsn1/projects/rech/hhb/uye44va/workspace/lm_benchmark/prod/code/src/lexical_benchmark/train/train-params.toml +05/23/2025 11:45:28 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading char tokenizer +05/23/2025 11:45:28 - lexical_benchmark.train.trainers.gpt2 - INFO - Tokenizer loaded with vocabulary size: 115 +05/23/2025 11:45:28 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading and processing datasets with joined utterances (max_length=1024) +05/23/2025 11:45:29 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/05/00/train.tokenized +05/23/2025 11:45:29 - lexical_benchmark.train.tokenizers - INFO - Read 187919 lines from file +05/23/2025 11:45:45 - lexical_benchmark.train.tokenizers - INFO - Created 18264 joined examples from 187919 original lines +05/23/2025 11:45:48 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/dev/dev.tokenized +05/23/2025 11:45:48 - lexical_benchmark.train.tokenizers - INFO - Read 179806 lines from file +05/23/2025 11:46:04 - lexical_benchmark.train.tokenizers - INFO - Created 18373 joined examples from 179806 original lines +05/23/2025 11:46:07 - lexical_benchmark.train.trainers.gpt2 - INFO - Training dataset size: 18264 +05/23/2025 11:46:07 - lexical_benchmark.train.trainers.gpt2 - INFO - Validation dataset size: 18373 +05/23/2025 11:46:07 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating GPT2 configuration +05/23/2025 11:46:08 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating standard Trainer +05/23/2025 11:46:08 - lexical_benchmark.train.run_train - INFO - Running training of stela_gpt2_EN_05_00 +05/23/2025 17:33:52 - lexical_benchmark.train.run_train - INFO - Loading gpt2 trainer class. +05/23/2025 17:34:04 - datasets - INFO - PyTorch version 2.5.1 available. +05/23/2025 17:34:04 - datasets - INFO - Polars version 1.26.0 available. +05/23/2025 17:34:04 - lexical_benchmark.train.train_params - INFO - Loading params from /lustre/fsn1/projects/rech/hhb/uye44va/workspace/lm_benchmark/prod/code/src/lexical_benchmark/train/train-params.toml +05/23/2025 17:34:04 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading char tokenizer +05/23/2025 17:34:04 - lexical_benchmark.train.trainers.gpt2 - INFO - Tokenizer loaded with vocabulary size: 115 +05/23/2025 17:34:04 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading and processing datasets with joined utterances (max_length=1024) +05/23/2025 17:34:04 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/05/00/train.tokenized +05/23/2025 17:34:04 - lexical_benchmark.train.tokenizers - INFO - Read 187919 lines from file +05/23/2025 17:34:20 - lexical_benchmark.train.tokenizers - INFO - Created 18264 joined examples from 187919 original lines +05/23/2025 17:34:23 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/dev/dev.tokenized +05/23/2025 17:34:23 - lexical_benchmark.train.tokenizers - INFO - Read 179806 lines from file +05/23/2025 17:34:40 - lexical_benchmark.train.tokenizers - INFO - Created 18373 joined examples from 179806 original lines +05/23/2025 17:34:43 - lexical_benchmark.train.trainers.gpt2 - INFO - Training dataset size: 18264 +05/23/2025 17:34:43 - lexical_benchmark.train.trainers.gpt2 - INFO - Validation dataset size: 18373 +05/23/2025 17:34:43 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating GPT2 configuration +05/23/2025 17:34:43 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating standard Trainer +05/23/2025 17:34:44 - lexical_benchmark.train.run_train - INFO - Running training of stela_gpt2_EN_05_00 +05/23/2025 18:48:00 - lexical_benchmark.train.run_train - INFO - Loading gpt2 trainer class. +05/23/2025 18:48:13 - datasets - INFO - PyTorch version 2.5.1 available. +05/23/2025 18:48:13 - datasets - INFO - Polars version 1.26.0 available. +05/23/2025 18:48:13 - lexical_benchmark.train.train_params - INFO - Loading params from /lustre/fsn1/projects/rech/hhb/uye44va/workspace/lm_benchmark/prod/code/src/lexical_benchmark/train/train-params.toml +05/23/2025 18:48:13 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading char tokenizer +05/23/2025 18:48:13 - lexical_benchmark.train.trainers.gpt2 - INFO - Tokenizer loaded with vocabulary size: 115 +05/23/2025 18:48:13 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading and processing datasets with joined utterances (max_length=1024) +05/23/2025 18:48:13 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/05/00/train.tokenized +05/23/2025 18:48:13 - lexical_benchmark.train.tokenizers - INFO - Read 187919 lines from file +05/23/2025 18:48:29 - lexical_benchmark.train.tokenizers - INFO - Created 18264 joined examples from 187919 original lines +05/23/2025 18:48:32 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/dev/dev.tokenized +05/23/2025 18:48:32 - lexical_benchmark.train.tokenizers - INFO - Read 179806 lines from file +05/23/2025 18:48:48 - lexical_benchmark.train.tokenizers - INFO - Created 18373 joined examples from 179806 original lines +05/23/2025 18:48:51 - lexical_benchmark.train.trainers.gpt2 - INFO - Training dataset size: 18264 +05/23/2025 18:48:51 - lexical_benchmark.train.trainers.gpt2 - INFO - Validation dataset size: 18373 +05/23/2025 18:48:51 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating GPT2 configuration +05/23/2025 18:48:52 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating standard Trainer +05/23/2025 18:48:52 - lexical_benchmark.train.run_train - INFO - Running training of stela_gpt2_EN_05_00 +05/25/2025 22:46:20 - lexical_benchmark.train.run_train - INFO - Loading gpt2 trainer class. +05/25/2025 22:46:32 - datasets - INFO - PyTorch version 2.5.1 available. +05/25/2025 22:46:32 - datasets - INFO - Polars version 1.26.0 available. +05/25/2025 22:46:32 - lexical_benchmark.train.train_params - INFO - Loading params from /lustre/fsn1/projects/rech/hhb/uye44va/workspace/lm_benchmark/prod/code/src/lexical_benchmark/train/train-params.toml +05/25/2025 22:46:32 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading char tokenizer +05/25/2025 22:46:32 - lexical_benchmark.train.trainers.gpt2 - INFO - Tokenizer loaded with vocabulary size: 115 +05/25/2025 22:46:32 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading and processing datasets with joined utterances (max_length=1024) +05/25/2025 22:46:32 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/05/00/train.tokenized +05/25/2025 22:46:32 - lexical_benchmark.train.tokenizers - INFO - Read 187919 lines from file +05/25/2025 22:46:49 - lexical_benchmark.train.tokenizers - INFO - Created 18264 joined examples from 187919 original lines +05/25/2025 22:46:52 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/dev/dev.tokenized +05/25/2025 22:46:52 - lexical_benchmark.train.tokenizers - INFO - Read 179806 lines from file +05/25/2025 22:47:09 - lexical_benchmark.train.tokenizers - INFO - Created 18373 joined examples from 179806 original lines +05/25/2025 22:47:12 - lexical_benchmark.train.trainers.gpt2 - INFO - Training dataset size: 18264 +05/25/2025 22:47:12 - lexical_benchmark.train.trainers.gpt2 - INFO - Validation dataset size: 18373 +05/25/2025 22:47:12 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating GPT2 configuration +05/25/2025 22:47:12 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating standard Trainer +05/25/2025 22:47:13 - lexical_benchmark.train.run_train - INFO - Running training of stela_gpt2_EN_05_00 +05/26/2025 00:32:20 - lexical_benchmark.train.run_train - INFO - Loading gpt2 trainer class. +05/26/2025 00:32:33 - datasets - INFO - PyTorch version 2.5.1 available. +05/26/2025 00:32:33 - datasets - INFO - Polars version 1.26.0 available. +05/26/2025 00:32:33 - lexical_benchmark.train.train_params - INFO - Loading params from /lustre/fsn1/projects/rech/hhb/uye44va/workspace/lm_benchmark/prod/code/src/lexical_benchmark/train/train-params.toml +05/26/2025 00:32:33 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading char tokenizer +05/26/2025 00:32:33 - lexical_benchmark.train.trainers.gpt2 - INFO - Tokenizer loaded with vocabulary size: 115 +05/26/2025 00:32:33 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading and processing datasets with joined utterances (max_length=1024) +05/26/2025 00:32:33 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/05/00/train.tokenized +05/26/2025 00:32:33 - lexical_benchmark.train.tokenizers - INFO - Read 187919 lines from file +05/26/2025 00:32:50 - lexical_benchmark.train.tokenizers - INFO - Created 18264 joined examples from 187919 original lines +05/26/2025 00:32:53 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/dev/dev.tokenized +05/26/2025 00:32:53 - lexical_benchmark.train.tokenizers - INFO - Read 179806 lines from file +05/26/2025 00:33:10 - lexical_benchmark.train.tokenizers - INFO - Created 18373 joined examples from 179806 original lines +05/26/2025 00:33:13 - lexical_benchmark.train.trainers.gpt2 - INFO - Training dataset size: 18264 +05/26/2025 00:33:13 - lexical_benchmark.train.trainers.gpt2 - INFO - Validation dataset size: 18373 +05/26/2025 00:33:13 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating GPT2 configuration +05/26/2025 00:33:13 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating standard Trainer +05/26/2025 00:33:14 - lexical_benchmark.train.run_train - INFO - Running training of stela_gpt2_EN_05_00 +05/26/2025 00:38:30 - lexical_benchmark.train.run_train - INFO - Loading gpt2 trainer class. +05/26/2025 00:38:42 - datasets - INFO - PyTorch version 2.5.1 available. +05/26/2025 00:38:42 - datasets - INFO - Polars version 1.26.0 available. +05/26/2025 00:38:42 - lexical_benchmark.train.train_params - INFO - Loading params from /lustre/fsn1/projects/rech/hhb/uye44va/workspace/lm_benchmark/prod/code/src/lexical_benchmark/train/train-params.toml +05/26/2025 00:38:42 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading char tokenizer +05/26/2025 00:38:42 - lexical_benchmark.train.trainers.gpt2 - INFO - Tokenizer loaded with vocabulary size: 115 +05/26/2025 00:38:42 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading and processing datasets with joined utterances (max_length=1024) +05/26/2025 00:38:42 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/05/00/train.tokenized +05/26/2025 00:38:42 - lexical_benchmark.train.tokenizers - INFO - Read 187919 lines from file +05/26/2025 00:39:02 - lexical_benchmark.train.tokenizers - INFO - Created 18264 joined examples from 187919 original lines +05/26/2025 00:39:06 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/dev/dev.tokenized +05/26/2025 00:39:06 - lexical_benchmark.train.tokenizers - INFO - Read 179806 lines from file +05/26/2025 00:39:27 - lexical_benchmark.train.tokenizers - INFO - Created 18373 joined examples from 179806 original lines +05/26/2025 00:39:30 - lexical_benchmark.train.trainers.gpt2 - INFO - Training dataset size: 18264 +05/26/2025 00:39:30 - lexical_benchmark.train.trainers.gpt2 - INFO - Validation dataset size: 18373 +05/26/2025 00:39:30 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating GPT2 configuration +05/26/2025 00:39:31 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating standard Trainer +05/26/2025 00:39:31 - lexical_benchmark.train.run_train - INFO - Running training of stela_gpt2_EN_05_00 +05/26/2025 02:31:04 - lexical_benchmark.train.run_train - INFO - Training of stela_gpt2_EN_05_00 has completed ! +05/26/2025 02:31:04 - lexical_benchmark.train.run_train - INFO - Model saved to /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/models/stela/EN/05/00/gpt2 diff --git a/stela/EN/05/00/orthographic_gpt2/training_args.bin b/stela/EN/05/00/orthographic_gpt2/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..18c8d69cd73de778cb638341b46506c1df47f4cb --- /dev/null +++ b/stela/EN/05/00/orthographic_gpt2/training_args.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:83201b637b85ada900a879f80ff9987e272c84bae9d60bae387e464993f6fde8 +size 5496 diff --git a/stela/EN/05/00/orthographic_lstm/config.json b/stela/EN/05/00/orthographic_lstm/config.json new file mode 100644 index 0000000000000000000000000000000000000000..5ef629887c7fc87d64278c9788e493ac34c60370 --- /dev/null +++ b/stela/EN/05/00/orthographic_lstm/config.json @@ -0,0 +1,13 @@ +{ + "architectures": [ + "LSTMForLanguageModeling" + ], + "dropout": 0.1, + "embedding_dim": 200, + "hidden_size": 1024, + "model_type": "lstm", + "num_layers": 3, + "torch_dtype": "float32", + "transformers_version": "4.49.0", + "vocab_size": 115 +} diff --git a/stela/EN/05/00/orthographic_lstm/generation_config.json b/stela/EN/05/00/orthographic_lstm/generation_config.json new file mode 100644 index 0000000000000000000000000000000000000000..c1818a95e07bdccb83305ce8f9d878755e3c6119 --- /dev/null +++ b/stela/EN/05/00/orthographic_lstm/generation_config.json @@ -0,0 +1,4 @@ +{ + "_from_model_config": true, + "transformers_version": "4.49.0" +} diff --git a/stela/EN/05/00/orthographic_lstm/model.safetensors b/stela/EN/05/00/orthographic_lstm/model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..bd4f20a7f9ae018c7113425dd7be3e6098e4256a --- /dev/null +++ b/stela/EN/05/00/orthographic_lstm/model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:27af8ecda08a3ba085144f8cf3bcf6e988dce4916e18f7ca892623e5eab03ad5 +size 87825996 diff --git a/stela/EN/05/00/orthographic_lstm/training_args.bin b/stela/EN/05/00/orthographic_lstm/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..f0dba1940d5547376d9cb05f73e9cc7d1216dbc9 --- /dev/null +++ b/stela/EN/05/00/orthographic_lstm/training_args.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7cf2ca65bde897f648555625f473e000e1c033457969a6b27e63924a2e1fd564 +size 5496 diff --git a/stela/EN/05/01/orthographic_gpt2/config.json b/stela/EN/05/01/orthographic_gpt2/config.json new file mode 100644 index 0000000000000000000000000000000000000000..493011df283c5726a4be3aa8f39c84d43991636e --- /dev/null +++ b/stela/EN/05/01/orthographic_gpt2/config.json @@ -0,0 +1,32 @@ +{ + "activation_function": "gelu_new", + "architectures": [ + "GPT2LMHeadModel" + ], + "attn_pdrop": 0.1, + "bos_token_id": 50256, + "embd_pdrop": 0.1, + "eos_token_id": 50256, + "initializer_range": 0.02, + "layer_norm_epsilon": 1e-05, + "model_type": "gpt2", + "n_ctx": 1024, + "n_embd": 768, + "n_head": 12, + "n_inner": null, + "n_layer": 12, + "n_positions": 1024, + "reorder_and_upcast_attn": false, + "resid_pdrop": 0.1, + "scale_attn_by_inverse_layer_idx": false, + "scale_attn_weights": true, + "summary_activation": null, + "summary_first_dropout": 0.1, + "summary_proj_to_labels": true, + "summary_type": "cls_index", + "summary_use_proj": true, + "torch_dtype": "float32", + "transformers_version": "4.49.0", + "use_cache": true, + "vocab_size": 115 +} diff --git a/stela/EN/05/01/orthographic_gpt2/generation_config.json b/stela/EN/05/01/orthographic_gpt2/generation_config.json new file mode 100644 index 0000000000000000000000000000000000000000..bd968f46ce39e754a4603c69f53a2a4a08a13dbb --- /dev/null +++ b/stela/EN/05/01/orthographic_gpt2/generation_config.json @@ -0,0 +1,6 @@ +{ + "_from_model_config": true, + "bos_token_id": 50256, + "eos_token_id": 50256, + "transformers_version": "4.49.0" +} diff --git a/stela/EN/05/01/orthographic_gpt2/model.safetensors b/stela/EN/05/01/orthographic_gpt2/model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..b43e78e871543c3290c6ef059643d4df4e84ba98 --- /dev/null +++ b/stela/EN/05/01/orthographic_gpt2/model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:418614bf348f0f67d135f84752ba9e6c590e0f8ec18cbdea8a9ac9d8b1745885 +size 343737984 diff --git a/stela/EN/05/01/orthographic_gpt2/training.logs b/stela/EN/05/01/orthographic_gpt2/training.logs new file mode 100644 index 0000000000000000000000000000000000000000..303f14cba3b769718de0e72c3717b35cb97b0e21 --- /dev/null +++ b/stela/EN/05/01/orthographic_gpt2/training.logs @@ -0,0 +1,20 @@ +06/03/2025 16:47:29 - lexical_benchmark.train.run_train - INFO - Loading gpt2 trainer class. +06/03/2025 16:47:41 - datasets - INFO - PyTorch version 2.5.1 available. +06/03/2025 16:47:41 - datasets - INFO - Polars version 1.26.0 available. +06/03/2025 16:47:41 - lexical_benchmark.train.train_params - INFO - Loading params from /lustre/fsn1/projects/rech/hhb/uye44va/workspace/lm_benchmark/prod/code/src/lexical_benchmark/train/train-params.toml +06/03/2025 16:47:41 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading char tokenizer +06/03/2025 16:47:41 - lexical_benchmark.train.trainers.gpt2 - INFO - Tokenizer loaded with vocabulary size: 115 +06/03/2025 16:47:41 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading and processing datasets with joined utterances (max_length=1024) +06/03/2025 16:47:42 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/05/01/train.tokenized +06/03/2025 16:47:42 - lexical_benchmark.train.tokenizers - INFO - Read 194118 lines from file +06/03/2025 16:47:58 - lexical_benchmark.train.tokenizers - INFO - Created 18325 joined examples from 194118 original lines +06/03/2025 16:48:01 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/dev/dev.tokenized +06/03/2025 16:48:01 - lexical_benchmark.train.tokenizers - INFO - Read 179806 lines from file +06/03/2025 16:48:17 - lexical_benchmark.train.tokenizers - INFO - Created 18373 joined examples from 179806 original lines +06/03/2025 16:48:20 - lexical_benchmark.train.trainers.gpt2 - INFO - Training dataset size: 18325 +06/03/2025 16:48:20 - lexical_benchmark.train.trainers.gpt2 - INFO - Validation dataset size: 18373 +06/03/2025 16:48:20 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating GPT2 configuration +06/03/2025 16:48:21 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating standard Trainer +06/03/2025 16:48:21 - lexical_benchmark.train.run_train - INFO - Running training of stela_gpt2_EN_05_01 +06/03/2025 18:39:15 - lexical_benchmark.train.run_train - INFO - Training of stela_gpt2_EN_05_01 has completed ! +06/03/2025 18:39:15 - lexical_benchmark.train.run_train - INFO - Model saved to /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/models/stela/EN/05/01/gpt2 diff --git a/stela/EN/05/01/orthographic_gpt2/training_args.bin b/stela/EN/05/01/orthographic_gpt2/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..04e3b6bbd54b2baaffb098b30dfffd0de13a4c1a --- /dev/null +++ b/stela/EN/05/01/orthographic_gpt2/training_args.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9db7d2332a11991cb92579234ef7244f216890c67f84b4082834db8a3e896ef4 +size 5496 diff --git a/stela/EN/05/01/orthographic_lstm/config.json b/stela/EN/05/01/orthographic_lstm/config.json new file mode 100644 index 0000000000000000000000000000000000000000..5ef629887c7fc87d64278c9788e493ac34c60370 --- /dev/null +++ b/stela/EN/05/01/orthographic_lstm/config.json @@ -0,0 +1,13 @@ +{ + "architectures": [ + "LSTMForLanguageModeling" + ], + "dropout": 0.1, + "embedding_dim": 200, + "hidden_size": 1024, + "model_type": "lstm", + "num_layers": 3, + "torch_dtype": "float32", + "transformers_version": "4.49.0", + "vocab_size": 115 +} diff --git a/stela/EN/05/01/orthographic_lstm/generation_config.json b/stela/EN/05/01/orthographic_lstm/generation_config.json new file mode 100644 index 0000000000000000000000000000000000000000..c1818a95e07bdccb83305ce8f9d878755e3c6119 --- /dev/null +++ b/stela/EN/05/01/orthographic_lstm/generation_config.json @@ -0,0 +1,4 @@ +{ + "_from_model_config": true, + "transformers_version": "4.49.0" +} diff --git a/stela/EN/05/01/orthographic_lstm/model.safetensors b/stela/EN/05/01/orthographic_lstm/model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..4600b8269e42292e6f3d594284f0d636009ecf1b --- /dev/null +++ b/stela/EN/05/01/orthographic_lstm/model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1e4579e1e2a1cdcdc9427d5c32e7fb6602eeb4175dca95deab1ef70f75e08194 +size 87825996 diff --git a/stela/EN/05/01/orthographic_lstm/training.logs b/stela/EN/05/01/orthographic_lstm/training.logs new file mode 100644 index 0000000000000000000000000000000000000000..08d7dda1c3c82ad7bfc756578d6baccfa8e245ff --- /dev/null +++ b/stela/EN/05/01/orthographic_lstm/training.logs @@ -0,0 +1,19 @@ +06/03/2025 16:47:29 - lexical_benchmark.train.run_train - INFO - Loading lstm trainer class. +06/03/2025 16:47:42 - datasets - INFO - PyTorch version 2.5.1 available. +06/03/2025 16:47:42 - datasets - INFO - Polars version 1.26.0 available. +06/03/2025 16:47:42 - lexical_benchmark.train.train_params - INFO - Loading params from /lustre/fsn1/projects/rech/hhb/uye44va/workspace/lm_benchmark/prod/code/src/lexical_benchmark/train/train-params.toml +06/03/2025 16:47:42 - lexical_benchmark.train.trainers.lstm - INFO - Loading char-tokenizer +06/03/2025 16:47:42 - lexical_benchmark.train.trainers.lstm - INFO - Tokenizer loaded with vocabulary size: 115 +06/03/2025 16:47:42 - lexical_benchmark.train.trainers.lstm - INFO - Loading and processing datasets with joined utterances (max_length=1024) +06/03/2025 16:47:42 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/05/01/train.tokenized +06/03/2025 16:47:42 - lexical_benchmark.train.tokenizers - INFO - Read 194118 lines from file +06/03/2025 16:47:59 - lexical_benchmark.train.tokenizers - INFO - Created 18325 joined examples from 194118 original lines +06/03/2025 16:48:02 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/dev/dev.tokenized +06/03/2025 16:48:02 - lexical_benchmark.train.tokenizers - INFO - Read 179806 lines from file +06/03/2025 16:48:18 - lexical_benchmark.train.tokenizers - INFO - Created 18373 joined examples from 179806 original lines +06/03/2025 16:48:21 - lexical_benchmark.train.trainers.lstm - INFO - Training dataset size: 18325 +06/03/2025 16:48:21 - lexical_benchmark.train.trainers.lstm - INFO - Validation dataset size: 18373 +06/03/2025 16:48:21 - lexical_benchmark.train.trainers.lstm - INFO - Loading configurations & initialising LSTM model trainer +06/03/2025 16:48:21 - lexical_benchmark.train.run_train - INFO - Running training of stela_lstm_EN_05_01 +06/03/2025 20:28:25 - lexical_benchmark.train.run_train - INFO - Training of stela_lstm_EN_05_01 has completed ! +06/03/2025 20:28:25 - lexical_benchmark.train.run_train - INFO - Model saved to /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/models/stela/EN/05/01/lstm diff --git a/stela/EN/05/01/orthographic_lstm/training_args.bin b/stela/EN/05/01/orthographic_lstm/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..6b1b1ab8f2793bbab04c56eddf62bb7a3e9bd453 --- /dev/null +++ b/stela/EN/05/01/orthographic_lstm/training_args.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6de95a0c8efee999ae7682bf0e58bf03bafd3682610e3dfbb38720ea35cdf200 +size 5496 diff --git a/stela/EN/05/05/orthographic_gpt2/training.logs b/stela/EN/05/05/orthographic_gpt2/training.logs new file mode 100644 index 0000000000000000000000000000000000000000..1f80d4edc8c2a1169cfd03e829bb1edcf39b61cf --- /dev/null +++ b/stela/EN/05/05/orthographic_gpt2/training.logs @@ -0,0 +1,6 @@ +05/22/2025 20:09:40 - urllib3.connectionpool - WARNING - Retrying (Retry(total=2, connect=None, read=None, redirect=None, status=None)) after connection broken by 'NewConnectionError(': Failed to establish a new connection: [Errno 101] Network is unreachable')': /api/4504800232407040/envelope/ +05/22/2025 20:09:40 - urllib3.connectionpool - WARNING - Retrying (Retry(total=1, connect=None, read=None, redirect=None, status=None)) after connection broken by 'NewConnectionError(': Failed to establish a new connection: [Errno 101] Network is unreachable')': /api/4504800232407040/envelope/ +05/22/2025 20:09:40 - urllib3.connectionpool - WARNING - Retrying (Retry(total=0, connect=None, read=None, redirect=None, status=None)) after connection broken by 'NewConnectionError(': Failed to establish a new connection: [Errno 101] Network is unreachable')': /api/4504800232407040/envelope/ +05/22/2025 20:09:40 - urllib3.connectionpool - WARNING - Retrying (Retry(total=2, connect=None, read=None, redirect=None, status=None)) after connection broken by 'NewConnectionError(': Failed to establish a new connection: [Errno 101] Network is unreachable')': /api/4504800232407040/envelope/ +05/22/2025 20:09:40 - urllib3.connectionpool - WARNING - Retrying (Retry(total=1, connect=None, read=None, redirect=None, status=None)) after connection broken by 'NewConnectionError(': Failed to establish a new connection: [Errno 101] Network is unreachable')': /api/4504800232407040/envelope/ +05/22/2025 20:09:40 - urllib3.connectionpool - WARNING - Retrying (Retry(total=0, connect=None, read=None, redirect=None, status=None)) after connection broken by 'NewConnectionError(': Failed to establish a new connection: [Errno 101] Network is unreachable')': /api/4504800232407040/envelope/ diff --git a/stela/EN/06/00/orthographic_gpt2/config.json b/stela/EN/06/00/orthographic_gpt2/config.json new file mode 100644 index 0000000000000000000000000000000000000000..493011df283c5726a4be3aa8f39c84d43991636e --- /dev/null +++ b/stela/EN/06/00/orthographic_gpt2/config.json @@ -0,0 +1,32 @@ +{ + "activation_function": "gelu_new", + "architectures": [ + "GPT2LMHeadModel" + ], + "attn_pdrop": 0.1, + "bos_token_id": 50256, + "embd_pdrop": 0.1, + "eos_token_id": 50256, + "initializer_range": 0.02, + "layer_norm_epsilon": 1e-05, + "model_type": "gpt2", + "n_ctx": 1024, + "n_embd": 768, + "n_head": 12, + "n_inner": null, + "n_layer": 12, + "n_positions": 1024, + "reorder_and_upcast_attn": false, + "resid_pdrop": 0.1, + "scale_attn_by_inverse_layer_idx": false, + "scale_attn_weights": true, + "summary_activation": null, + "summary_first_dropout": 0.1, + "summary_proj_to_labels": true, + "summary_type": "cls_index", + "summary_use_proj": true, + "torch_dtype": "float32", + "transformers_version": "4.49.0", + "use_cache": true, + "vocab_size": 115 +} diff --git a/stela/EN/06/00/orthographic_gpt2/generation_config.json b/stela/EN/06/00/orthographic_gpt2/generation_config.json new file mode 100644 index 0000000000000000000000000000000000000000..bd968f46ce39e754a4603c69f53a2a4a08a13dbb --- /dev/null +++ b/stela/EN/06/00/orthographic_gpt2/generation_config.json @@ -0,0 +1,6 @@ +{ + "_from_model_config": true, + "bos_token_id": 50256, + "eos_token_id": 50256, + "transformers_version": "4.49.0" +} diff --git a/stela/EN/06/00/orthographic_gpt2/model.safetensors b/stela/EN/06/00/orthographic_gpt2/model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..b34615be7fde567c5e7b9a8b1e2425410dd83191 --- /dev/null +++ b/stela/EN/06/00/orthographic_gpt2/model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:485344ca54174a7955156f5dee40e617dd2b98b7fe5520f9713eb40712e5e6d3 +size 343737984 diff --git a/stela/EN/06/00/orthographic_gpt2/training.logs b/stela/EN/06/00/orthographic_gpt2/training.logs new file mode 100644 index 0000000000000000000000000000000000000000..6f1ba2cf37ecb86d43d91f7fd3477220bfabf7c4 --- /dev/null +++ b/stela/EN/06/00/orthographic_gpt2/training.logs @@ -0,0 +1,170 @@ +04/30/2025 18:21:35 - lexical_benchmark.train.run_train - INFO - Loading gpt2 trainer class. +04/30/2025 18:21:47 - datasets - INFO - PyTorch version 2.5.1 available. +04/30/2025 18:21:47 - datasets - INFO - Polars version 1.26.0 available. +04/30/2025 18:21:48 - lexical_benchmark.train.train_params - INFO - Loading params from /lustre/fsn1/projects/rech/hhb/uye44va/workspace/lm_benchmark/prod/code/src/lexical_benchmark/train/train-params.toml +04/30/2025 18:21:48 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading char tokenizer +04/30/2025 18:21:48 - lexical_benchmark.train.trainers.gpt2 - INFO - Tokenizer loaded with vocabulary size: 115 +04/30/2025 18:21:48 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading and processing datasets with joined utterances (max_length=1024) +04/30/2025 18:21:48 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/06/06/train.tokenized +04/30/2025 18:21:48 - lexical_benchmark.train.tokenizers - INFO - Read 229786 lines from file +04/30/2025 18:22:10 - lexical_benchmark.train.tokenizers - INFO - Created 22176 joined examples from 229786 original lines +04/30/2025 18:22:13 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/dev/dev.tokenized +04/30/2025 18:22:13 - lexical_benchmark.train.tokenizers - INFO - Read 179806 lines from file +04/30/2025 18:22:31 - lexical_benchmark.train.tokenizers - INFO - Created 18373 joined examples from 179806 original lines +04/30/2025 18:22:34 - lexical_benchmark.train.trainers.gpt2 - INFO - Training dataset size: 22176 +04/30/2025 18:22:34 - lexical_benchmark.train.trainers.gpt2 - INFO - Validation dataset size: 18373 +04/30/2025 18:22:34 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating GPT2 configuration +04/30/2025 18:22:35 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating standard Trainer +04/30/2025 18:22:35 - lexical_benchmark.train.run_train - INFO - Running training of stela_gpt2_EN_06_06 +04/30/2025 19:27:35 - urllib3.connectionpool - WARNING - Retrying (Retry(total=2, connect=None, read=None, redirect=None, status=None)) after connection broken by 'NewConnectionError(': Failed to establish a new connection: [Errno 101] Network is unreachable')': /api/4504800232407040/envelope/ +04/30/2025 19:27:35 - urllib3.connectionpool - WARNING - Retrying (Retry(total=1, connect=None, read=None, redirect=None, status=None)) after connection broken by 'NewConnectionError(': Failed to establish a new connection: [Errno 101] Network is unreachable')': /api/4504800232407040/envelope/ +04/30/2025 19:27:35 - urllib3.connectionpool - WARNING - Retrying (Retry(total=0, connect=None, read=None, redirect=None, status=None)) after connection broken by 'NewConnectionError(': Failed to establish a new connection: [Errno 101] Network is unreachable')': /api/4504800232407040/envelope/ +04/30/2025 19:27:35 - urllib3.connectionpool - WARNING - Retrying (Retry(total=2, connect=None, read=None, redirect=None, status=None)) after connection broken by 'NewConnectionError(': Failed to establish a new connection: [Errno 101] Network is unreachable')': /api/4504800232407040/envelope/ +04/30/2025 19:27:35 - urllib3.connectionpool - WARNING - Retrying (Retry(total=1, connect=None, read=None, redirect=None, status=None)) after connection broken by 'NewConnectionError(': Failed to establish a new connection: [Errno 101] Network is unreachable')': /api/4504800232407040/envelope/ +04/30/2025 19:27:35 - urllib3.connectionpool - WARNING - Retrying (Retry(total=0, connect=None, read=None, redirect=None, status=None)) after connection broken by 'NewConnectionError(': Failed to establish a new connection: [Errno 101] Network is unreachable')': /api/4504800232407040/envelope/ +04/30/2025 19:58:17 - lexical_benchmark.train.run_train - INFO - Loading gpt2 trainer class. +04/30/2025 19:58:30 - datasets - INFO - PyTorch version 2.5.1 available. +04/30/2025 19:58:30 - datasets - INFO - Polars version 1.26.0 available. +04/30/2025 19:58:30 - lexical_benchmark.train.train_params - INFO - Loading params from /lustre/fsn1/projects/rech/hhb/ucx81cx/workspace/lm_benchmark/dev/code/src/lexical_benchmark/train/train-params.toml +04/30/2025 19:58:30 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading char tokenizer +04/30/2025 19:58:30 - lexical_benchmark.train.trainers.gpt2 - INFO - Tokenizer loaded with vocabulary size: 115 +04/30/2025 19:58:30 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading and processing datasets with joined utterances (max_length=1024) +04/30/2025 19:58:30 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/06/06/train.tokenized +04/30/2025 19:58:30 - lexical_benchmark.train.tokenizers - INFO - Read 229786 lines from file +04/30/2025 19:58:50 - lexical_benchmark.train.tokenizers - INFO - Created 22176 joined examples from 229786 original lines +04/30/2025 19:58:54 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/dev/dev.tokenized +04/30/2025 19:58:54 - lexical_benchmark.train.tokenizers - INFO - Read 179806 lines from file +04/30/2025 19:59:10 - lexical_benchmark.train.tokenizers - INFO - Created 18373 joined examples from 179806 original lines +04/30/2025 19:59:13 - lexical_benchmark.train.trainers.gpt2 - INFO - Training dataset size: 22176 +04/30/2025 19:59:13 - lexical_benchmark.train.trainers.gpt2 - INFO - Validation dataset size: 18373 +04/30/2025 19:59:13 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating GPT2 configuration +04/30/2025 19:59:14 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating standard Trainer +04/30/2025 19:59:14 - lexical_benchmark.train.run_train - INFO - Running training of stela_gpt2_EN_06_06 +04/30/2025 20:08:33 - lexical_benchmark.train.run_train - INFO - Loading gpt2 trainer class. +04/30/2025 20:08:45 - datasets - INFO - PyTorch version 2.5.1 available. +04/30/2025 20:08:45 - datasets - INFO - Polars version 1.26.0 available. +04/30/2025 20:08:46 - lexical_benchmark.train.train_params - INFO - Loading params from /lustre/fsn1/projects/rech/hhb/ucx81cx/workspace/lm_benchmark/dev/code/src/lexical_benchmark/train/train-params.toml +04/30/2025 20:08:46 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading char tokenizer +04/30/2025 20:08:46 - lexical_benchmark.train.trainers.gpt2 - INFO - Tokenizer loaded with vocabulary size: 115 +04/30/2025 20:08:46 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading and processing datasets with joined utterances (max_length=1024) +04/30/2025 20:08:46 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/06/06/train.tokenized +04/30/2025 20:08:46 - lexical_benchmark.train.tokenizers - INFO - Read 229786 lines from file +04/30/2025 20:09:06 - lexical_benchmark.train.tokenizers - INFO - Created 22176 joined examples from 229786 original lines +04/30/2025 20:09:09 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/dev/dev.tokenized +04/30/2025 20:09:09 - lexical_benchmark.train.tokenizers - INFO - Read 179806 lines from file +04/30/2025 20:09:27 - lexical_benchmark.train.tokenizers - INFO - Created 18373 joined examples from 179806 original lines +04/30/2025 20:09:30 - lexical_benchmark.train.trainers.gpt2 - INFO - Training dataset size: 22176 +04/30/2025 20:09:30 - lexical_benchmark.train.trainers.gpt2 - INFO - Validation dataset size: 18373 +04/30/2025 20:09:30 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating GPT2 configuration +04/30/2025 20:09:31 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating standard Trainer +04/30/2025 20:09:31 - lexical_benchmark.train.run_train - INFO - Running training of stela_gpt2_EN_06_06 +05/23/2025 11:45:14 - lexical_benchmark.train.run_train - INFO - Loading gpt2 trainer class. +05/23/2025 11:45:27 - datasets - INFO - PyTorch version 2.5.1 available. +05/23/2025 11:45:27 - datasets - INFO - Polars version 1.26.0 available. +05/23/2025 11:45:28 - lexical_benchmark.train.train_params - INFO - Loading params from /lustre/fsn1/projects/rech/hhb/uye44va/workspace/lm_benchmark/prod/code/src/lexical_benchmark/train/train-params.toml +05/23/2025 11:45:28 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading char tokenizer +05/23/2025 11:45:28 - lexical_benchmark.train.trainers.gpt2 - INFO - Tokenizer loaded with vocabulary size: 115 +05/23/2025 11:45:28 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading and processing datasets with joined utterances (max_length=1024) +05/23/2025 11:45:29 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/06/00/train.tokenized +05/23/2025 11:45:29 - lexical_benchmark.train.tokenizers - INFO - Read 225915 lines from file +05/23/2025 11:45:48 - lexical_benchmark.train.tokenizers - INFO - Created 21892 joined examples from 225915 original lines +05/23/2025 11:45:52 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/dev/dev.tokenized +05/23/2025 11:45:52 - lexical_benchmark.train.tokenizers - INFO - Read 179806 lines from file +05/23/2025 11:46:08 - lexical_benchmark.train.tokenizers - INFO - Created 18373 joined examples from 179806 original lines +05/23/2025 11:46:10 - lexical_benchmark.train.trainers.gpt2 - INFO - Training dataset size: 21892 +05/23/2025 11:46:10 - lexical_benchmark.train.trainers.gpt2 - INFO - Validation dataset size: 18373 +05/23/2025 11:46:10 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating GPT2 configuration +05/23/2025 11:46:11 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating standard Trainer +05/23/2025 11:46:11 - lexical_benchmark.train.run_train - INFO - Running training of stela_gpt2_EN_06_00 +05/23/2025 17:33:51 - lexical_benchmark.train.run_train - INFO - Loading gpt2 trainer class. +05/23/2025 17:34:03 - datasets - INFO - PyTorch version 2.5.1 available. +05/23/2025 17:34:03 - datasets - INFO - Polars version 1.26.0 available. +05/23/2025 17:34:03 - lexical_benchmark.train.train_params - INFO - Loading params from /lustre/fsn1/projects/rech/hhb/uye44va/workspace/lm_benchmark/prod/code/src/lexical_benchmark/train/train-params.toml +05/23/2025 17:34:03 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading char tokenizer +05/23/2025 17:34:03 - lexical_benchmark.train.trainers.gpt2 - INFO - Tokenizer loaded with vocabulary size: 115 +05/23/2025 17:34:03 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading and processing datasets with joined utterances (max_length=1024) +05/23/2025 17:34:03 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/06/00/train.tokenized +05/23/2025 17:34:03 - lexical_benchmark.train.tokenizers - INFO - Read 225915 lines from file +05/23/2025 17:34:23 - lexical_benchmark.train.tokenizers - INFO - Created 21892 joined examples from 225915 original lines +05/23/2025 17:34:26 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/dev/dev.tokenized +05/23/2025 17:34:26 - lexical_benchmark.train.tokenizers - INFO - Read 179806 lines from file +05/23/2025 17:34:43 - lexical_benchmark.train.tokenizers - INFO - Created 18373 joined examples from 179806 original lines +05/23/2025 17:34:45 - lexical_benchmark.train.trainers.gpt2 - INFO - Training dataset size: 21892 +05/23/2025 17:34:45 - lexical_benchmark.train.trainers.gpt2 - INFO - Validation dataset size: 18373 +05/23/2025 17:34:45 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating GPT2 configuration +05/23/2025 17:34:46 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating standard Trainer +05/23/2025 17:34:46 - lexical_benchmark.train.run_train - INFO - Running training of stela_gpt2_EN_06_00 +05/23/2025 18:48:00 - lexical_benchmark.train.run_train - INFO - Loading gpt2 trainer class. +05/23/2025 18:48:12 - datasets - INFO - PyTorch version 2.5.1 available. +05/23/2025 18:48:12 - datasets - INFO - Polars version 1.26.0 available. +05/23/2025 18:48:13 - lexical_benchmark.train.train_params - INFO - Loading params from /lustre/fsn1/projects/rech/hhb/uye44va/workspace/lm_benchmark/prod/code/src/lexical_benchmark/train/train-params.toml +05/23/2025 18:48:13 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading char tokenizer +05/23/2025 18:48:13 - lexical_benchmark.train.trainers.gpt2 - INFO - Tokenizer loaded with vocabulary size: 115 +05/23/2025 18:48:13 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading and processing datasets with joined utterances (max_length=1024) +05/23/2025 18:48:13 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/06/00/train.tokenized +05/23/2025 18:48:13 - lexical_benchmark.train.tokenizers - INFO - Read 225915 lines from file +05/23/2025 18:48:33 - lexical_benchmark.train.tokenizers - INFO - Created 21892 joined examples from 225915 original lines +05/23/2025 18:48:36 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/dev/dev.tokenized +05/23/2025 18:48:36 - lexical_benchmark.train.tokenizers - INFO - Read 179806 lines from file +05/23/2025 18:48:53 - lexical_benchmark.train.tokenizers - INFO - Created 18373 joined examples from 179806 original lines +05/23/2025 18:48:56 - lexical_benchmark.train.trainers.gpt2 - INFO - Training dataset size: 21892 +05/23/2025 18:48:56 - lexical_benchmark.train.trainers.gpt2 - INFO - Validation dataset size: 18373 +05/23/2025 18:48:56 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating GPT2 configuration +05/23/2025 18:48:57 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating standard Trainer +05/23/2025 18:48:57 - lexical_benchmark.train.run_train - INFO - Running training of stela_gpt2_EN_06_00 +05/25/2025 22:46:20 - lexical_benchmark.train.run_train - INFO - Loading gpt2 trainer class. +05/25/2025 22:46:31 - datasets - INFO - PyTorch version 2.5.1 available. +05/25/2025 22:46:31 - datasets - INFO - Polars version 1.26.0 available. +05/25/2025 22:46:32 - lexical_benchmark.train.train_params - INFO - Loading params from /lustre/fsn1/projects/rech/hhb/uye44va/workspace/lm_benchmark/prod/code/src/lexical_benchmark/train/train-params.toml +05/25/2025 22:46:32 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading char tokenizer +05/25/2025 22:46:32 - lexical_benchmark.train.trainers.gpt2 - INFO - Tokenizer loaded with vocabulary size: 115 +05/25/2025 22:46:32 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading and processing datasets with joined utterances (max_length=1024) +05/25/2025 22:46:32 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/06/00/train.tokenized +05/25/2025 22:46:32 - lexical_benchmark.train.tokenizers - INFO - Read 225915 lines from file +05/25/2025 22:46:53 - lexical_benchmark.train.tokenizers - INFO - Created 21892 joined examples from 225915 original lines +05/25/2025 22:46:56 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/dev/dev.tokenized +05/25/2025 22:46:56 - lexical_benchmark.train.tokenizers - INFO - Read 179806 lines from file +05/25/2025 22:47:14 - lexical_benchmark.train.tokenizers - INFO - Created 18373 joined examples from 179806 original lines +05/25/2025 22:47:16 - lexical_benchmark.train.trainers.gpt2 - INFO - Training dataset size: 21892 +05/25/2025 22:47:16 - lexical_benchmark.train.trainers.gpt2 - INFO - Validation dataset size: 18373 +05/25/2025 22:47:16 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating GPT2 configuration +05/25/2025 22:47:17 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating standard Trainer +05/25/2025 22:47:18 - lexical_benchmark.train.run_train - INFO - Running training of stela_gpt2_EN_06_00 +05/26/2025 00:32:20 - lexical_benchmark.train.run_train - INFO - Loading gpt2 trainer class. +05/26/2025 00:32:34 - datasets - INFO - PyTorch version 2.5.1 available. +05/26/2025 00:32:34 - datasets - INFO - Polars version 1.26.0 available. +05/26/2025 00:32:34 - lexical_benchmark.train.train_params - INFO - Loading params from /lustre/fsn1/projects/rech/hhb/uye44va/workspace/lm_benchmark/prod/code/src/lexical_benchmark/train/train-params.toml +05/26/2025 00:32:34 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading char tokenizer +05/26/2025 00:32:34 - lexical_benchmark.train.trainers.gpt2 - INFO - Tokenizer loaded with vocabulary size: 115 +05/26/2025 00:32:34 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading and processing datasets with joined utterances (max_length=1024) +05/26/2025 00:32:34 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/06/00/train.tokenized +05/26/2025 00:32:34 - lexical_benchmark.train.tokenizers - INFO - Read 225915 lines from file +05/26/2025 00:32:54 - lexical_benchmark.train.tokenizers - INFO - Created 21892 joined examples from 225915 original lines +05/26/2025 00:32:57 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/dev/dev.tokenized +05/26/2025 00:32:57 - lexical_benchmark.train.tokenizers - INFO - Read 179806 lines from file +05/26/2025 00:33:14 - lexical_benchmark.train.tokenizers - INFO - Created 18373 joined examples from 179806 original lines +05/26/2025 00:33:17 - lexical_benchmark.train.trainers.gpt2 - INFO - Training dataset size: 21892 +05/26/2025 00:33:17 - lexical_benchmark.train.trainers.gpt2 - INFO - Validation dataset size: 18373 +05/26/2025 00:33:17 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating GPT2 configuration +05/26/2025 00:33:17 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating standard Trainer +05/26/2025 00:33:18 - lexical_benchmark.train.run_train - INFO - Running training of stela_gpt2_EN_06_00 +05/26/2025 00:38:30 - lexical_benchmark.train.run_train - INFO - Loading gpt2 trainer class. +05/26/2025 00:38:42 - datasets - INFO - PyTorch version 2.5.1 available. +05/26/2025 00:38:42 - datasets - INFO - Polars version 1.26.0 available. +05/26/2025 00:38:42 - lexical_benchmark.train.train_params - INFO - Loading params from /lustre/fsn1/projects/rech/hhb/uye44va/workspace/lm_benchmark/prod/code/src/lexical_benchmark/train/train-params.toml +05/26/2025 00:38:42 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading char tokenizer +05/26/2025 00:38:42 - lexical_benchmark.train.trainers.gpt2 - INFO - Tokenizer loaded with vocabulary size: 115 +05/26/2025 00:38:42 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading and processing datasets with joined utterances (max_length=1024) +05/26/2025 00:38:42 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/06/00/train.tokenized +05/26/2025 00:38:42 - lexical_benchmark.train.tokenizers - INFO - Read 225915 lines from file +05/26/2025 00:39:06 - lexical_benchmark.train.tokenizers - INFO - Created 21892 joined examples from 225915 original lines +05/26/2025 00:39:10 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/dev/dev.tokenized +05/26/2025 00:39:10 - lexical_benchmark.train.tokenizers - INFO - Read 179806 lines from file +05/26/2025 00:39:31 - lexical_benchmark.train.tokenizers - INFO - Created 18373 joined examples from 179806 original lines +05/26/2025 00:39:34 - lexical_benchmark.train.trainers.gpt2 - INFO - Training dataset size: 21892 +05/26/2025 00:39:34 - lexical_benchmark.train.trainers.gpt2 - INFO - Validation dataset size: 18373 +05/26/2025 00:39:34 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating GPT2 configuration +05/26/2025 00:39:35 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating standard Trainer +05/26/2025 00:39:35 - lexical_benchmark.train.run_train - INFO - Running training of stela_gpt2_EN_06_00 +05/26/2025 02:46:29 - lexical_benchmark.train.run_train - INFO - Training of stela_gpt2_EN_06_00 has completed ! +05/26/2025 02:46:30 - lexical_benchmark.train.run_train - INFO - Model saved to /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/models/stela/EN/06/00/gpt2 diff --git a/stela/EN/06/00/orthographic_gpt2/training_args.bin b/stela/EN/06/00/orthographic_gpt2/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..0d2d067ce4fcd5f55377c5dd8250c352043e1fc1 --- /dev/null +++ b/stela/EN/06/00/orthographic_gpt2/training_args.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1024b81f2d6a13380ceeb2866bbaac24e2792dc277b0b32fc3399264b84d2030 +size 5496 diff --git a/stela/EN/06/00/orthographic_lstm/config.json b/stela/EN/06/00/orthographic_lstm/config.json new file mode 100644 index 0000000000000000000000000000000000000000..5ef629887c7fc87d64278c9788e493ac34c60370 --- /dev/null +++ b/stela/EN/06/00/orthographic_lstm/config.json @@ -0,0 +1,13 @@ +{ + "architectures": [ + "LSTMForLanguageModeling" + ], + "dropout": 0.1, + "embedding_dim": 200, + "hidden_size": 1024, + "model_type": "lstm", + "num_layers": 3, + "torch_dtype": "float32", + "transformers_version": "4.49.0", + "vocab_size": 115 +} diff --git a/stela/EN/06/00/orthographic_lstm/generation_config.json b/stela/EN/06/00/orthographic_lstm/generation_config.json new file mode 100644 index 0000000000000000000000000000000000000000..c1818a95e07bdccb83305ce8f9d878755e3c6119 --- /dev/null +++ b/stela/EN/06/00/orthographic_lstm/generation_config.json @@ -0,0 +1,4 @@ +{ + "_from_model_config": true, + "transformers_version": "4.49.0" +} diff --git a/stela/EN/06/00/orthographic_lstm/model.safetensors b/stela/EN/06/00/orthographic_lstm/model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..1bafd8696082f15230795a29bb2f0d241bd0eba8 --- /dev/null +++ b/stela/EN/06/00/orthographic_lstm/model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9833334a6853f9c1f41f8728e22e7705a0d1b2c88a2f1a0c8c1bea7089cb63d0 +size 87825996 diff --git a/stela/EN/06/00/orthographic_lstm/training_args.bin b/stela/EN/06/00/orthographic_lstm/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..143ab9af83feeecd681fbfafbc13797395a8875a --- /dev/null +++ b/stela/EN/06/00/orthographic_lstm/training_args.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:07241b2a21d2015804806052cc890948cdbe8d6afe6b841dd978bfb7d140ef5e +size 5496 diff --git a/stela/EN/06/01/orthographic_gpt2/config.json b/stela/EN/06/01/orthographic_gpt2/config.json new file mode 100644 index 0000000000000000000000000000000000000000..493011df283c5726a4be3aa8f39c84d43991636e --- /dev/null +++ b/stela/EN/06/01/orthographic_gpt2/config.json @@ -0,0 +1,32 @@ +{ + "activation_function": "gelu_new", + "architectures": [ + "GPT2LMHeadModel" + ], + "attn_pdrop": 0.1, + "bos_token_id": 50256, + "embd_pdrop": 0.1, + "eos_token_id": 50256, + "initializer_range": 0.02, + "layer_norm_epsilon": 1e-05, + "model_type": "gpt2", + "n_ctx": 1024, + "n_embd": 768, + "n_head": 12, + "n_inner": null, + "n_layer": 12, + "n_positions": 1024, + "reorder_and_upcast_attn": false, + "resid_pdrop": 0.1, + "scale_attn_by_inverse_layer_idx": false, + "scale_attn_weights": true, + "summary_activation": null, + "summary_first_dropout": 0.1, + "summary_proj_to_labels": true, + "summary_type": "cls_index", + "summary_use_proj": true, + "torch_dtype": "float32", + "transformers_version": "4.49.0", + "use_cache": true, + "vocab_size": 115 +} diff --git a/stela/EN/06/01/orthographic_gpt2/generation_config.json b/stela/EN/06/01/orthographic_gpt2/generation_config.json new file mode 100644 index 0000000000000000000000000000000000000000..bd968f46ce39e754a4603c69f53a2a4a08a13dbb --- /dev/null +++ b/stela/EN/06/01/orthographic_gpt2/generation_config.json @@ -0,0 +1,6 @@ +{ + "_from_model_config": true, + "bos_token_id": 50256, + "eos_token_id": 50256, + "transformers_version": "4.49.0" +} diff --git a/stela/EN/06/01/orthographic_gpt2/model.safetensors b/stela/EN/06/01/orthographic_gpt2/model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..b2be78fe6c4dd58dfdb3ea2e650b9ba0b518703b --- /dev/null +++ b/stela/EN/06/01/orthographic_gpt2/model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4b45bec072468f1fe2fa1cd746a7412f94322a2be963351b65771f42a383fd5d +size 343737984 diff --git a/stela/EN/06/01/orthographic_gpt2/training.logs b/stela/EN/06/01/orthographic_gpt2/training.logs new file mode 100644 index 0000000000000000000000000000000000000000..40aba14c793669957744a59b92a916fe1894419f --- /dev/null +++ b/stela/EN/06/01/orthographic_gpt2/training.logs @@ -0,0 +1,20 @@ +06/03/2025 16:47:29 - lexical_benchmark.train.run_train - INFO - Loading gpt2 trainer class. +06/03/2025 16:47:41 - datasets - INFO - PyTorch version 2.5.1 available. +06/03/2025 16:47:41 - datasets - INFO - Polars version 1.26.0 available. +06/03/2025 16:47:41 - lexical_benchmark.train.train_params - INFO - Loading params from /lustre/fsn1/projects/rech/hhb/uye44va/workspace/lm_benchmark/prod/code/src/lexical_benchmark/train/train-params.toml +06/03/2025 16:47:41 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading char tokenizer +06/03/2025 16:47:41 - lexical_benchmark.train.trainers.gpt2 - INFO - Tokenizer loaded with vocabulary size: 115 +06/03/2025 16:47:41 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading and processing datasets with joined utterances (max_length=1024) +06/03/2025 16:47:42 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/06/01/train.tokenized +06/03/2025 16:47:42 - lexical_benchmark.train.tokenizers - INFO - Read 231862 lines from file +06/03/2025 16:48:02 - lexical_benchmark.train.tokenizers - INFO - Created 21956 joined examples from 231862 original lines +06/03/2025 16:48:05 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/dev/dev.tokenized +06/03/2025 16:48:05 - lexical_benchmark.train.tokenizers - INFO - Read 179806 lines from file +06/03/2025 16:48:22 - lexical_benchmark.train.tokenizers - INFO - Created 18373 joined examples from 179806 original lines +06/03/2025 16:48:24 - lexical_benchmark.train.trainers.gpt2 - INFO - Training dataset size: 21956 +06/03/2025 16:48:24 - lexical_benchmark.train.trainers.gpt2 - INFO - Validation dataset size: 18373 +06/03/2025 16:48:24 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating GPT2 configuration +06/03/2025 16:48:25 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating standard Trainer +06/03/2025 16:48:25 - lexical_benchmark.train.run_train - INFO - Running training of stela_gpt2_EN_06_01 +06/03/2025 18:39:47 - lexical_benchmark.train.run_train - INFO - Training of stela_gpt2_EN_06_01 has completed ! +06/03/2025 18:39:47 - lexical_benchmark.train.run_train - INFO - Model saved to /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/models/stela/EN/06/01/gpt2 diff --git a/stela/EN/06/01/orthographic_gpt2/training_args.bin b/stela/EN/06/01/orthographic_gpt2/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..d2ae73306f7ca073278c286e38f15af206a9ed85 --- /dev/null +++ b/stela/EN/06/01/orthographic_gpt2/training_args.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9403eb82b56f4904a474e5171fc4709d7955aa80d5b788e913a33d8781fb4305 +size 5496 diff --git a/stela/EN/06/01/orthographic_lstm/config.json b/stela/EN/06/01/orthographic_lstm/config.json new file mode 100644 index 0000000000000000000000000000000000000000..5ef629887c7fc87d64278c9788e493ac34c60370 --- /dev/null +++ b/stela/EN/06/01/orthographic_lstm/config.json @@ -0,0 +1,13 @@ +{ + "architectures": [ + "LSTMForLanguageModeling" + ], + "dropout": 0.1, + "embedding_dim": 200, + "hidden_size": 1024, + "model_type": "lstm", + "num_layers": 3, + "torch_dtype": "float32", + "transformers_version": "4.49.0", + "vocab_size": 115 +} diff --git a/stela/EN/06/01/orthographic_lstm/generation_config.json b/stela/EN/06/01/orthographic_lstm/generation_config.json new file mode 100644 index 0000000000000000000000000000000000000000..c1818a95e07bdccb83305ce8f9d878755e3c6119 --- /dev/null +++ b/stela/EN/06/01/orthographic_lstm/generation_config.json @@ -0,0 +1,4 @@ +{ + "_from_model_config": true, + "transformers_version": "4.49.0" +} diff --git a/stela/EN/06/01/orthographic_lstm/model.safetensors b/stela/EN/06/01/orthographic_lstm/model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..470d037b15e301826d1e2ea99474e117b64ab5ea --- /dev/null +++ b/stela/EN/06/01/orthographic_lstm/model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:08aebda5a15fdedd7a7b30ba39c6974f99ddec8f1e9f315ec1555c47183458ce +size 87825996 diff --git a/stela/EN/06/01/orthographic_lstm/training.logs b/stela/EN/06/01/orthographic_lstm/training.logs new file mode 100644 index 0000000000000000000000000000000000000000..4201ee6188badb826100ca825759fd9501f61328 --- /dev/null +++ b/stela/EN/06/01/orthographic_lstm/training.logs @@ -0,0 +1,19 @@ +06/03/2025 16:47:29 - lexical_benchmark.train.run_train - INFO - Loading lstm trainer class. +06/03/2025 16:47:41 - datasets - INFO - PyTorch version 2.5.1 available. +06/03/2025 16:47:41 - datasets - INFO - Polars version 1.26.0 available. +06/03/2025 16:47:41 - lexical_benchmark.train.train_params - INFO - Loading params from /lustre/fsn1/projects/rech/hhb/uye44va/workspace/lm_benchmark/prod/code/src/lexical_benchmark/train/train-params.toml +06/03/2025 16:47:41 - lexical_benchmark.train.trainers.lstm - INFO - Loading char-tokenizer +06/03/2025 16:47:41 - lexical_benchmark.train.trainers.lstm - INFO - Tokenizer loaded with vocabulary size: 115 +06/03/2025 16:47:41 - lexical_benchmark.train.trainers.lstm - INFO - Loading and processing datasets with joined utterances (max_length=1024) +06/03/2025 16:47:42 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/06/01/train.tokenized +06/03/2025 16:47:42 - lexical_benchmark.train.tokenizers - INFO - Read 231862 lines from file +06/03/2025 16:48:02 - lexical_benchmark.train.tokenizers - INFO - Created 21956 joined examples from 231862 original lines +06/03/2025 16:48:05 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/dev/dev.tokenized +06/03/2025 16:48:05 - lexical_benchmark.train.tokenizers - INFO - Read 179806 lines from file +06/03/2025 16:48:21 - lexical_benchmark.train.tokenizers - INFO - Created 18373 joined examples from 179806 original lines +06/03/2025 16:48:24 - lexical_benchmark.train.trainers.lstm - INFO - Training dataset size: 21956 +06/03/2025 16:48:24 - lexical_benchmark.train.trainers.lstm - INFO - Validation dataset size: 18373 +06/03/2025 16:48:24 - lexical_benchmark.train.trainers.lstm - INFO - Loading configurations & initialising LSTM model trainer +06/03/2025 16:48:25 - lexical_benchmark.train.run_train - INFO - Running training of stela_lstm_EN_06_01 +06/03/2025 22:13:04 - lexical_benchmark.train.run_train - INFO - Training of stela_lstm_EN_06_01 has completed ! +06/03/2025 22:13:04 - lexical_benchmark.train.run_train - INFO - Model saved to /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/models/stela/EN/06/01/lstm diff --git a/stela/EN/06/01/orthographic_lstm/training_args.bin b/stela/EN/06/01/orthographic_lstm/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..db3e56c1af98edf0f8702b0fc45a4e500ce0bc6e --- /dev/null +++ b/stela/EN/06/01/orthographic_lstm/training_args.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d3bc6dc0746f2c0ee877aa6f88453d343e00039d9183093f6d00f91e526acd22 +size 5496 diff --git a/stela/EN/06/06/orthographic_gpt2/training.logs b/stela/EN/06/06/orthographic_gpt2/training.logs new file mode 100644 index 0000000000000000000000000000000000000000..3bba71c00e2f2971d894f97114febcaf1a2afe0d --- /dev/null +++ b/stela/EN/06/06/orthographic_gpt2/training.logs @@ -0,0 +1,6 @@ +05/22/2025 20:09:40 - urllib3.connectionpool - WARNING - Retrying (Retry(total=2, connect=None, read=None, redirect=None, status=None)) after connection broken by 'NewConnectionError(': Failed to establish a new connection: [Errno 101] Network is unreachable')': /api/4504800232407040/envelope/ +05/22/2025 20:09:40 - urllib3.connectionpool - WARNING - Retrying (Retry(total=1, connect=None, read=None, redirect=None, status=None)) after connection broken by 'NewConnectionError(': Failed to establish a new connection: [Errno 101] Network is unreachable')': /api/4504800232407040/envelope/ +05/22/2025 20:09:40 - urllib3.connectionpool - WARNING - Retrying (Retry(total=0, connect=None, read=None, redirect=None, status=None)) after connection broken by 'NewConnectionError(': Failed to establish a new connection: [Errno 101] Network is unreachable')': /api/4504800232407040/envelope/ +05/22/2025 20:09:40 - urllib3.connectionpool - WARNING - Retrying (Retry(total=2, connect=None, read=None, redirect=None, status=None)) after connection broken by 'NewConnectionError(': Failed to establish a new connection: [Errno 101] Network is unreachable')': /api/4504800232407040/envelope/ +05/22/2025 20:09:40 - urllib3.connectionpool - WARNING - Retrying (Retry(total=1, connect=None, read=None, redirect=None, status=None)) after connection broken by 'NewConnectionError(': Failed to establish a new connection: [Errno 101] Network is unreachable')': /api/4504800232407040/envelope/ +05/22/2025 20:09:40 - urllib3.connectionpool - WARNING - Retrying (Retry(total=0, connect=None, read=None, redirect=None, status=None)) after connection broken by 'NewConnectionError(': Failed to establish a new connection: [Errno 101] Network is unreachable')': /api/4504800232407040/envelope/ diff --git a/stela/EN/10/00/orthographic_gpt2/config.json b/stela/EN/10/00/orthographic_gpt2/config.json new file mode 100644 index 0000000000000000000000000000000000000000..493011df283c5726a4be3aa8f39c84d43991636e --- /dev/null +++ b/stela/EN/10/00/orthographic_gpt2/config.json @@ -0,0 +1,32 @@ +{ + "activation_function": "gelu_new", + "architectures": [ + "GPT2LMHeadModel" + ], + "attn_pdrop": 0.1, + "bos_token_id": 50256, + "embd_pdrop": 0.1, + "eos_token_id": 50256, + "initializer_range": 0.02, + "layer_norm_epsilon": 1e-05, + "model_type": "gpt2", + "n_ctx": 1024, + "n_embd": 768, + "n_head": 12, + "n_inner": null, + "n_layer": 12, + "n_positions": 1024, + "reorder_and_upcast_attn": false, + "resid_pdrop": 0.1, + "scale_attn_by_inverse_layer_idx": false, + "scale_attn_weights": true, + "summary_activation": null, + "summary_first_dropout": 0.1, + "summary_proj_to_labels": true, + "summary_type": "cls_index", + "summary_use_proj": true, + "torch_dtype": "float32", + "transformers_version": "4.49.0", + "use_cache": true, + "vocab_size": 115 +} diff --git a/stela/EN/10/00/orthographic_gpt2/generation_config.json b/stela/EN/10/00/orthographic_gpt2/generation_config.json new file mode 100644 index 0000000000000000000000000000000000000000..bd968f46ce39e754a4603c69f53a2a4a08a13dbb --- /dev/null +++ b/stela/EN/10/00/orthographic_gpt2/generation_config.json @@ -0,0 +1,6 @@ +{ + "_from_model_config": true, + "bos_token_id": 50256, + "eos_token_id": 50256, + "transformers_version": "4.49.0" +} diff --git a/stela/EN/10/00/orthographic_gpt2/model.safetensors b/stela/EN/10/00/orthographic_gpt2/model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..f4324e1731ad072bc9e654ab7939835060793220 --- /dev/null +++ b/stela/EN/10/00/orthographic_gpt2/model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:11a0f50bc99a2bc075515496d798719f62c30f0aebb6e51bd38984a036af7719 +size 343737984 diff --git a/stela/EN/10/00/orthographic_gpt2/training.logs b/stela/EN/10/00/orthographic_gpt2/training.logs new file mode 100644 index 0000000000000000000000000000000000000000..41fed7ad82eb44b8d0e4a7521ec2b7b050e9c5ca --- /dev/null +++ b/stela/EN/10/00/orthographic_gpt2/training.logs @@ -0,0 +1,56 @@ +05/23/2025 16:41:55 - lexical_benchmark.train.run_train - INFO - Loading gpt2 trainer class. +05/23/2025 16:41:59 - datasets - INFO - PyTorch version 2.5.1 available. +05/23/2025 16:41:59 - datasets - INFO - Polars version 1.26.0 available. +05/23/2025 16:42:00 - lexical_benchmark.train.train_params - INFO - Loading params from /lustre/fsn1/projects/rech/hhb/uye44va/workspace/lm_benchmark/prod/code/src/lexical_benchmark/train/train-params.toml +05/23/2025 16:42:00 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading char tokenizer +05/23/2025 16:42:00 - lexical_benchmark.train.trainers.gpt2 - INFO - Tokenizer loaded with vocabulary size: 115 +05/23/2025 16:42:00 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading and processing datasets with joined utterances (max_length=1024) +05/23/2025 16:42:00 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/10/00/train.tokenized +05/23/2025 16:42:00 - lexical_benchmark.train.tokenizers - INFO - Read 381551 lines from file +05/23/2025 16:42:35 - lexical_benchmark.train.tokenizers - INFO - Created 36559 joined examples from 381551 original lines +05/23/2025 16:42:41 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/dev/dev.tokenized +05/23/2025 16:42:41 - lexical_benchmark.train.tokenizers - INFO - Read 179806 lines from file +05/23/2025 16:42:58 - lexical_benchmark.train.tokenizers - INFO - Created 18373 joined examples from 179806 original lines +05/23/2025 16:43:01 - lexical_benchmark.train.trainers.gpt2 - INFO - Training dataset size: 36559 +05/23/2025 16:43:01 - lexical_benchmark.train.trainers.gpt2 - INFO - Validation dataset size: 18373 +05/23/2025 16:43:01 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating GPT2 configuration +05/23/2025 16:43:01 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating standard Trainer +05/23/2025 16:43:02 - lexical_benchmark.train.run_train - INFO - Running training of stela_gpt2_EN_10_00 +05/23/2025 16:43:48 - lexical_benchmark.train.run_train - INFO - Loading gpt2 trainer class. +05/23/2025 16:43:52 - datasets - INFO - PyTorch version 2.5.1 available. +05/23/2025 16:43:52 - datasets - INFO - Polars version 1.26.0 available. +05/23/2025 16:43:53 - lexical_benchmark.train.train_params - INFO - Loading params from /lustre/fsn1/projects/rech/hhb/uye44va/workspace/lm_benchmark/prod/code/src/lexical_benchmark/train/train-params.toml +05/23/2025 16:43:53 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading char tokenizer +05/23/2025 16:43:53 - lexical_benchmark.train.trainers.gpt2 - INFO - Tokenizer loaded with vocabulary size: 115 +05/23/2025 16:43:53 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading and processing datasets with joined utterances (max_length=1024) +05/23/2025 16:43:53 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/10/00/train.tokenized +05/23/2025 16:43:53 - lexical_benchmark.train.tokenizers - INFO - Read 381551 lines from file +05/23/2025 16:44:29 - lexical_benchmark.train.tokenizers - INFO - Created 36559 joined examples from 381551 original lines +05/23/2025 16:44:34 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/dev/dev.tokenized +05/23/2025 16:44:34 - lexical_benchmark.train.tokenizers - INFO - Read 179806 lines from file +05/23/2025 16:44:51 - lexical_benchmark.train.tokenizers - INFO - Created 18373 joined examples from 179806 original lines +05/23/2025 16:44:54 - lexical_benchmark.train.trainers.gpt2 - INFO - Training dataset size: 36559 +05/23/2025 16:44:54 - lexical_benchmark.train.trainers.gpt2 - INFO - Validation dataset size: 18373 +05/23/2025 16:44:54 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating GPT2 configuration +05/23/2025 16:44:55 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating standard Trainer +05/23/2025 16:44:55 - lexical_benchmark.train.run_train - INFO - Running training of stela_gpt2_EN_10_00 +05/28/2025 17:02:55 - lexical_benchmark.train.run_train - INFO - Loading gpt2 trainer class. +05/28/2025 17:03:07 - datasets - INFO - PyTorch version 2.5.1 available. +05/28/2025 17:03:07 - datasets - INFO - Polars version 1.26.0 available. +05/28/2025 17:03:07 - lexical_benchmark.train.train_params - INFO - Loading params from /lustre/fsn1/projects/rech/hhb/uye44va/workspace/lm_benchmark/prod/code/src/lexical_benchmark/train/train-params.toml +05/28/2025 17:03:07 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading char tokenizer +05/28/2025 17:03:07 - lexical_benchmark.train.trainers.gpt2 - INFO - Tokenizer loaded with vocabulary size: 115 +05/28/2025 17:03:07 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading and processing datasets with joined utterances (max_length=1024) +05/28/2025 17:03:07 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/10/00/train.tokenized +05/28/2025 17:03:07 - lexical_benchmark.train.tokenizers - INFO - Read 381551 lines from file +05/28/2025 17:03:42 - lexical_benchmark.train.tokenizers - INFO - Created 36559 joined examples from 381551 original lines +05/28/2025 17:03:47 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/dev/dev.tokenized +05/28/2025 17:03:47 - lexical_benchmark.train.tokenizers - INFO - Read 179806 lines from file +05/28/2025 17:04:04 - lexical_benchmark.train.tokenizers - INFO - Created 18373 joined examples from 179806 original lines +05/28/2025 17:04:07 - lexical_benchmark.train.trainers.gpt2 - INFO - Training dataset size: 36559 +05/28/2025 17:04:07 - lexical_benchmark.train.trainers.gpt2 - INFO - Validation dataset size: 18373 +05/28/2025 17:04:07 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating GPT2 configuration +05/28/2025 17:04:07 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating standard Trainer +05/28/2025 17:04:08 - lexical_benchmark.train.run_train - INFO - Running training of stela_gpt2_EN_10_00 +05/28/2025 19:27:19 - lexical_benchmark.train.run_train - INFO - Training of stela_gpt2_EN_10_00 has completed ! +05/28/2025 19:27:19 - lexical_benchmark.train.run_train - INFO - Model saved to /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/models/stela/EN/10/00/gpt2 diff --git a/stela/EN/10/00/orthographic_gpt2/training_args.bin b/stela/EN/10/00/orthographic_gpt2/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..d8389b8d433bbe795651dc9787642824f384772f --- /dev/null +++ b/stela/EN/10/00/orthographic_gpt2/training_args.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:187f185b8783d49d8dc8f755247c66abaada55daf5c56905ebefc4f87b2c9ce9 +size 5496 diff --git a/stela/EN/10/00/orthographic_lstm/config.json b/stela/EN/10/00/orthographic_lstm/config.json new file mode 100644 index 0000000000000000000000000000000000000000..5ef629887c7fc87d64278c9788e493ac34c60370 --- /dev/null +++ b/stela/EN/10/00/orthographic_lstm/config.json @@ -0,0 +1,13 @@ +{ + "architectures": [ + "LSTMForLanguageModeling" + ], + "dropout": 0.1, + "embedding_dim": 200, + "hidden_size": 1024, + "model_type": "lstm", + "num_layers": 3, + "torch_dtype": "float32", + "transformers_version": "4.49.0", + "vocab_size": 115 +} diff --git a/stela/EN/10/00/orthographic_lstm/generation_config.json b/stela/EN/10/00/orthographic_lstm/generation_config.json new file mode 100644 index 0000000000000000000000000000000000000000..c1818a95e07bdccb83305ce8f9d878755e3c6119 --- /dev/null +++ b/stela/EN/10/00/orthographic_lstm/generation_config.json @@ -0,0 +1,4 @@ +{ + "_from_model_config": true, + "transformers_version": "4.49.0" +} diff --git a/stela/EN/10/00/orthographic_lstm/model.safetensors b/stela/EN/10/00/orthographic_lstm/model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..b991e4d487c973a4668bfcfc42a1e97b01470874 --- /dev/null +++ b/stela/EN/10/00/orthographic_lstm/model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:bf07ac8b83cd0168493d155c1849d63b00b0906d6507f1360dfeef9f233b0e62 +size 87825996 diff --git a/stela/EN/10/00/orthographic_lstm/training.logs b/stela/EN/10/00/orthographic_lstm/training.logs new file mode 100644 index 0000000000000000000000000000000000000000..b0c491c4278487426241bc9bdc3a574c1a43e7c8 --- /dev/null +++ b/stela/EN/10/00/orthographic_lstm/training.logs @@ -0,0 +1,36 @@ +05/23/2025 16:40:13 - lexical_benchmark.train.run_train - INFO - Loading lstm trainer class. +05/23/2025 16:40:17 - datasets - INFO - PyTorch version 2.5.1 available. +05/23/2025 16:40:17 - datasets - INFO - Polars version 1.26.0 available. +05/23/2025 16:40:17 - lexical_benchmark.train.train_params - INFO - Loading params from /lustre/fsn1/projects/rech/hhb/uye44va/workspace/lm_benchmark/prod/code/src/lexical_benchmark/train/train-params.toml +05/23/2025 16:40:17 - lexical_benchmark.train.trainers.lstm - INFO - Loading char-tokenizer +05/23/2025 16:40:17 - lexical_benchmark.train.trainers.lstm - INFO - Tokenizer loaded with vocabulary size: 115 +05/23/2025 16:40:17 - lexical_benchmark.train.trainers.lstm - INFO - Loading and processing datasets with joined utterances (max_length=1024) +05/23/2025 16:40:18 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/10/00/train.tokenized +05/23/2025 16:40:18 - lexical_benchmark.train.tokenizers - INFO - Read 381551 lines from file +05/23/2025 16:40:54 - lexical_benchmark.train.tokenizers - INFO - Created 36559 joined examples from 381551 original lines +05/23/2025 16:40:59 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/dev/dev.tokenized +05/23/2025 16:40:59 - lexical_benchmark.train.tokenizers - INFO - Read 179806 lines from file +05/23/2025 16:41:17 - lexical_benchmark.train.tokenizers - INFO - Created 18373 joined examples from 179806 original lines +05/23/2025 16:41:19 - lexical_benchmark.train.trainers.lstm - INFO - Training dataset size: 36559 +05/23/2025 16:41:19 - lexical_benchmark.train.trainers.lstm - INFO - Validation dataset size: 18373 +05/23/2025 16:41:19 - lexical_benchmark.train.trainers.lstm - INFO - Loading configurations & initialising LSTM model trainer +05/23/2025 16:41:20 - lexical_benchmark.train.run_train - INFO - Running training of stela_lstm_EN_10_00 +05/28/2025 17:02:55 - lexical_benchmark.train.run_train - INFO - Loading lstm trainer class. +05/28/2025 17:03:07 - datasets - INFO - PyTorch version 2.5.1 available. +05/28/2025 17:03:07 - datasets - INFO - Polars version 1.26.0 available. +05/28/2025 17:03:07 - lexical_benchmark.train.train_params - INFO - Loading params from /lustre/fsn1/projects/rech/hhb/uye44va/workspace/lm_benchmark/prod/code/src/lexical_benchmark/train/train-params.toml +05/28/2025 17:03:07 - lexical_benchmark.train.trainers.lstm - INFO - Loading char-tokenizer +05/28/2025 17:03:07 - lexical_benchmark.train.trainers.lstm - INFO - Tokenizer loaded with vocabulary size: 115 +05/28/2025 17:03:07 - lexical_benchmark.train.trainers.lstm - INFO - Loading and processing datasets with joined utterances (max_length=1024) +05/28/2025 17:03:07 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/10/00/train.tokenized +05/28/2025 17:03:07 - lexical_benchmark.train.tokenizers - INFO - Read 381551 lines from file +05/28/2025 17:03:42 - lexical_benchmark.train.tokenizers - INFO - Created 36559 joined examples from 381551 original lines +05/28/2025 17:03:47 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/dev/dev.tokenized +05/28/2025 17:03:47 - lexical_benchmark.train.tokenizers - INFO - Read 179806 lines from file +05/28/2025 17:04:03 - lexical_benchmark.train.tokenizers - INFO - Created 18373 joined examples from 179806 original lines +05/28/2025 17:04:06 - lexical_benchmark.train.trainers.lstm - INFO - Training dataset size: 36559 +05/28/2025 17:04:06 - lexical_benchmark.train.trainers.lstm - INFO - Validation dataset size: 18373 +05/28/2025 17:04:06 - lexical_benchmark.train.trainers.lstm - INFO - Loading configurations & initialising LSTM model trainer +05/28/2025 17:04:07 - lexical_benchmark.train.run_train - INFO - Running training of stela_lstm_EN_10_00 +05/29/2025 01:44:58 - lexical_benchmark.train.run_train - INFO - Training of stela_lstm_EN_10_00 has completed ! +05/29/2025 01:44:58 - lexical_benchmark.train.run_train - INFO - Model saved to /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/models/stela/EN/10/00/lstm diff --git a/stela/EN/10/00/orthographic_lstm/training_args.bin b/stela/EN/10/00/orthographic_lstm/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..15ebc86f72e44f7d9ae61d1dade973b96ed52f78 --- /dev/null +++ b/stela/EN/10/00/orthographic_lstm/training_args.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3bd691538faa5ef22d7e2f8d4469bdd550befd08469aec258b44c3bb56261a51 +size 5496 diff --git a/stela/EN/10/01/orthographic_gpt2/config.json b/stela/EN/10/01/orthographic_gpt2/config.json new file mode 100644 index 0000000000000000000000000000000000000000..493011df283c5726a4be3aa8f39c84d43991636e --- /dev/null +++ b/stela/EN/10/01/orthographic_gpt2/config.json @@ -0,0 +1,32 @@ +{ + "activation_function": "gelu_new", + "architectures": [ + "GPT2LMHeadModel" + ], + "attn_pdrop": 0.1, + "bos_token_id": 50256, + "embd_pdrop": 0.1, + "eos_token_id": 50256, + "initializer_range": 0.02, + "layer_norm_epsilon": 1e-05, + "model_type": "gpt2", + "n_ctx": 1024, + "n_embd": 768, + "n_head": 12, + "n_inner": null, + "n_layer": 12, + "n_positions": 1024, + "reorder_and_upcast_attn": false, + "resid_pdrop": 0.1, + "scale_attn_by_inverse_layer_idx": false, + "scale_attn_weights": true, + "summary_activation": null, + "summary_first_dropout": 0.1, + "summary_proj_to_labels": true, + "summary_type": "cls_index", + "summary_use_proj": true, + "torch_dtype": "float32", + "transformers_version": "4.49.0", + "use_cache": true, + "vocab_size": 115 +} diff --git a/stela/EN/10/01/orthographic_gpt2/generation_config.json b/stela/EN/10/01/orthographic_gpt2/generation_config.json new file mode 100644 index 0000000000000000000000000000000000000000..bd968f46ce39e754a4603c69f53a2a4a08a13dbb --- /dev/null +++ b/stela/EN/10/01/orthographic_gpt2/generation_config.json @@ -0,0 +1,6 @@ +{ + "_from_model_config": true, + "bos_token_id": 50256, + "eos_token_id": 50256, + "transformers_version": "4.49.0" +} diff --git a/stela/EN/10/01/orthographic_gpt2/model.safetensors b/stela/EN/10/01/orthographic_gpt2/model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..c6a9de86e448e7c10c5da0d1a1ae956efd1bf9ee --- /dev/null +++ b/stela/EN/10/01/orthographic_gpt2/model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b6d888bab68dd74b432740c58e856d1a3c24d3d0e03ab210d7e145bd9ee88f38 +size 343737984 diff --git a/stela/EN/10/01/orthographic_gpt2/training.logs b/stela/EN/10/01/orthographic_gpt2/training.logs new file mode 100644 index 0000000000000000000000000000000000000000..eb4fb255be980e579133f87b6d15fe3785b0550e --- /dev/null +++ b/stela/EN/10/01/orthographic_gpt2/training.logs @@ -0,0 +1,20 @@ +05/28/2025 17:02:55 - lexical_benchmark.train.run_train - INFO - Loading gpt2 trainer class. +05/28/2025 17:03:07 - datasets - INFO - PyTorch version 2.5.1 available. +05/28/2025 17:03:07 - datasets - INFO - Polars version 1.26.0 available. +05/28/2025 17:03:07 - lexical_benchmark.train.train_params - INFO - Loading params from /lustre/fsn1/projects/rech/hhb/uye44va/workspace/lm_benchmark/prod/code/src/lexical_benchmark/train/train-params.toml +05/28/2025 17:03:07 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading char tokenizer +05/28/2025 17:03:07 - lexical_benchmark.train.trainers.gpt2 - INFO - Tokenizer loaded with vocabulary size: 115 +05/28/2025 17:03:07 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading and processing datasets with joined utterances (max_length=1024) +05/28/2025 17:03:08 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/10/01/train.tokenized +05/28/2025 17:03:08 - lexical_benchmark.train.tokenizers - INFO - Read 389401 lines from file +05/28/2025 17:03:43 - lexical_benchmark.train.tokenizers - INFO - Created 36641 joined examples from 389401 original lines +05/28/2025 17:03:49 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/dev/dev.tokenized +05/28/2025 17:03:49 - lexical_benchmark.train.tokenizers - INFO - Read 179806 lines from file +05/28/2025 17:04:05 - lexical_benchmark.train.tokenizers - INFO - Created 18373 joined examples from 179806 original lines +05/28/2025 17:04:08 - lexical_benchmark.train.trainers.gpt2 - INFO - Training dataset size: 36641 +05/28/2025 17:04:08 - lexical_benchmark.train.trainers.gpt2 - INFO - Validation dataset size: 18373 +05/28/2025 17:04:08 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating GPT2 configuration +05/28/2025 17:04:08 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating standard Trainer +05/28/2025 17:04:09 - lexical_benchmark.train.run_train - INFO - Running training of stela_gpt2_EN_10_01 +05/28/2025 19:59:54 - lexical_benchmark.train.run_train - INFO - Training of stela_gpt2_EN_10_01 has completed ! +05/28/2025 19:59:54 - lexical_benchmark.train.run_train - INFO - Model saved to /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/models/stela/EN/10/01/gpt2 diff --git a/stela/EN/10/01/orthographic_gpt2/training_args.bin b/stela/EN/10/01/orthographic_gpt2/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..a37b5375043f5a74e4c1a7dc13747fa84cddf43f --- /dev/null +++ b/stela/EN/10/01/orthographic_gpt2/training_args.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b6dab279d8424bdb792a8b636c662191ca22f1f011f757ee16ccdcdfcd7894cf +size 5496 diff --git a/stela/EN/10/01/orthographic_lstm/config.json b/stela/EN/10/01/orthographic_lstm/config.json new file mode 100644 index 0000000000000000000000000000000000000000..5ef629887c7fc87d64278c9788e493ac34c60370 --- /dev/null +++ b/stela/EN/10/01/orthographic_lstm/config.json @@ -0,0 +1,13 @@ +{ + "architectures": [ + "LSTMForLanguageModeling" + ], + "dropout": 0.1, + "embedding_dim": 200, + "hidden_size": 1024, + "model_type": "lstm", + "num_layers": 3, + "torch_dtype": "float32", + "transformers_version": "4.49.0", + "vocab_size": 115 +} diff --git a/stela/EN/10/01/orthographic_lstm/generation_config.json b/stela/EN/10/01/orthographic_lstm/generation_config.json new file mode 100644 index 0000000000000000000000000000000000000000..c1818a95e07bdccb83305ce8f9d878755e3c6119 --- /dev/null +++ b/stela/EN/10/01/orthographic_lstm/generation_config.json @@ -0,0 +1,4 @@ +{ + "_from_model_config": true, + "transformers_version": "4.49.0" +} diff --git a/stela/EN/10/01/orthographic_lstm/model.safetensors b/stela/EN/10/01/orthographic_lstm/model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..60f90a9eae50d731b234faeb7a84e619f2165f10 --- /dev/null +++ b/stela/EN/10/01/orthographic_lstm/model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8c1b19c7ffd82666fa4b21ae852460907212cd2ce6d2c33e16b5f06995fb6375 +size 87825996 diff --git a/stela/EN/10/01/orthographic_lstm/training.logs b/stela/EN/10/01/orthographic_lstm/training.logs new file mode 100644 index 0000000000000000000000000000000000000000..509f24f45dd0b3f2029e61ee1801aa4edb4286b5 --- /dev/null +++ b/stela/EN/10/01/orthographic_lstm/training.logs @@ -0,0 +1,19 @@ +05/28/2025 17:02:55 - lexical_benchmark.train.run_train - INFO - Loading lstm trainer class. +05/28/2025 17:03:07 - datasets - INFO - PyTorch version 2.5.1 available. +05/28/2025 17:03:07 - datasets - INFO - Polars version 1.26.0 available. +05/28/2025 17:03:07 - lexical_benchmark.train.train_params - INFO - Loading params from /lustre/fsn1/projects/rech/hhb/uye44va/workspace/lm_benchmark/prod/code/src/lexical_benchmark/train/train-params.toml +05/28/2025 17:03:07 - lexical_benchmark.train.trainers.lstm - INFO - Loading char-tokenizer +05/28/2025 17:03:07 - lexical_benchmark.train.trainers.lstm - INFO - Tokenizer loaded with vocabulary size: 115 +05/28/2025 17:03:07 - lexical_benchmark.train.trainers.lstm - INFO - Loading and processing datasets with joined utterances (max_length=1024) +05/28/2025 17:03:08 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/10/01/train.tokenized +05/28/2025 17:03:09 - lexical_benchmark.train.tokenizers - INFO - Read 389401 lines from file +05/28/2025 17:03:49 - lexical_benchmark.train.tokenizers - INFO - Created 36641 joined examples from 389401 original lines +05/28/2025 17:03:55 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/dev/dev.tokenized +05/28/2025 17:03:56 - lexical_benchmark.train.tokenizers - INFO - Read 179806 lines from file +05/28/2025 17:04:15 - lexical_benchmark.train.tokenizers - INFO - Created 18373 joined examples from 179806 original lines +05/28/2025 17:04:18 - lexical_benchmark.train.trainers.lstm - INFO - Training dataset size: 36641 +05/28/2025 17:04:18 - lexical_benchmark.train.trainers.lstm - INFO - Validation dataset size: 18373 +05/28/2025 17:04:18 - lexical_benchmark.train.trainers.lstm - INFO - Loading configurations & initialising LSTM model trainer +05/28/2025 17:04:19 - lexical_benchmark.train.run_train - INFO - Running training of stela_lstm_EN_10_01 +05/29/2025 02:42:03 - lexical_benchmark.train.run_train - INFO - Training of stela_lstm_EN_10_01 has completed ! +05/29/2025 02:42:03 - lexical_benchmark.train.run_train - INFO - Model saved to /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/models/stela/EN/10/01/lstm diff --git a/stela/EN/10/01/orthographic_lstm/training_args.bin b/stela/EN/10/01/orthographic_lstm/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..286688db566654c1ac27d39ec81862bb5c65949d --- /dev/null +++ b/stela/EN/10/01/orthographic_lstm/training_args.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:641a1a40aa499465d8beb4312fb16cd9b1e60ae479cef06e0765dd826b363bb7 +size 5496 diff --git a/stela/EN/15/00/orthographic_gpt2/config.json b/stela/EN/15/00/orthographic_gpt2/config.json new file mode 100644 index 0000000000000000000000000000000000000000..493011df283c5726a4be3aa8f39c84d43991636e --- /dev/null +++ b/stela/EN/15/00/orthographic_gpt2/config.json @@ -0,0 +1,32 @@ +{ + "activation_function": "gelu_new", + "architectures": [ + "GPT2LMHeadModel" + ], + "attn_pdrop": 0.1, + "bos_token_id": 50256, + "embd_pdrop": 0.1, + "eos_token_id": 50256, + "initializer_range": 0.02, + "layer_norm_epsilon": 1e-05, + "model_type": "gpt2", + "n_ctx": 1024, + "n_embd": 768, + "n_head": 12, + "n_inner": null, + "n_layer": 12, + "n_positions": 1024, + "reorder_and_upcast_attn": false, + "resid_pdrop": 0.1, + "scale_attn_by_inverse_layer_idx": false, + "scale_attn_weights": true, + "summary_activation": null, + "summary_first_dropout": 0.1, + "summary_proj_to_labels": true, + "summary_type": "cls_index", + "summary_use_proj": true, + "torch_dtype": "float32", + "transformers_version": "4.49.0", + "use_cache": true, + "vocab_size": 115 +} diff --git a/stela/EN/15/00/orthographic_gpt2/generation_config.json b/stela/EN/15/00/orthographic_gpt2/generation_config.json new file mode 100644 index 0000000000000000000000000000000000000000..bd968f46ce39e754a4603c69f53a2a4a08a13dbb --- /dev/null +++ b/stela/EN/15/00/orthographic_gpt2/generation_config.json @@ -0,0 +1,6 @@ +{ + "_from_model_config": true, + "bos_token_id": 50256, + "eos_token_id": 50256, + "transformers_version": "4.49.0" +} diff --git a/stela/EN/15/00/orthographic_gpt2/model.safetensors b/stela/EN/15/00/orthographic_gpt2/model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..73ee32aff215386858965d04c7f2876d314c8e64 --- /dev/null +++ b/stela/EN/15/00/orthographic_gpt2/model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0ad1bec556e609c66eba8fa73131969052dc11d9d34347d9e6f85f22718b99ce +size 343737984 diff --git a/stela/EN/15/00/orthographic_gpt2/training.logs b/stela/EN/15/00/orthographic_gpt2/training.logs new file mode 100644 index 0000000000000000000000000000000000000000..403fbcf9e4b25ab5aabd8816853ccd7972912468 --- /dev/null +++ b/stela/EN/15/00/orthographic_gpt2/training.logs @@ -0,0 +1,20 @@ +05/28/2025 17:02:55 - lexical_benchmark.train.run_train - INFO - Loading gpt2 trainer class. +05/28/2025 17:03:08 - datasets - INFO - PyTorch version 2.5.1 available. +05/28/2025 17:03:08 - datasets - INFO - Polars version 1.26.0 available. +05/28/2025 17:03:08 - lexical_benchmark.train.train_params - INFO - Loading params from /lustre/fsn1/projects/rech/hhb/uye44va/workspace/lm_benchmark/prod/code/src/lexical_benchmark/train/train-params.toml +05/28/2025 17:03:08 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading char tokenizer +05/28/2025 17:03:08 - lexical_benchmark.train.trainers.gpt2 - INFO - Tokenizer loaded with vocabulary size: 115 +05/28/2025 17:03:08 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading and processing datasets with joined utterances (max_length=1024) +05/28/2025 17:03:10 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/15/00/train.tokenized +05/28/2025 17:03:10 - lexical_benchmark.train.tokenizers - INFO - Read 551813 lines from file +05/28/2025 17:04:05 - lexical_benchmark.train.tokenizers - INFO - Created 55035 joined examples from 551813 original lines +05/28/2025 17:04:13 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/dev/dev.tokenized +05/28/2025 17:04:13 - lexical_benchmark.train.tokenizers - INFO - Read 179806 lines from file +05/28/2025 17:04:30 - lexical_benchmark.train.tokenizers - INFO - Created 18373 joined examples from 179806 original lines +05/28/2025 17:04:33 - lexical_benchmark.train.trainers.gpt2 - INFO - Training dataset size: 55035 +05/28/2025 17:04:33 - lexical_benchmark.train.trainers.gpt2 - INFO - Validation dataset size: 18373 +05/28/2025 17:04:33 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating GPT2 configuration +05/28/2025 17:04:34 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating standard Trainer +05/28/2025 17:04:34 - lexical_benchmark.train.run_train - INFO - Running training of stela_gpt2_EN_15_00 +05/28/2025 21:03:28 - lexical_benchmark.train.run_train - INFO - Training of stela_gpt2_EN_15_00 has completed ! +05/28/2025 21:03:28 - lexical_benchmark.train.run_train - INFO - Model saved to /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/models/stela/EN/15/00/gpt2 diff --git a/stela/EN/15/00/orthographic_gpt2/training_args.bin b/stela/EN/15/00/orthographic_gpt2/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..745d9d1158f6797eb8a21980e87a0df51f9e239b --- /dev/null +++ b/stela/EN/15/00/orthographic_gpt2/training_args.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:417c25d1b2bf0258de21a18f2ffb7bb0e7eaa9042bada3d21d47f05a5338d3b8 +size 5496 diff --git a/stela/EN/15/00/orthographic_lstm/config.json b/stela/EN/15/00/orthographic_lstm/config.json new file mode 100644 index 0000000000000000000000000000000000000000..5ef629887c7fc87d64278c9788e493ac34c60370 --- /dev/null +++ b/stela/EN/15/00/orthographic_lstm/config.json @@ -0,0 +1,13 @@ +{ + "architectures": [ + "LSTMForLanguageModeling" + ], + "dropout": 0.1, + "embedding_dim": 200, + "hidden_size": 1024, + "model_type": "lstm", + "num_layers": 3, + "torch_dtype": "float32", + "transformers_version": "4.49.0", + "vocab_size": 115 +} diff --git a/stela/EN/15/00/orthographic_lstm/generation_config.json b/stela/EN/15/00/orthographic_lstm/generation_config.json new file mode 100644 index 0000000000000000000000000000000000000000..c1818a95e07bdccb83305ce8f9d878755e3c6119 --- /dev/null +++ b/stela/EN/15/00/orthographic_lstm/generation_config.json @@ -0,0 +1,4 @@ +{ + "_from_model_config": true, + "transformers_version": "4.49.0" +} diff --git a/stela/EN/15/00/orthographic_lstm/model.safetensors b/stela/EN/15/00/orthographic_lstm/model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..51cb6ac1135ff61359f28f75507c6beb1f642fec --- /dev/null +++ b/stela/EN/15/00/orthographic_lstm/model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3d52cc7f3fb34510d90369b51d3130fe48d02b5dfd2bacac4422d32e0187e6cc +size 87825996 diff --git a/stela/EN/15/00/orthographic_lstm/training.logs b/stela/EN/15/00/orthographic_lstm/training.logs new file mode 100644 index 0000000000000000000000000000000000000000..732d2488b5e5a31fba41d17885edeadb137bbf59 --- /dev/null +++ b/stela/EN/15/00/orthographic_lstm/training.logs @@ -0,0 +1,19 @@ +05/28/2025 17:02:55 - lexical_benchmark.train.run_train - INFO - Loading lstm trainer class. +05/28/2025 17:03:07 - datasets - INFO - PyTorch version 2.5.1 available. +05/28/2025 17:03:07 - datasets - INFO - Polars version 1.26.0 available. +05/28/2025 17:03:07 - lexical_benchmark.train.train_params - INFO - Loading params from /lustre/fsn1/projects/rech/hhb/uye44va/workspace/lm_benchmark/prod/code/src/lexical_benchmark/train/train-params.toml +05/28/2025 17:03:07 - lexical_benchmark.train.trainers.lstm - INFO - Loading char-tokenizer +05/28/2025 17:03:07 - lexical_benchmark.train.trainers.lstm - INFO - Tokenizer loaded with vocabulary size: 115 +05/28/2025 17:03:07 - lexical_benchmark.train.trainers.lstm - INFO - Loading and processing datasets with joined utterances (max_length=1024) +05/28/2025 17:03:09 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/15/00/train.tokenized +05/28/2025 17:03:09 - lexical_benchmark.train.tokenizers - INFO - Read 551813 lines from file +05/28/2025 17:04:02 - lexical_benchmark.train.tokenizers - INFO - Created 55035 joined examples from 551813 original lines +05/28/2025 17:04:10 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/dev/dev.tokenized +05/28/2025 17:04:10 - lexical_benchmark.train.tokenizers - INFO - Read 179806 lines from file +05/28/2025 17:04:27 - lexical_benchmark.train.tokenizers - INFO - Created 18373 joined examples from 179806 original lines +05/28/2025 17:04:29 - lexical_benchmark.train.trainers.lstm - INFO - Training dataset size: 55035 +05/28/2025 17:04:29 - lexical_benchmark.train.trainers.lstm - INFO - Validation dataset size: 18373 +05/28/2025 17:04:29 - lexical_benchmark.train.trainers.lstm - INFO - Loading configurations & initialising LSTM model trainer +05/28/2025 17:04:30 - lexical_benchmark.train.run_train - INFO - Running training of stela_lstm_EN_15_00 +05/29/2025 01:46:36 - lexical_benchmark.train.run_train - INFO - Training of stela_lstm_EN_15_00 has completed ! +05/29/2025 01:46:36 - lexical_benchmark.train.run_train - INFO - Model saved to /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/models/stela/EN/15/00/lstm diff --git a/stela/EN/15/00/orthographic_lstm/training_args.bin b/stela/EN/15/00/orthographic_lstm/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..c48e9d7615293d3b9bc6ccdaed90b72c0bb598e9 --- /dev/null +++ b/stela/EN/15/00/orthographic_lstm/training_args.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:bee275ce00960a1520d1da93db6473723a5d563fbc69b55d35f1e7f899cbf8e3 +size 5496 diff --git a/stela/EN/15/01/orthographic_gpt2/config.json b/stela/EN/15/01/orthographic_gpt2/config.json new file mode 100644 index 0000000000000000000000000000000000000000..493011df283c5726a4be3aa8f39c84d43991636e --- /dev/null +++ b/stela/EN/15/01/orthographic_gpt2/config.json @@ -0,0 +1,32 @@ +{ + "activation_function": "gelu_new", + "architectures": [ + "GPT2LMHeadModel" + ], + "attn_pdrop": 0.1, + "bos_token_id": 50256, + "embd_pdrop": 0.1, + "eos_token_id": 50256, + "initializer_range": 0.02, + "layer_norm_epsilon": 1e-05, + "model_type": "gpt2", + "n_ctx": 1024, + "n_embd": 768, + "n_head": 12, + "n_inner": null, + "n_layer": 12, + "n_positions": 1024, + "reorder_and_upcast_attn": false, + "resid_pdrop": 0.1, + "scale_attn_by_inverse_layer_idx": false, + "scale_attn_weights": true, + "summary_activation": null, + "summary_first_dropout": 0.1, + "summary_proj_to_labels": true, + "summary_type": "cls_index", + "summary_use_proj": true, + "torch_dtype": "float32", + "transformers_version": "4.49.0", + "use_cache": true, + "vocab_size": 115 +} diff --git a/stela/EN/15/01/orthographic_gpt2/generation_config.json b/stela/EN/15/01/orthographic_gpt2/generation_config.json new file mode 100644 index 0000000000000000000000000000000000000000..bd968f46ce39e754a4603c69f53a2a4a08a13dbb --- /dev/null +++ b/stela/EN/15/01/orthographic_gpt2/generation_config.json @@ -0,0 +1,6 @@ +{ + "_from_model_config": true, + "bos_token_id": 50256, + "eos_token_id": 50256, + "transformers_version": "4.49.0" +} diff --git a/stela/EN/15/01/orthographic_gpt2/model.safetensors b/stela/EN/15/01/orthographic_gpt2/model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..2704b43306c842783209852be8659fee20193e7f --- /dev/null +++ b/stela/EN/15/01/orthographic_gpt2/model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f84add0a4a992f629f4906f2cff875c7ad1dfc53b265b9fcafca3255e44b0826 +size 343737984 diff --git a/stela/EN/15/01/orthographic_gpt2/training.logs b/stela/EN/15/01/orthographic_gpt2/training.logs new file mode 100644 index 0000000000000000000000000000000000000000..f9f6735d4257507a2a31db995ab8b92564860278 --- /dev/null +++ b/stela/EN/15/01/orthographic_gpt2/training.logs @@ -0,0 +1,20 @@ +05/28/2025 17:02:55 - lexical_benchmark.train.run_train - INFO - Loading gpt2 trainer class. +05/28/2025 17:03:07 - datasets - INFO - PyTorch version 2.5.1 available. +05/28/2025 17:03:07 - datasets - INFO - Polars version 1.26.0 available. +05/28/2025 17:03:07 - lexical_benchmark.train.train_params - INFO - Loading params from /lustre/fsn1/projects/rech/hhb/uye44va/workspace/lm_benchmark/prod/code/src/lexical_benchmark/train/train-params.toml +05/28/2025 17:03:07 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading char tokenizer +05/28/2025 17:03:07 - lexical_benchmark.train.trainers.gpt2 - INFO - Tokenizer loaded with vocabulary size: 115 +05/28/2025 17:03:07 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading and processing datasets with joined utterances (max_length=1024) +05/28/2025 17:03:09 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/15/01/train.tokenized +05/28/2025 17:03:09 - lexical_benchmark.train.tokenizers - INFO - Read 567136 lines from file +05/28/2025 17:04:08 - lexical_benchmark.train.tokenizers - INFO - Created 55130 joined examples from 567136 original lines +05/28/2025 17:04:16 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/dev/dev.tokenized +05/28/2025 17:04:16 - lexical_benchmark.train.tokenizers - INFO - Read 179806 lines from file +05/28/2025 17:04:34 - lexical_benchmark.train.tokenizers - INFO - Created 18373 joined examples from 179806 original lines +05/28/2025 17:04:36 - lexical_benchmark.train.trainers.gpt2 - INFO - Training dataset size: 55130 +05/28/2025 17:04:36 - lexical_benchmark.train.trainers.gpt2 - INFO - Validation dataset size: 18373 +05/28/2025 17:04:36 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating GPT2 configuration +05/28/2025 17:04:37 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating standard Trainer +05/28/2025 17:04:38 - lexical_benchmark.train.run_train - INFO - Running training of stela_gpt2_EN_15_01 +05/28/2025 21:36:25 - lexical_benchmark.train.run_train - INFO - Training of stela_gpt2_EN_15_01 has completed ! +05/28/2025 21:36:25 - lexical_benchmark.train.run_train - INFO - Model saved to /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/models/stela/EN/15/01/gpt2 diff --git a/stela/EN/15/01/orthographic_gpt2/training_args.bin b/stela/EN/15/01/orthographic_gpt2/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..afb03726bd40ee69983357795978b2f03ebd2a9c --- /dev/null +++ b/stela/EN/15/01/orthographic_gpt2/training_args.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:fdbc3c96e6afd1b4975be5b05ae2ac655349413a8774d5b117e167893aa538c6 +size 5496 diff --git a/stela/EN/15/01/orthographic_lstm/config.json b/stela/EN/15/01/orthographic_lstm/config.json new file mode 100644 index 0000000000000000000000000000000000000000..5ef629887c7fc87d64278c9788e493ac34c60370 --- /dev/null +++ b/stela/EN/15/01/orthographic_lstm/config.json @@ -0,0 +1,13 @@ +{ + "architectures": [ + "LSTMForLanguageModeling" + ], + "dropout": 0.1, + "embedding_dim": 200, + "hidden_size": 1024, + "model_type": "lstm", + "num_layers": 3, + "torch_dtype": "float32", + "transformers_version": "4.49.0", + "vocab_size": 115 +} diff --git a/stela/EN/15/01/orthographic_lstm/generation_config.json b/stela/EN/15/01/orthographic_lstm/generation_config.json new file mode 100644 index 0000000000000000000000000000000000000000..c1818a95e07bdccb83305ce8f9d878755e3c6119 --- /dev/null +++ b/stela/EN/15/01/orthographic_lstm/generation_config.json @@ -0,0 +1,4 @@ +{ + "_from_model_config": true, + "transformers_version": "4.49.0" +} diff --git a/stela/EN/15/01/orthographic_lstm/model.safetensors b/stela/EN/15/01/orthographic_lstm/model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..2a64aa6e97b58f23a4576721fbf66b8452a675eb --- /dev/null +++ b/stela/EN/15/01/orthographic_lstm/model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f26af8d8303a0da749bfab3eefa0d8332613e2c1148ff311af0a9e5383ea296e +size 87825996 diff --git a/stela/EN/15/01/orthographic_lstm/training.logs b/stela/EN/15/01/orthographic_lstm/training.logs new file mode 100644 index 0000000000000000000000000000000000000000..b7d7a278c940f1311ce48eb3072b9c8db411f8ae --- /dev/null +++ b/stela/EN/15/01/orthographic_lstm/training.logs @@ -0,0 +1,19 @@ +05/28/2025 17:02:55 - lexical_benchmark.train.run_train - INFO - Loading lstm trainer class. +05/28/2025 17:03:07 - datasets - INFO - PyTorch version 2.5.1 available. +05/28/2025 17:03:07 - datasets - INFO - Polars version 1.26.0 available. +05/28/2025 17:03:07 - lexical_benchmark.train.train_params - INFO - Loading params from /lustre/fsn1/projects/rech/hhb/uye44va/workspace/lm_benchmark/prod/code/src/lexical_benchmark/train/train-params.toml +05/28/2025 17:03:07 - lexical_benchmark.train.trainers.lstm - INFO - Loading char-tokenizer +05/28/2025 17:03:07 - lexical_benchmark.train.trainers.lstm - INFO - Tokenizer loaded with vocabulary size: 115 +05/28/2025 17:03:07 - lexical_benchmark.train.trainers.lstm - INFO - Loading and processing datasets with joined utterances (max_length=1024) +05/28/2025 17:03:09 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/15/01/train.tokenized +05/28/2025 17:03:09 - lexical_benchmark.train.tokenizers - INFO - Read 134202 lines from file +05/28/2025 17:03:21 - lexical_benchmark.train.tokenizers - INFO - Created 12729 joined examples from 134202 original lines +05/28/2025 17:03:23 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/dev/dev.tokenized +05/28/2025 17:03:23 - lexical_benchmark.train.tokenizers - INFO - Read 179806 lines from file +05/28/2025 17:03:40 - lexical_benchmark.train.tokenizers - INFO - Created 18373 joined examples from 179806 original lines +05/28/2025 17:03:43 - lexical_benchmark.train.trainers.lstm - INFO - Training dataset size: 12729 +05/28/2025 17:03:43 - lexical_benchmark.train.trainers.lstm - INFO - Validation dataset size: 18373 +05/28/2025 17:03:43 - lexical_benchmark.train.trainers.lstm - INFO - Loading configurations & initialising LSTM model trainer +05/28/2025 17:03:43 - lexical_benchmark.train.run_train - INFO - Running training of stela_lstm_EN_15_01 +05/28/2025 19:53:51 - lexical_benchmark.train.run_train - INFO - Training of stela_lstm_EN_15_01 has completed ! +05/28/2025 19:53:51 - lexical_benchmark.train.run_train - INFO - Model saved to /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/models/stela/EN/15/01/lstm diff --git a/stela/EN/15/01/orthographic_lstm/training_args.bin b/stela/EN/15/01/orthographic_lstm/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..c782664957fe3880289cbf191a63048949c871c3 --- /dev/null +++ b/stela/EN/15/01/orthographic_lstm/training_args.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:64f771560ca5b17d7ee43f0782553a7c78f802e8d70850b6ba8f2e7fa71a70a4 +size 5496 diff --git a/stela/EN/20/00/orthographic_gpt2/config.json b/stela/EN/20/00/orthographic_gpt2/config.json new file mode 100644 index 0000000000000000000000000000000000000000..493011df283c5726a4be3aa8f39c84d43991636e --- /dev/null +++ b/stela/EN/20/00/orthographic_gpt2/config.json @@ -0,0 +1,32 @@ +{ + "activation_function": "gelu_new", + "architectures": [ + "GPT2LMHeadModel" + ], + "attn_pdrop": 0.1, + "bos_token_id": 50256, + "embd_pdrop": 0.1, + "eos_token_id": 50256, + "initializer_range": 0.02, + "layer_norm_epsilon": 1e-05, + "model_type": "gpt2", + "n_ctx": 1024, + "n_embd": 768, + "n_head": 12, + "n_inner": null, + "n_layer": 12, + "n_positions": 1024, + "reorder_and_upcast_attn": false, + "resid_pdrop": 0.1, + "scale_attn_by_inverse_layer_idx": false, + "scale_attn_weights": true, + "summary_activation": null, + "summary_first_dropout": 0.1, + "summary_proj_to_labels": true, + "summary_type": "cls_index", + "summary_use_proj": true, + "torch_dtype": "float32", + "transformers_version": "4.49.0", + "use_cache": true, + "vocab_size": 115 +} diff --git a/stela/EN/20/00/orthographic_gpt2/generation_config.json b/stela/EN/20/00/orthographic_gpt2/generation_config.json new file mode 100644 index 0000000000000000000000000000000000000000..bd968f46ce39e754a4603c69f53a2a4a08a13dbb --- /dev/null +++ b/stela/EN/20/00/orthographic_gpt2/generation_config.json @@ -0,0 +1,6 @@ +{ + "_from_model_config": true, + "bos_token_id": 50256, + "eos_token_id": 50256, + "transformers_version": "4.49.0" +} diff --git a/stela/EN/20/00/orthographic_gpt2/model.safetensors b/stela/EN/20/00/orthographic_gpt2/model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..a3f6dffa08641d06f0a2f024a7eef6c111ed345c --- /dev/null +++ b/stela/EN/20/00/orthographic_gpt2/model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b003e4841b2716d2b50d052480fe30f62f6f6948257102a9fc6d424294dcb743 +size 343737984 diff --git a/stela/EN/20/00/orthographic_gpt2/training.logs b/stela/EN/20/00/orthographic_gpt2/training.logs new file mode 100644 index 0000000000000000000000000000000000000000..1027fd89de816be275d646abe3a3901ecb6625c0 --- /dev/null +++ b/stela/EN/20/00/orthographic_gpt2/training.logs @@ -0,0 +1,20 @@ +05/28/2025 17:02:55 - lexical_benchmark.train.run_train - INFO - Loading gpt2 trainer class. +05/28/2025 17:03:07 - datasets - INFO - PyTorch version 2.5.1 available. +05/28/2025 17:03:07 - datasets - INFO - Polars version 1.26.0 available. +05/28/2025 17:03:07 - lexical_benchmark.train.train_params - INFO - Loading params from /lustre/fsn1/projects/rech/hhb/uye44va/workspace/lm_benchmark/prod/code/src/lexical_benchmark/train/train-params.toml +05/28/2025 17:03:07 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading char tokenizer +05/28/2025 17:03:07 - lexical_benchmark.train.trainers.gpt2 - INFO - Tokenizer loaded with vocabulary size: 115 +05/28/2025 17:03:07 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading and processing datasets with joined utterances (max_length=1024) +05/28/2025 17:03:09 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/20/00/train.tokenized +05/28/2025 17:03:09 - lexical_benchmark.train.tokenizers - INFO - Read 752582 lines from file +05/28/2025 17:04:25 - lexical_benchmark.train.tokenizers - INFO - Created 73162 joined examples from 752582 original lines +05/28/2025 17:04:36 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/dev/dev.tokenized +05/28/2025 17:04:36 - lexical_benchmark.train.tokenizers - INFO - Read 179806 lines from file +05/28/2025 17:04:53 - lexical_benchmark.train.tokenizers - INFO - Created 18373 joined examples from 179806 original lines +05/28/2025 17:04:55 - lexical_benchmark.train.trainers.gpt2 - INFO - Training dataset size: 73162 +05/28/2025 17:04:55 - lexical_benchmark.train.trainers.gpt2 - INFO - Validation dataset size: 18373 +05/28/2025 17:04:55 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating GPT2 configuration +05/28/2025 17:04:56 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating standard Trainer +05/28/2025 17:04:57 - lexical_benchmark.train.run_train - INFO - Running training of stela_gpt2_EN_20_00 +05/28/2025 22:39:48 - lexical_benchmark.train.run_train - INFO - Training of stela_gpt2_EN_20_00 has completed ! +05/28/2025 22:39:48 - lexical_benchmark.train.run_train - INFO - Model saved to /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/models/stela/EN/20/00/gpt2 diff --git a/stela/EN/20/00/orthographic_gpt2/training_args.bin b/stela/EN/20/00/orthographic_gpt2/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..d88c8f73129019cb03f29ff426f96b5bf30d6380 --- /dev/null +++ b/stela/EN/20/00/orthographic_gpt2/training_args.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d3b82ee53a89d40cde2c4d162beb8edb26f50d4b5b0af9d0b75e858a9eafa9e1 +size 5496 diff --git a/stela/EN/20/00/orthographic_lstm/config.json b/stela/EN/20/00/orthographic_lstm/config.json new file mode 100644 index 0000000000000000000000000000000000000000..5ef629887c7fc87d64278c9788e493ac34c60370 --- /dev/null +++ b/stela/EN/20/00/orthographic_lstm/config.json @@ -0,0 +1,13 @@ +{ + "architectures": [ + "LSTMForLanguageModeling" + ], + "dropout": 0.1, + "embedding_dim": 200, + "hidden_size": 1024, + "model_type": "lstm", + "num_layers": 3, + "torch_dtype": "float32", + "transformers_version": "4.49.0", + "vocab_size": 115 +} diff --git a/stela/EN/20/00/orthographic_lstm/generation_config.json b/stela/EN/20/00/orthographic_lstm/generation_config.json new file mode 100644 index 0000000000000000000000000000000000000000..c1818a95e07bdccb83305ce8f9d878755e3c6119 --- /dev/null +++ b/stela/EN/20/00/orthographic_lstm/generation_config.json @@ -0,0 +1,4 @@ +{ + "_from_model_config": true, + "transformers_version": "4.49.0" +} diff --git a/stela/EN/20/00/orthographic_lstm/model.safetensors b/stela/EN/20/00/orthographic_lstm/model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..c03555fc25994c7da48583144ddef168a68e541f --- /dev/null +++ b/stela/EN/20/00/orthographic_lstm/model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1fa2fe25af3be222e78aa25a135577c9d36763dfc8aa78a37d1c93984c6503ad +size 87825996 diff --git a/stela/EN/20/00/orthographic_lstm/training.logs b/stela/EN/20/00/orthographic_lstm/training.logs new file mode 100644 index 0000000000000000000000000000000000000000..2e1c2a21bb2f33c337bd54b6a6517e20e1d1ff06 --- /dev/null +++ b/stela/EN/20/00/orthographic_lstm/training.logs @@ -0,0 +1,19 @@ +05/28/2025 17:02:55 - lexical_benchmark.train.run_train - INFO - Loading lstm trainer class. +05/28/2025 17:03:07 - datasets - INFO - PyTorch version 2.5.1 available. +05/28/2025 17:03:07 - datasets - INFO - Polars version 1.26.0 available. +05/28/2025 17:03:07 - lexical_benchmark.train.train_params - INFO - Loading params from /lustre/fsn1/projects/rech/hhb/uye44va/workspace/lm_benchmark/prod/code/src/lexical_benchmark/train/train-params.toml +05/28/2025 17:03:07 - lexical_benchmark.train.trainers.lstm - INFO - Loading char-tokenizer +05/28/2025 17:03:07 - lexical_benchmark.train.trainers.lstm - INFO - Tokenizer loaded with vocabulary size: 115 +05/28/2025 17:03:07 - lexical_benchmark.train.trainers.lstm - INFO - Loading and processing datasets with joined utterances (max_length=1024) +05/28/2025 17:03:09 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/20/00/train.tokenized +05/28/2025 17:03:10 - lexical_benchmark.train.tokenizers - INFO - Read 752582 lines from file +05/28/2025 17:04:25 - lexical_benchmark.train.tokenizers - INFO - Created 73162 joined examples from 752582 original lines +05/28/2025 17:04:36 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/dev/dev.tokenized +05/28/2025 17:04:36 - lexical_benchmark.train.tokenizers - INFO - Read 179806 lines from file +05/28/2025 17:04:53 - lexical_benchmark.train.tokenizers - INFO - Created 18373 joined examples from 179806 original lines +05/28/2025 17:04:55 - lexical_benchmark.train.trainers.lstm - INFO - Training dataset size: 73162 +05/28/2025 17:04:55 - lexical_benchmark.train.trainers.lstm - INFO - Validation dataset size: 18373 +05/28/2025 17:04:55 - lexical_benchmark.train.trainers.lstm - INFO - Loading configurations & initialising LSTM model trainer +05/28/2025 17:04:56 - lexical_benchmark.train.run_train - INFO - Running training of stela_lstm_EN_20_00 +05/29/2025 01:47:14 - lexical_benchmark.train.run_train - INFO - Training of stela_lstm_EN_20_00 has completed ! +05/29/2025 01:47:14 - lexical_benchmark.train.run_train - INFO - Model saved to /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/models/stela/EN/20/00/lstm diff --git a/stela/EN/20/00/orthographic_lstm/training_args.bin b/stela/EN/20/00/orthographic_lstm/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..88b47dc41d8f059e7a16b02c7b3ca4df42744c2c --- /dev/null +++ b/stela/EN/20/00/orthographic_lstm/training_args.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9031d3661b1f532481acbcc630a8b78cec27f63be364ac15f45087d1dbe259c5 +size 5496 diff --git a/stela/EN/20/01/orthographic_gpt2/config.json b/stela/EN/20/01/orthographic_gpt2/config.json new file mode 100644 index 0000000000000000000000000000000000000000..493011df283c5726a4be3aa8f39c84d43991636e --- /dev/null +++ b/stela/EN/20/01/orthographic_gpt2/config.json @@ -0,0 +1,32 @@ +{ + "activation_function": "gelu_new", + "architectures": [ + "GPT2LMHeadModel" + ], + "attn_pdrop": 0.1, + "bos_token_id": 50256, + "embd_pdrop": 0.1, + "eos_token_id": 50256, + "initializer_range": 0.02, + "layer_norm_epsilon": 1e-05, + "model_type": "gpt2", + "n_ctx": 1024, + "n_embd": 768, + "n_head": 12, + "n_inner": null, + "n_layer": 12, + "n_positions": 1024, + "reorder_and_upcast_attn": false, + "resid_pdrop": 0.1, + "scale_attn_by_inverse_layer_idx": false, + "scale_attn_weights": true, + "summary_activation": null, + "summary_first_dropout": 0.1, + "summary_proj_to_labels": true, + "summary_type": "cls_index", + "summary_use_proj": true, + "torch_dtype": "float32", + "transformers_version": "4.49.0", + "use_cache": true, + "vocab_size": 115 +} diff --git a/stela/EN/20/01/orthographic_gpt2/generation_config.json b/stela/EN/20/01/orthographic_gpt2/generation_config.json new file mode 100644 index 0000000000000000000000000000000000000000..bd968f46ce39e754a4603c69f53a2a4a08a13dbb --- /dev/null +++ b/stela/EN/20/01/orthographic_gpt2/generation_config.json @@ -0,0 +1,6 @@ +{ + "_from_model_config": true, + "bos_token_id": 50256, + "eos_token_id": 50256, + "transformers_version": "4.49.0" +} diff --git a/stela/EN/20/01/orthographic_gpt2/model.safetensors b/stela/EN/20/01/orthographic_gpt2/model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..da0d85e9acdf9aaed209da08f5b5195799c01034 --- /dev/null +++ b/stela/EN/20/01/orthographic_gpt2/model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a9d3d61a5569a3e5ecefb960c2f36e3182b89c73020d8e874d7d744bb2b47b7a +size 343737984 diff --git a/stela/EN/20/01/orthographic_gpt2/training.logs b/stela/EN/20/01/orthographic_gpt2/training.logs new file mode 100644 index 0000000000000000000000000000000000000000..c8cb4c3bc0af94d536aca8f29df5b85a303de4bd --- /dev/null +++ b/stela/EN/20/01/orthographic_gpt2/training.logs @@ -0,0 +1,20 @@ +05/28/2025 17:02:56 - lexical_benchmark.train.run_train - INFO - Loading gpt2 trainer class. +05/28/2025 17:03:08 - datasets - INFO - PyTorch version 2.5.1 available. +05/28/2025 17:03:08 - datasets - INFO - Polars version 1.26.0 available. +05/28/2025 17:03:09 - lexical_benchmark.train.train_params - INFO - Loading params from /lustre/fsn1/projects/rech/hhb/uye44va/workspace/lm_benchmark/prod/code/src/lexical_benchmark/train/train-params.toml +05/28/2025 17:03:09 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading char tokenizer +05/28/2025 17:03:09 - lexical_benchmark.train.trainers.gpt2 - INFO - Tokenizer loaded with vocabulary size: 115 +05/28/2025 17:03:09 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading and processing datasets with joined utterances (max_length=1024) +05/28/2025 17:03:11 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/20/01/train.tokenized +05/28/2025 17:03:11 - lexical_benchmark.train.tokenizers - INFO - Read 756352 lines from file +05/28/2025 17:04:27 - lexical_benchmark.train.tokenizers - INFO - Created 73577 joined examples from 756352 original lines +05/28/2025 17:04:38 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/dev/dev.tokenized +05/28/2025 17:04:38 - lexical_benchmark.train.tokenizers - INFO - Read 179806 lines from file +05/28/2025 17:04:55 - lexical_benchmark.train.tokenizers - INFO - Created 18373 joined examples from 179806 original lines +05/28/2025 17:04:57 - lexical_benchmark.train.trainers.gpt2 - INFO - Training dataset size: 73577 +05/28/2025 17:04:57 - lexical_benchmark.train.trainers.gpt2 - INFO - Validation dataset size: 18373 +05/28/2025 17:04:57 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating GPT2 configuration +05/28/2025 17:04:58 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating standard Trainer +05/28/2025 17:04:59 - lexical_benchmark.train.run_train - INFO - Running training of stela_gpt2_EN_20_01 +05/28/2025 22:07:50 - lexical_benchmark.train.run_train - INFO - Training of stela_gpt2_EN_20_01 has completed ! +05/28/2025 22:07:50 - lexical_benchmark.train.run_train - INFO - Model saved to /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/models/stela/EN/20/01/gpt2 diff --git a/stela/EN/20/01/orthographic_gpt2/training_args.bin b/stela/EN/20/01/orthographic_gpt2/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..368bfcc335aa194406e1d223b45d92a56575a063 --- /dev/null +++ b/stela/EN/20/01/orthographic_gpt2/training_args.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7a414be7f973de72825d90d9ecca881db9650f30915e752baa3500bfbd238fdb +size 5496 diff --git a/stela/EN/20/01/orthographic_lstm/config.json b/stela/EN/20/01/orthographic_lstm/config.json new file mode 100644 index 0000000000000000000000000000000000000000..5ef629887c7fc87d64278c9788e493ac34c60370 --- /dev/null +++ b/stela/EN/20/01/orthographic_lstm/config.json @@ -0,0 +1,13 @@ +{ + "architectures": [ + "LSTMForLanguageModeling" + ], + "dropout": 0.1, + "embedding_dim": 200, + "hidden_size": 1024, + "model_type": "lstm", + "num_layers": 3, + "torch_dtype": "float32", + "transformers_version": "4.49.0", + "vocab_size": 115 +} diff --git a/stela/EN/20/01/orthographic_lstm/generation_config.json b/stela/EN/20/01/orthographic_lstm/generation_config.json new file mode 100644 index 0000000000000000000000000000000000000000..c1818a95e07bdccb83305ce8f9d878755e3c6119 --- /dev/null +++ b/stela/EN/20/01/orthographic_lstm/generation_config.json @@ -0,0 +1,4 @@ +{ + "_from_model_config": true, + "transformers_version": "4.49.0" +} diff --git a/stela/EN/20/01/orthographic_lstm/model.safetensors b/stela/EN/20/01/orthographic_lstm/model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..bb9d98e94263b20c0f3ddfe073ba28f2e3453265 --- /dev/null +++ b/stela/EN/20/01/orthographic_lstm/model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:024acc61cc3eb17e5d4828fbadf4642b032ef2c7e00ee5edef3ed228b644b2ba +size 87825996 diff --git a/stela/EN/20/01/orthographic_lstm/training.logs b/stela/EN/20/01/orthographic_lstm/training.logs new file mode 100644 index 0000000000000000000000000000000000000000..c072b0bcd98f7db0b5c61e3553f8783ad3f0edfd --- /dev/null +++ b/stela/EN/20/01/orthographic_lstm/training.logs @@ -0,0 +1,19 @@ +05/28/2025 17:02:56 - lexical_benchmark.train.run_train - INFO - Loading lstm trainer class. +05/28/2025 17:03:08 - datasets - INFO - PyTorch version 2.5.1 available. +05/28/2025 17:03:08 - datasets - INFO - Polars version 1.26.0 available. +05/28/2025 17:03:08 - lexical_benchmark.train.train_params - INFO - Loading params from /lustre/fsn1/projects/rech/hhb/uye44va/workspace/lm_benchmark/prod/code/src/lexical_benchmark/train/train-params.toml +05/28/2025 17:03:08 - lexical_benchmark.train.trainers.lstm - INFO - Loading char-tokenizer +05/28/2025 17:03:08 - lexical_benchmark.train.trainers.lstm - INFO - Tokenizer loaded with vocabulary size: 115 +05/28/2025 17:03:08 - lexical_benchmark.train.trainers.lstm - INFO - Loading and processing datasets with joined utterances (max_length=1024) +05/28/2025 17:03:10 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/20/01/train.tokenized +05/28/2025 17:03:10 - lexical_benchmark.train.tokenizers - INFO - Read 756352 lines from file +05/28/2025 17:04:28 - lexical_benchmark.train.tokenizers - INFO - Created 73577 joined examples from 756352 original lines +05/28/2025 17:04:39 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/dev/dev.tokenized +05/28/2025 17:04:39 - lexical_benchmark.train.tokenizers - INFO - Read 179806 lines from file +05/28/2025 17:04:56 - lexical_benchmark.train.tokenizers - INFO - Created 18373 joined examples from 179806 original lines +05/28/2025 17:04:59 - lexical_benchmark.train.trainers.lstm - INFO - Training dataset size: 73577 +05/28/2025 17:04:59 - lexical_benchmark.train.trainers.lstm - INFO - Validation dataset size: 18373 +05/28/2025 17:04:59 - lexical_benchmark.train.trainers.lstm - INFO - Loading configurations & initialising LSTM model trainer +05/28/2025 17:04:59 - lexical_benchmark.train.run_train - INFO - Running training of stela_lstm_EN_20_01 +05/29/2025 02:00:37 - lexical_benchmark.train.run_train - INFO - Training of stela_lstm_EN_20_01 has completed ! +05/29/2025 02:00:37 - lexical_benchmark.train.run_train - INFO - Model saved to /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/models/stela/EN/20/01/lstm diff --git a/stela/EN/20/01/orthographic_lstm/training_args.bin b/stela/EN/20/01/orthographic_lstm/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..bc3044f5e47e6aa995cc83f879143d0742e2c68e --- /dev/null +++ b/stela/EN/20/01/orthographic_lstm/training_args.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:065e306ac81cd6ea93ffde9f11c07f97b4dcc9a584718eda16900b9e3632ed36 +size 5496 diff --git a/stela/EN/25/00/orthographic_gpt2/config.json b/stela/EN/25/00/orthographic_gpt2/config.json new file mode 100644 index 0000000000000000000000000000000000000000..493011df283c5726a4be3aa8f39c84d43991636e --- /dev/null +++ b/stela/EN/25/00/orthographic_gpt2/config.json @@ -0,0 +1,32 @@ +{ + "activation_function": "gelu_new", + "architectures": [ + "GPT2LMHeadModel" + ], + "attn_pdrop": 0.1, + "bos_token_id": 50256, + "embd_pdrop": 0.1, + "eos_token_id": 50256, + "initializer_range": 0.02, + "layer_norm_epsilon": 1e-05, + "model_type": "gpt2", + "n_ctx": 1024, + "n_embd": 768, + "n_head": 12, + "n_inner": null, + "n_layer": 12, + "n_positions": 1024, + "reorder_and_upcast_attn": false, + "resid_pdrop": 0.1, + "scale_attn_by_inverse_layer_idx": false, + "scale_attn_weights": true, + "summary_activation": null, + "summary_first_dropout": 0.1, + "summary_proj_to_labels": true, + "summary_type": "cls_index", + "summary_use_proj": true, + "torch_dtype": "float32", + "transformers_version": "4.49.0", + "use_cache": true, + "vocab_size": 115 +} diff --git a/stela/EN/25/00/orthographic_gpt2/generation_config.json b/stela/EN/25/00/orthographic_gpt2/generation_config.json new file mode 100644 index 0000000000000000000000000000000000000000..bd968f46ce39e754a4603c69f53a2a4a08a13dbb --- /dev/null +++ b/stela/EN/25/00/orthographic_gpt2/generation_config.json @@ -0,0 +1,6 @@ +{ + "_from_model_config": true, + "bos_token_id": 50256, + "eos_token_id": 50256, + "transformers_version": "4.49.0" +} diff --git a/stela/EN/25/00/orthographic_gpt2/model.safetensors b/stela/EN/25/00/orthographic_gpt2/model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..780ad2d1613e0701c3bf379e1dc79c8ceb39394d --- /dev/null +++ b/stela/EN/25/00/orthographic_gpt2/model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:971c50e1fc2d4e3acdcb6d55962b8b594c0e0c07bcedb0a65963930fc8f28261 +size 343737984 diff --git a/stela/EN/25/00/orthographic_gpt2/training.logs b/stela/EN/25/00/orthographic_gpt2/training.logs new file mode 100644 index 0000000000000000000000000000000000000000..635271b8aa083098e29de836956a11607a35e903 --- /dev/null +++ b/stela/EN/25/00/orthographic_gpt2/training.logs @@ -0,0 +1,20 @@ +05/28/2025 17:02:56 - lexical_benchmark.train.run_train - INFO - Loading gpt2 trainer class. +05/28/2025 17:03:08 - datasets - INFO - PyTorch version 2.5.1 available. +05/28/2025 17:03:08 - datasets - INFO - Polars version 1.26.0 available. +05/28/2025 17:03:08 - lexical_benchmark.train.train_params - INFO - Loading params from /lustre/fsn1/projects/rech/hhb/uye44va/workspace/lm_benchmark/prod/code/src/lexical_benchmark/train/train-params.toml +05/28/2025 17:03:08 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading char tokenizer +05/28/2025 17:03:08 - lexical_benchmark.train.trainers.gpt2 - INFO - Tokenizer loaded with vocabulary size: 115 +05/28/2025 17:03:08 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading and processing datasets with joined utterances (max_length=1024) +05/28/2025 17:03:11 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/25/00/train.tokenized +05/28/2025 17:03:11 - lexical_benchmark.train.tokenizers - INFO - Read 940240 lines from file +05/28/2025 17:04:48 - lexical_benchmark.train.tokenizers - INFO - Created 90908 joined examples from 940240 original lines +05/28/2025 17:05:02 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/dev/dev.tokenized +05/28/2025 17:05:02 - lexical_benchmark.train.tokenizers - INFO - Read 179806 lines from file +05/28/2025 17:05:18 - lexical_benchmark.train.tokenizers - INFO - Created 18373 joined examples from 179806 original lines +05/28/2025 17:05:20 - lexical_benchmark.train.trainers.gpt2 - INFO - Training dataset size: 90908 +05/28/2025 17:05:20 - lexical_benchmark.train.trainers.gpt2 - INFO - Validation dataset size: 18373 +05/28/2025 17:05:20 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating GPT2 configuration +05/28/2025 17:05:21 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating standard Trainer +05/28/2025 17:05:21 - lexical_benchmark.train.run_train - INFO - Running training of stela_gpt2_EN_25_00 +05/29/2025 01:03:19 - lexical_benchmark.train.run_train - INFO - Training of stela_gpt2_EN_25_00 has completed ! +05/29/2025 01:03:19 - lexical_benchmark.train.run_train - INFO - Model saved to /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/models/stela/EN/25/00/gpt2 diff --git a/stela/EN/25/00/orthographic_gpt2/training_args.bin b/stela/EN/25/00/orthographic_gpt2/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..ec1daf92e82b2505793e091c9967010c1405a915 --- /dev/null +++ b/stela/EN/25/00/orthographic_gpt2/training_args.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9b493c5ee0e292dd0d92a9a31995cbb8d2c27e2fed7574a825f44042729327f1 +size 5496 diff --git a/stela/EN/25/00/orthographic_lstm/config.json b/stela/EN/25/00/orthographic_lstm/config.json new file mode 100644 index 0000000000000000000000000000000000000000..5ef629887c7fc87d64278c9788e493ac34c60370 --- /dev/null +++ b/stela/EN/25/00/orthographic_lstm/config.json @@ -0,0 +1,13 @@ +{ + "architectures": [ + "LSTMForLanguageModeling" + ], + "dropout": 0.1, + "embedding_dim": 200, + "hidden_size": 1024, + "model_type": "lstm", + "num_layers": 3, + "torch_dtype": "float32", + "transformers_version": "4.49.0", + "vocab_size": 115 +} diff --git a/stela/EN/25/00/orthographic_lstm/generation_config.json b/stela/EN/25/00/orthographic_lstm/generation_config.json new file mode 100644 index 0000000000000000000000000000000000000000..c1818a95e07bdccb83305ce8f9d878755e3c6119 --- /dev/null +++ b/stela/EN/25/00/orthographic_lstm/generation_config.json @@ -0,0 +1,4 @@ +{ + "_from_model_config": true, + "transformers_version": "4.49.0" +} diff --git a/stela/EN/25/00/orthographic_lstm/model.safetensors b/stela/EN/25/00/orthographic_lstm/model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..7ba5807b9259788e4e0162290304d91b87c8ad48 --- /dev/null +++ b/stela/EN/25/00/orthographic_lstm/model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:bb1db802c477d76c3d4a42734d9c981a9fa2b9ec53f6263e4ad05f916e43ccdd +size 87825996 diff --git a/stela/EN/25/00/orthographic_lstm/training.logs b/stela/EN/25/00/orthographic_lstm/training.logs new file mode 100644 index 0000000000000000000000000000000000000000..f82b9dd55f8c515878d1f6c7a46df7e5d5a9679b --- /dev/null +++ b/stela/EN/25/00/orthographic_lstm/training.logs @@ -0,0 +1,19 @@ +05/28/2025 17:02:56 - lexical_benchmark.train.run_train - INFO - Loading lstm trainer class. +05/28/2025 17:03:09 - datasets - INFO - PyTorch version 2.5.1 available. +05/28/2025 17:03:09 - datasets - INFO - Polars version 1.26.0 available. +05/28/2025 17:03:10 - lexical_benchmark.train.train_params - INFO - Loading params from /lustre/fsn1/projects/rech/hhb/uye44va/workspace/lm_benchmark/prod/code/src/lexical_benchmark/train/train-params.toml +05/28/2025 17:03:10 - lexical_benchmark.train.trainers.lstm - INFO - Loading char-tokenizer +05/28/2025 17:03:10 - lexical_benchmark.train.trainers.lstm - INFO - Tokenizer loaded with vocabulary size: 115 +05/28/2025 17:03:10 - lexical_benchmark.train.trainers.lstm - INFO - Loading and processing datasets with joined utterances (max_length=1024) +05/28/2025 17:03:12 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/25/00/train.tokenized +05/28/2025 17:03:12 - lexical_benchmark.train.tokenizers - INFO - Read 940240 lines from file +05/28/2025 17:04:48 - lexical_benchmark.train.tokenizers - INFO - Created 90908 joined examples from 940240 original lines +05/28/2025 17:05:02 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/dev/dev.tokenized +05/28/2025 17:05:02 - lexical_benchmark.train.tokenizers - INFO - Read 179806 lines from file +05/28/2025 17:05:19 - lexical_benchmark.train.tokenizers - INFO - Created 18373 joined examples from 179806 original lines +05/28/2025 17:05:21 - lexical_benchmark.train.trainers.lstm - INFO - Training dataset size: 90908 +05/28/2025 17:05:21 - lexical_benchmark.train.trainers.lstm - INFO - Validation dataset size: 18373 +05/28/2025 17:05:21 - lexical_benchmark.train.trainers.lstm - INFO - Loading configurations & initialising LSTM model trainer +05/28/2025 17:05:22 - lexical_benchmark.train.run_train - INFO - Running training of stela_lstm_EN_25_00 +05/29/2025 02:00:43 - lexical_benchmark.train.run_train - INFO - Training of stela_lstm_EN_25_00 has completed ! +05/29/2025 02:00:43 - lexical_benchmark.train.run_train - INFO - Model saved to /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/models/stela/EN/25/00/lstm diff --git a/stela/EN/25/00/orthographic_lstm/training_args.bin b/stela/EN/25/00/orthographic_lstm/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..6d005b189f0e06d0a8898375db9b8c58f66208f0 --- /dev/null +++ b/stela/EN/25/00/orthographic_lstm/training_args.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a91cf960b115cd1884f5794796edd135eaacba58d634925a6bdf7a2f683425ef +size 5496 diff --git a/stela/EN/25/01/orthographic_gpt2/config.json b/stela/EN/25/01/orthographic_gpt2/config.json new file mode 100644 index 0000000000000000000000000000000000000000..493011df283c5726a4be3aa8f39c84d43991636e --- /dev/null +++ b/stela/EN/25/01/orthographic_gpt2/config.json @@ -0,0 +1,32 @@ +{ + "activation_function": "gelu_new", + "architectures": [ + "GPT2LMHeadModel" + ], + "attn_pdrop": 0.1, + "bos_token_id": 50256, + "embd_pdrop": 0.1, + "eos_token_id": 50256, + "initializer_range": 0.02, + "layer_norm_epsilon": 1e-05, + "model_type": "gpt2", + "n_ctx": 1024, + "n_embd": 768, + "n_head": 12, + "n_inner": null, + "n_layer": 12, + "n_positions": 1024, + "reorder_and_upcast_attn": false, + "resid_pdrop": 0.1, + "scale_attn_by_inverse_layer_idx": false, + "scale_attn_weights": true, + "summary_activation": null, + "summary_first_dropout": 0.1, + "summary_proj_to_labels": true, + "summary_type": "cls_index", + "summary_use_proj": true, + "torch_dtype": "float32", + "transformers_version": "4.49.0", + "use_cache": true, + "vocab_size": 115 +} diff --git a/stela/EN/25/01/orthographic_gpt2/generation_config.json b/stela/EN/25/01/orthographic_gpt2/generation_config.json new file mode 100644 index 0000000000000000000000000000000000000000..bd968f46ce39e754a4603c69f53a2a4a08a13dbb --- /dev/null +++ b/stela/EN/25/01/orthographic_gpt2/generation_config.json @@ -0,0 +1,6 @@ +{ + "_from_model_config": true, + "bos_token_id": 50256, + "eos_token_id": 50256, + "transformers_version": "4.49.0" +} diff --git a/stela/EN/25/01/orthographic_gpt2/model.safetensors b/stela/EN/25/01/orthographic_gpt2/model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..764c13f391af43efa9333c4df5fca2942b603041 --- /dev/null +++ b/stela/EN/25/01/orthographic_gpt2/model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:083759137c4ac200fb0fc8d2e609a7a641210034551630a009fab29a4e6efc07 +size 343737984 diff --git a/stela/EN/25/01/orthographic_gpt2/training.logs b/stela/EN/25/01/orthographic_gpt2/training.logs new file mode 100644 index 0000000000000000000000000000000000000000..80e7baabc0e6a66fa9bfa647d2587c380c3534f0 --- /dev/null +++ b/stela/EN/25/01/orthographic_gpt2/training.logs @@ -0,0 +1,20 @@ +05/28/2025 17:02:55 - lexical_benchmark.train.run_train - INFO - Loading gpt2 trainer class. +05/28/2025 17:03:07 - datasets - INFO - PyTorch version 2.5.1 available. +05/28/2025 17:03:07 - datasets - INFO - Polars version 1.26.0 available. +05/28/2025 17:03:08 - lexical_benchmark.train.train_params - INFO - Loading params from /lustre/fsn1/projects/rech/hhb/uye44va/workspace/lm_benchmark/prod/code/src/lexical_benchmark/train/train-params.toml +05/28/2025 17:03:08 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading char tokenizer +05/28/2025 17:03:08 - lexical_benchmark.train.trainers.gpt2 - INFO - Tokenizer loaded with vocabulary size: 115 +05/28/2025 17:03:08 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading and processing datasets with joined utterances (max_length=1024) +05/28/2025 17:03:10 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/25/01/train.tokenized +05/28/2025 17:03:10 - lexical_benchmark.train.tokenizers - INFO - Read 942140 lines from file +05/28/2025 17:04:51 - lexical_benchmark.train.tokenizers - INFO - Created 92100 joined examples from 942140 original lines +05/28/2025 17:05:07 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/dev/dev.tokenized +05/28/2025 17:05:07 - lexical_benchmark.train.tokenizers - INFO - Read 179806 lines from file +05/28/2025 17:05:23 - lexical_benchmark.train.tokenizers - INFO - Created 18373 joined examples from 179806 original lines +05/28/2025 17:05:26 - lexical_benchmark.train.trainers.gpt2 - INFO - Training dataset size: 92100 +05/28/2025 17:05:26 - lexical_benchmark.train.trainers.gpt2 - INFO - Validation dataset size: 18373 +05/28/2025 17:05:26 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating GPT2 configuration +05/28/2025 17:05:27 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating standard Trainer +05/28/2025 17:05:27 - lexical_benchmark.train.run_train - INFO - Running training of stela_gpt2_EN_25_01 +05/29/2025 00:31:08 - lexical_benchmark.train.run_train - INFO - Training of stela_gpt2_EN_25_01 has completed ! +05/29/2025 00:31:08 - lexical_benchmark.train.run_train - INFO - Model saved to /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/models/stela/EN/25/01/gpt2 diff --git a/stela/EN/25/01/orthographic_gpt2/training_args.bin b/stela/EN/25/01/orthographic_gpt2/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..10faf566b77992fc623e15e25488c412ccfa2cc9 --- /dev/null +++ b/stela/EN/25/01/orthographic_gpt2/training_args.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d6f590c61970ac99db5c3043a71bcf5f69a0b99ddbbc9fb5eabee2b9bf8a7a10 +size 5496 diff --git a/stela/EN/25/01/orthographic_lstm/config.json b/stela/EN/25/01/orthographic_lstm/config.json new file mode 100644 index 0000000000000000000000000000000000000000..5ef629887c7fc87d64278c9788e493ac34c60370 --- /dev/null +++ b/stela/EN/25/01/orthographic_lstm/config.json @@ -0,0 +1,13 @@ +{ + "architectures": [ + "LSTMForLanguageModeling" + ], + "dropout": 0.1, + "embedding_dim": 200, + "hidden_size": 1024, + "model_type": "lstm", + "num_layers": 3, + "torch_dtype": "float32", + "transformers_version": "4.49.0", + "vocab_size": 115 +} diff --git a/stela/EN/25/01/orthographic_lstm/generation_config.json b/stela/EN/25/01/orthographic_lstm/generation_config.json new file mode 100644 index 0000000000000000000000000000000000000000..c1818a95e07bdccb83305ce8f9d878755e3c6119 --- /dev/null +++ b/stela/EN/25/01/orthographic_lstm/generation_config.json @@ -0,0 +1,4 @@ +{ + "_from_model_config": true, + "transformers_version": "4.49.0" +} diff --git a/stela/EN/25/01/orthographic_lstm/model.safetensors b/stela/EN/25/01/orthographic_lstm/model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..4c228274d0ded82088758d3d6d5fca3be6562a84 --- /dev/null +++ b/stela/EN/25/01/orthographic_lstm/model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3f174ab1793a8ae0bb4a0b029a10fa6c6994c2b706d885ef37585e4594d3a033 +size 87825996 diff --git a/stela/EN/25/01/orthographic_lstm/training.logs b/stela/EN/25/01/orthographic_lstm/training.logs new file mode 100644 index 0000000000000000000000000000000000000000..172a9bc3c145bfd77269386af957e0dc4947595d --- /dev/null +++ b/stela/EN/25/01/orthographic_lstm/training.logs @@ -0,0 +1,19 @@ +05/28/2025 17:02:55 - lexical_benchmark.train.run_train - INFO - Loading lstm trainer class. +05/28/2025 17:03:07 - datasets - INFO - PyTorch version 2.5.1 available. +05/28/2025 17:03:07 - datasets - INFO - Polars version 1.26.0 available. +05/28/2025 17:03:07 - lexical_benchmark.train.train_params - INFO - Loading params from /lustre/fsn1/projects/rech/hhb/uye44va/workspace/lm_benchmark/prod/code/src/lexical_benchmark/train/train-params.toml +05/28/2025 17:03:07 - lexical_benchmark.train.trainers.lstm - INFO - Loading char-tokenizer +05/28/2025 17:03:07 - lexical_benchmark.train.trainers.lstm - INFO - Tokenizer loaded with vocabulary size: 115 +05/28/2025 17:03:07 - lexical_benchmark.train.trainers.lstm - INFO - Loading and processing datasets with joined utterances (max_length=1024) +05/28/2025 17:03:10 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/25/01/train.tokenized +05/28/2025 17:03:10 - lexical_benchmark.train.tokenizers - INFO - Read 942140 lines from file +05/28/2025 17:04:47 - lexical_benchmark.train.tokenizers - INFO - Created 92100 joined examples from 942140 original lines +05/28/2025 17:05:02 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/dev/dev.tokenized +05/28/2025 17:05:02 - lexical_benchmark.train.tokenizers - INFO - Read 179806 lines from file +05/28/2025 17:05:18 - lexical_benchmark.train.tokenizers - INFO - Created 18373 joined examples from 179806 original lines +05/28/2025 17:05:20 - lexical_benchmark.train.trainers.lstm - INFO - Training dataset size: 92100 +05/28/2025 17:05:20 - lexical_benchmark.train.trainers.lstm - INFO - Validation dataset size: 18373 +05/28/2025 17:05:20 - lexical_benchmark.train.trainers.lstm - INFO - Loading configurations & initialising LSTM model trainer +05/28/2025 17:05:21 - lexical_benchmark.train.run_train - INFO - Running training of stela_lstm_EN_25_01 +05/29/2025 01:52:28 - lexical_benchmark.train.run_train - INFO - Training of stela_lstm_EN_25_01 has completed ! +05/29/2025 01:52:29 - lexical_benchmark.train.run_train - INFO - Model saved to /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/models/stela/EN/25/01/lstm diff --git a/stela/EN/25/01/orthographic_lstm/training_args.bin b/stela/EN/25/01/orthographic_lstm/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..3ff6dddf79797d177fca4d03950d33df19bf658d --- /dev/null +++ b/stela/EN/25/01/orthographic_lstm/training_args.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:cd93209e841be5c41dfb34e71dec6113df6fcf765aa04a878a10ebf9ee6939d7 +size 5496 diff --git a/stela/EN/30/00/orthographic_gpt2/config.json b/stela/EN/30/00/orthographic_gpt2/config.json new file mode 100644 index 0000000000000000000000000000000000000000..493011df283c5726a4be3aa8f39c84d43991636e --- /dev/null +++ b/stela/EN/30/00/orthographic_gpt2/config.json @@ -0,0 +1,32 @@ +{ + "activation_function": "gelu_new", + "architectures": [ + "GPT2LMHeadModel" + ], + "attn_pdrop": 0.1, + "bos_token_id": 50256, + "embd_pdrop": 0.1, + "eos_token_id": 50256, + "initializer_range": 0.02, + "layer_norm_epsilon": 1e-05, + "model_type": "gpt2", + "n_ctx": 1024, + "n_embd": 768, + "n_head": 12, + "n_inner": null, + "n_layer": 12, + "n_positions": 1024, + "reorder_and_upcast_attn": false, + "resid_pdrop": 0.1, + "scale_attn_by_inverse_layer_idx": false, + "scale_attn_weights": true, + "summary_activation": null, + "summary_first_dropout": 0.1, + "summary_proj_to_labels": true, + "summary_type": "cls_index", + "summary_use_proj": true, + "torch_dtype": "float32", + "transformers_version": "4.49.0", + "use_cache": true, + "vocab_size": 115 +} diff --git a/stela/EN/30/00/orthographic_gpt2/generation_config.json b/stela/EN/30/00/orthographic_gpt2/generation_config.json new file mode 100644 index 0000000000000000000000000000000000000000..bd968f46ce39e754a4603c69f53a2a4a08a13dbb --- /dev/null +++ b/stela/EN/30/00/orthographic_gpt2/generation_config.json @@ -0,0 +1,6 @@ +{ + "_from_model_config": true, + "bos_token_id": 50256, + "eos_token_id": 50256, + "transformers_version": "4.49.0" +} diff --git a/stela/EN/30/00/orthographic_gpt2/model.safetensors b/stela/EN/30/00/orthographic_gpt2/model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..898266a598b8444c1503ceafd7afbd6b78eb4876 --- /dev/null +++ b/stela/EN/30/00/orthographic_gpt2/model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:66511585839bb086c631ad7c294ba9f83052e7b6a758fd65127026ba5ce0ae5b +size 343737984 diff --git a/stela/EN/30/00/orthographic_gpt2/training.logs b/stela/EN/30/00/orthographic_gpt2/training.logs new file mode 100644 index 0000000000000000000000000000000000000000..d55357d406694a4bdd985225548d47e9f8c9d782 --- /dev/null +++ b/stela/EN/30/00/orthographic_gpt2/training.logs @@ -0,0 +1,20 @@ +05/28/2025 17:02:55 - lexical_benchmark.train.run_train - INFO - Loading gpt2 trainer class. +05/28/2025 17:03:08 - datasets - INFO - PyTorch version 2.5.1 available. +05/28/2025 17:03:08 - datasets - INFO - Polars version 1.26.0 available. +05/28/2025 17:03:08 - lexical_benchmark.train.train_params - INFO - Loading params from /lustre/fsn1/projects/rech/hhb/uye44va/workspace/lm_benchmark/prod/code/src/lexical_benchmark/train/train-params.toml +05/28/2025 17:03:08 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading char tokenizer +05/28/2025 17:03:08 - lexical_benchmark.train.trainers.gpt2 - INFO - Tokenizer loaded with vocabulary size: 115 +05/28/2025 17:03:08 - lexical_benchmark.train.trainers.gpt2 - INFO - Loading and processing datasets with joined utterances (max_length=1024) +05/28/2025 17:03:11 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/30/00/train.tokenized +05/28/2025 17:03:11 - lexical_benchmark.train.tokenizers - INFO - Read 1118012 lines from file +05/28/2025 17:05:08 - lexical_benchmark.train.tokenizers - INFO - Created 109380 joined examples from 1118012 original lines +05/28/2025 17:05:24 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/dev/dev.tokenized +05/28/2025 17:05:24 - lexical_benchmark.train.tokenizers - INFO - Read 179806 lines from file +05/28/2025 17:05:40 - lexical_benchmark.train.tokenizers - INFO - Created 18373 joined examples from 179806 original lines +05/28/2025 17:05:43 - lexical_benchmark.train.trainers.gpt2 - INFO - Training dataset size: 109380 +05/28/2025 17:05:43 - lexical_benchmark.train.trainers.gpt2 - INFO - Validation dataset size: 18373 +05/28/2025 17:05:43 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating GPT2 configuration +05/28/2025 17:05:44 - lexical_benchmark.train.trainers.gpt2 - INFO - Creating standard Trainer +05/28/2025 17:05:44 - lexical_benchmark.train.run_train - INFO - Running training of stela_gpt2_EN_30_00 +05/29/2025 01:35:15 - lexical_benchmark.train.run_train - INFO - Training of stela_gpt2_EN_30_00 has completed ! +05/29/2025 01:35:15 - lexical_benchmark.train.run_train - INFO - Model saved to /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/models/stela/EN/30/00/gpt2 diff --git a/stela/EN/30/00/orthographic_gpt2/training_args.bin b/stela/EN/30/00/orthographic_gpt2/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..e6f8ad56423a0390a22148f90889e240289042f8 --- /dev/null +++ b/stela/EN/30/00/orthographic_gpt2/training_args.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a6c62d5a0ad6cafdfa40ca6f10e8e09d29112c29d9f45ad767fec568f33b9cc4 +size 5496 diff --git a/stela/EN/30/00/orthographic_lstm/config.json b/stela/EN/30/00/orthographic_lstm/config.json new file mode 100644 index 0000000000000000000000000000000000000000..5ef629887c7fc87d64278c9788e493ac34c60370 --- /dev/null +++ b/stela/EN/30/00/orthographic_lstm/config.json @@ -0,0 +1,13 @@ +{ + "architectures": [ + "LSTMForLanguageModeling" + ], + "dropout": 0.1, + "embedding_dim": 200, + "hidden_size": 1024, + "model_type": "lstm", + "num_layers": 3, + "torch_dtype": "float32", + "transformers_version": "4.49.0", + "vocab_size": 115 +} diff --git a/stela/EN/30/00/orthographic_lstm/generation_config.json b/stela/EN/30/00/orthographic_lstm/generation_config.json new file mode 100644 index 0000000000000000000000000000000000000000..c1818a95e07bdccb83305ce8f9d878755e3c6119 --- /dev/null +++ b/stela/EN/30/00/orthographic_lstm/generation_config.json @@ -0,0 +1,4 @@ +{ + "_from_model_config": true, + "transformers_version": "4.49.0" +} diff --git a/stela/EN/30/00/orthographic_lstm/model.safetensors b/stela/EN/30/00/orthographic_lstm/model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..0303baf134c7a0f41fe9863e0386b502481008e0 --- /dev/null +++ b/stela/EN/30/00/orthographic_lstm/model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d30885199f86945592d374bb0da2d28d0f1797e81bab065d7e65942e94fd1908 +size 87825996 diff --git a/stela/EN/30/00/orthographic_lstm/training.logs b/stela/EN/30/00/orthographic_lstm/training.logs new file mode 100644 index 0000000000000000000000000000000000000000..2d89f68e44a2d2cb2e2d10620c0c5dedc32e5238 --- /dev/null +++ b/stela/EN/30/00/orthographic_lstm/training.logs @@ -0,0 +1,19 @@ +05/28/2025 17:02:55 - lexical_benchmark.train.run_train - INFO - Loading lstm trainer class. +05/28/2025 17:03:07 - datasets - INFO - PyTorch version 2.5.1 available. +05/28/2025 17:03:07 - datasets - INFO - Polars version 1.26.0 available. +05/28/2025 17:03:07 - lexical_benchmark.train.train_params - INFO - Loading params from /lustre/fsn1/projects/rech/hhb/uye44va/workspace/lm_benchmark/prod/code/src/lexical_benchmark/train/train-params.toml +05/28/2025 17:03:07 - lexical_benchmark.train.trainers.lstm - INFO - Loading char-tokenizer +05/28/2025 17:03:07 - lexical_benchmark.train.trainers.lstm - INFO - Tokenizer loaded with vocabulary size: 115 +05/28/2025 17:03:07 - lexical_benchmark.train.trainers.lstm - INFO - Loading and processing datasets with joined utterances (max_length=1024) +05/28/2025 17:03:10 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/30/00/train.tokenized +05/28/2025 17:03:10 - lexical_benchmark.train.tokenizers - INFO - Read 1118012 lines from file +05/28/2025 17:05:08 - lexical_benchmark.train.tokenizers - INFO - Created 109380 joined examples from 1118012 original lines +05/28/2025 17:05:24 - lexical_benchmark.train.tokenizers - INFO - Reading text from /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/datasets/stela/by_size/EN/dev/dev.tokenized +05/28/2025 17:05:24 - lexical_benchmark.train.tokenizers - INFO - Read 179806 lines from file +05/28/2025 17:05:40 - lexical_benchmark.train.tokenizers - INFO - Created 18373 joined examples from 179806 original lines +05/28/2025 17:05:43 - lexical_benchmark.train.trainers.lstm - INFO - Training dataset size: 109380 +05/28/2025 17:05:43 - lexical_benchmark.train.trainers.lstm - INFO - Validation dataset size: 18373 +05/28/2025 17:05:43 - lexical_benchmark.train.trainers.lstm - INFO - Loading configurations & initialising LSTM model trainer +05/28/2025 17:05:43 - lexical_benchmark.train.run_train - INFO - Running training of stela_lstm_EN_30_00 +05/29/2025 01:49:14 - lexical_benchmark.train.run_train - INFO - Training of stela_lstm_EN_30_00 has completed ! +05/29/2025 01:49:14 - lexical_benchmark.train.run_train - INFO - Model saved to /lustre/fswork/projects/rech/hhb/commun/lexical-benchmark/models/stela/EN/30/00/lstm diff --git a/stela/EN/30/00/orthographic_lstm/training_args.bin b/stela/EN/30/00/orthographic_lstm/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..78354ca553c58162f1f06dada7374c682d72f622 --- /dev/null +++ b/stela/EN/30/00/orthographic_lstm/training_args.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c252c2d1e5ead939c3ead6ebe2f4a282efa6b69bfe530ddfdad0ae380fb16328 +size 5496 diff --git a/stela/EN/30/01/orthographic_gpt2/config.json b/stela/EN/30/01/orthographic_gpt2/config.json new file mode 100644 index 0000000000000000000000000000000000000000..493011df283c5726a4be3aa8f39c84d43991636e --- /dev/null +++ b/stela/EN/30/01/orthographic_gpt2/config.json @@ -0,0 +1,32 @@ +{ + "activation_function": "gelu_new", + "architectures": [ + "GPT2LMHeadModel" + ], + "attn_pdrop": 0.1, + "bos_token_id": 50256, + "embd_pdrop": 0.1, + "eos_token_id": 50256, + "initializer_range": 0.02, + "layer_norm_epsilon": 1e-05, + "model_type": "gpt2", + "n_ctx": 1024, + "n_embd": 768, + "n_head": 12, + "n_inner": null, + "n_layer": 12, + "n_positions": 1024, + "reorder_and_upcast_attn": false, + "resid_pdrop": 0.1, + "scale_attn_by_inverse_layer_idx": false, + "scale_attn_weights": true, + "summary_activation": null, + "summary_first_dropout": 0.1, + "summary_proj_to_labels": true, + "summary_type": "cls_index", + "summary_use_proj": true, + "torch_dtype": "float32", + "transformers_version": "4.49.0", + "use_cache": true, + "vocab_size": 115 +}