e55f11f7380cf6d6a0dd84a16c0951a9

This model is a fine-tuned version of google-bert/bert-base-multilingual-uncased on the dair-ai/emotion [split] dataset. It achieves the following results on the evaluation set:

• Loss: 0.2640
• Data Size: 1.0
• Epoch Runtime: 27.4815
• Accuracy: 0.9199
• F1 Macro: 0.8807

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	1.7736	0	1.5029	0.3458	0.0859
No log	1	500	1.5967	0.0078	1.9621	0.2999	0.0878
No log	2	1000	1.5693	0.0156	2.1892	0.3488	0.0862
No log	3	1500	1.3821	0.0312	2.8265	0.5060	0.2024
No log	4	2000	1.0741	0.0625	3.7015	0.6099	0.3092
0.065	5	2500	0.5465	0.125	5.4274	0.8362	0.7731
0.4022	6	3000	0.3237	0.25	8.6450	0.8896	0.8204
0.0444	7	3500	0.2789	0.5	14.7843	0.9178	0.8692
0.2108	8.0	4000	0.1982	1.0	28.8195	0.9173	0.8672
0.138	9.0	4500	0.1903	1.0	28.0745	0.9249	0.8776
0.1396	10.0	5000	0.1988	1.0	27.8459	0.9309	0.8922
0.1027	11.0	5500	0.1665	1.0	28.3796	0.9294	0.8967
0.1295	12.0	6000	0.2243	1.0	27.2293	0.9204	0.8719
0.157	13.0	6500	0.2233	1.0	27.7196	0.9204	0.8804
0.1134	14.0	7000	0.2371	1.0	28.0744	0.9199	0.8740
0.1262	15.0	7500	0.2640	1.0	27.4815	0.9199	0.8807

Framework versions

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