fa62f91a20d689a0327d1a7cd0f64d37

This model is a fine-tuned version of google-bert/bert-large-uncased-whole-word-masking on the contemmcm/trec dataset. It achieves the following results on the evaluation set:

• Loss: 0.2072
• Data Size: 1.0
• Epoch Runtime: 19.6301
• Accuracy: 0.9688
• F1 Macro: 0.9575

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 5e-05
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• distributed_type: multi-GPU
• num_devices: 4
• total_train_batch_size: 32
• total_eval_batch_size: 32
• optimizer: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: constant
• num_epochs: 50

Training results

Training Loss	Epoch	Step	Validation Loss	Data Size	Epoch Runtime	Accuracy	F1 Macro
No log	0	0	2.1016	0	0.9615	0.2146	0.0603
No log	1	170	1.7396	0.0078	1.3741	0.3146	0.1410
No log	2	340	1.7595	0.0156	1.6452	0.2667	0.0854
No log	3	510	1.7183	0.0312	2.5708	0.1854	0.1220
No log	4	680	1.3450	0.0625	3.6894	0.4188	0.3637
0.0782	5	850	0.3433	0.125	4.7394	0.9062	0.7605
0.0782	6	1020	0.2220	0.25	7.1136	0.9375	0.8201
0.2885	7	1190	0.3527	0.5	11.6226	0.9146	0.9032
0.1871	8.0	1360	0.2188	1.0	19.3197	0.9583	0.9333
0.2363	9.0	1530	0.1507	1.0	18.9049	0.9729	0.9751
0.1095	10.0	1700	0.2446	1.0	19.3125	0.9646	0.9682
0.1041	11.0	1870	0.2158	1.0	18.7699	0.9646	0.9470
0.0902	12.0	2040	0.1731	1.0	19.0156	0.9729	0.9683
0.067	13.0	2210	0.2072	1.0	19.6301	0.9688	0.9575

Framework versions

• Transformers 4.57.0
• Pytorch 2.8.0+cu128
• Datasets 4.3.0
• Tokenizers 0.22.1