ES-ENG-mBERT-sentiment

This model is a fine-tuned version of bert-base-multilingual-cased on a Custom dataset.

The best model (stopped after 14 epochs) achieves the following results on the evaluation set:

• Loss: 0.8110
• Accuracy: 0.6307
• F1: 0.6298
• Precision: 0.6291
• Recall: 0.6307

Intended uses & limitations

Note that commercial use with this model is prohibited.

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 2e-06
• train_batch_size: 64
• eval_batch_size: 64
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 30

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	F1	Precision	Recall
1.063	1.0	208	0.9989	0.4731	0.4044	0.4885	0.4731
0.9664	2.0	416	0.9144	0.5262	0.4845	0.5270	0.5262
0.9067	3.0	624	0.8648	0.5896	0.5844	0.5935	0.5896
0.8572	4.0	832	0.8294	0.6065	0.5984	0.6102	0.6065
0.8168	5.0	1040	0.8101	0.6107	0.6092	0.6119	0.6107
0.7897	6.0	1248	0.8213	0.6074	0.6015	0.6018	0.6074
0.7568	7.0	1456	0.7992	0.6194	0.6181	0.6176	0.6194
0.7465	8.0	1664	0.8089	0.6246	0.6183	0.6206	0.6246
0.7223	9.0	1872	0.7988	0.6236	0.6214	0.6207	0.6236
0.7045	10.0	2080	0.8390	0.6165	0.6080	0.6126	0.6165
0.6888	11.0	2288	0.8042	0.6291	0.6260	0.6257	0.6291
0.671	12.0	2496	0.8088	0.6239	0.6212	0.6216	0.6239
0.6543	13.0	2704	0.8104	0.6256	0.6227	0.6216	0.6256
0.6409	14.0	2912	0.8110	0.6307	0.6298	0.6291	0.6307
0.6275	15.0	3120	0.8127	0.6298	0.6292	0.6299	0.6298
0.6176	16.0	3328	0.8334	0.6252	0.6217	0.6206	0.6252
0.6096	17.0	3536	0.8331	0.6256	0.6210	0.6210	0.6256

Framework versions

• Transformers 4.29.2
• Pytorch 2.0.1+cu117
• Datasets 2.12.0
• Tokenizers 0.13.3