556b812af635a7c074e915cb04fa916d

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

• Loss: 0.7418
• Data Size: 1.0
• Epoch Runtime: 8.8846
• Accuracy: 0.8333
• F1 Macro: 0.8347

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	1.8902	0	0.8036	0.1792	0.0836
No log	1	170	1.7651	0.0078	1.2378	0.1979	0.1176
No log	2	340	1.7125	0.0156	1.0274	0.3542	0.1770
No log	3	510	1.6302	0.0312	1.3792	0.2875	0.1774
No log	4	680	1.5510	0.0625	1.6990	0.3729	0.2519
0.0959	5	850	1.2756	0.125	2.2889	0.525	0.3790
0.0959	6	1020	0.8803	0.25	3.1743	0.6937	0.5680
0.8674	7	1190	0.8325	0.5	5.1661	0.7063	0.7165
0.6337	8.0	1360	0.6727	1.0	9.1500	0.7833	0.7893
0.4516	9.0	1530	0.5310	1.0	8.9803	0.85	0.8478
0.3229	10.0	1700	0.5425	1.0	9.4322	0.8396	0.8222
0.2642	11.0	1870	0.5433	1.0	9.7524	0.825	0.8301
0.2402	12.0	2040	0.6689	1.0	8.7796	0.8208	0.7996
0.2125	13.0	2210	0.7418	1.0	8.8846	0.8333	0.8347

Framework versions

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