2120mslsamples

This model is a fine-tuned version of Helsinki-NLP/opus-mt-es-es on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.4982
• Bleu Msl: 79.5487
• Bleu Asl: 0
• Ter Msl: 11.4533
• Ter Asl: 100

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: 0.0001
• train_batch_size: 32
• eval_batch_size: 64
• seed: 42
• optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• num_epochs: 30
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Bleu Msl	Bleu Asl	Ter Msl	Ter Asl
No log	1.0	67	1.0924	16.5734	0	78.3784	100
No log	2.0	134	0.5283	66.9864	0	17.2037	100
No log	3.0	201	0.4500	32.7340	0	24.8960	100
No log	4.0	268	0.4546	64.2241	0	17.0998	100
No log	5.0	335	0.4371	79.5661	0	12.3181	100
No log	6.0	402	0.4183	62.3356	0	12.4740	100
No log	7.0	469	0.4174	78.8409	0	12.2141	100
0.5741	8.0	536	0.4236	73.4058	0	13.0977	100
0.5741	9.0	603	0.4365	80.4263	0	11.6424	100
0.5741	10.0	670	0.4230	78.2192	0	11.5904	100
0.5741	11.0	737	0.4461	81.5222	0	11.0187	100
0.5741	12.0	804	0.4361	81.0763	0	10.8108	100
0.5741	13.0	871	0.4462	80.2271	0	11.4345	100
0.5741	14.0	938	0.4487	81.2932	0	10.0832	100
0.0399	15.0	1005	0.4535	82.0574	0	10.1351	100
0.0399	16.0	1072	0.4561	81.1900	0	10.7069	100
0.0399	17.0	1139	0.4498	80.5848	0	10.3430	100
0.0399	18.0	1206	0.4704	81.2086	0	10.4470	100
0.0399	19.0	1273	0.4780	83.3481	0	9.7713	100
0.0399	20.0	1340	0.4697	82.5737	0	9.8753	100
0.0399	21.0	1407	0.4675	82.5187	0	9.8753	100
0.0399	22.0	1474	0.4678	82.9501	0	10.0312	100
0.0155	23.0	1541	0.4708	82.3160	0	9.9792	100
0.0155	24.0	1608	0.4732	82.1759	0	10.0832	100
0.0155	25.0	1675	0.4777	82.5015	0	9.8233	100
0.0155	26.0	1742	0.4723	81.7682	0	10.1351	100
0.0155	27.0	1809	0.4755	82.1104	0	9.9792	100
0.0155	28.0	1876	0.4763	82.2547	0	10.1351	100
0.0155	29.0	1943	0.4756	82.2633	0	10.0832	100
0.01	30.0	2010	0.4761	82.2721	0	10.0312	100

Framework versions

• Transformers 4.49.0
• Pytorch 2.5.1+cu124
• Datasets 3.3.2
• Tokenizers 0.21.0