1946a7117d7a8124779a1d08e32b6d9b

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

• Loss: 0.2571
• Data Size: 1.0
• Epoch Runtime: 28.6617
• Accuracy: 0.9199
• F1 Macro: 0.8795

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	1.8301	0	1.6031	0.0761	0.0409
No log	1	500	1.6370	0.0078	2.0821	0.2908	0.0751
No log	2	1000	1.6015	0.0156	2.2148	0.3488	0.0862
No log	3	1500	1.5740	0.0312	2.8934	0.3488	0.0862
No log	4	2000	1.5640	0.0625	3.7276	0.3543	0.1315
0.0805	5	2500	1.0444	0.125	5.4327	0.6537	0.4254
0.6917	6	3000	0.4829	0.25	8.6456	0.8508	0.7472
0.0509	7	3500	0.3162	0.5	15.5057	0.8967	0.8480
0.2485	8.0	4000	0.2465	1.0	29.6891	0.9088	0.8741
0.1891	9.0	4500	0.2971	1.0	27.3833	0.9078	0.8598
0.1618	10.0	5000	0.2139	1.0	28.0175	0.9173	0.8686
0.1376	11.0	5500	0.2075	1.0	29.1736	0.9163	0.8794
0.1359	12.0	6000	0.2999	1.0	28.7744	0.9158	0.8597
0.1703	13.0	6500	0.2485	1.0	27.9716	0.9239	0.8748
0.1163	14.0	7000	0.2355	1.0	30.0916	0.9209	0.8851
0.1684	15.0	7500	0.2571	1.0	28.6617	0.9199	0.8795

Framework versions

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