83428c86b643eae677ae2a6cb681d6f1

This model is a fine-tuned version of google-bert/bert-base-multilingual-cased on the contemmcm/trec dataset. It achieves the following results on the evaluation set:

• Loss: 0.1905
• Data Size: 1.0
• Epoch Runtime: 10.0110
• Accuracy: 0.9708
• F1 Macro: 0.9663

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	1.7960	0	0.8129	0.1646	0.0471
No log	1	170	1.6696	0.0078	1.1363	0.2771	0.0723
No log	2	340	1.7087	0.0156	1.2767	0.3833	0.1882
No log	3	510	1.5574	0.0312	1.7468	0.4979	0.3438
No log	4	680	1.0149	0.0625	2.0031	0.6979	0.5681
0.075	5	850	0.3924	0.125	2.8257	0.9042	0.7592
0.075	6	1020	0.2977	0.25	3.9929	0.9083	0.7585
0.3136	7	1190	0.2164	0.5	6.1424	0.9437	0.9339
0.2265	8.0	1360	0.2480	1.0	10.8015	0.925	0.9269
0.1588	9.0	1530	0.1436	1.0	10.5611	0.9625	0.9611
0.1348	10.0	1700	0.1786	1.0	11.1877	0.9625	0.9533
0.0944	11.0	1870	0.2590	1.0	10.3626	0.9479	0.9394
0.0716	12.0	2040	0.2618	1.0	10.0806	0.9625	0.9609
0.046	13.0	2210	0.1905	1.0	10.0110	0.9708	0.9663

Framework versions

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