aslandmsl

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: 0.1038
• Bleu: 93.2487
• Ter: 3.7344

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: 1e-05
• train_batch_size: 32
• 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
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Bleu	Ter
No log	1.0	386	0.5826	70.7840	19.9252
0.9041	2.0	772	0.3933	80.1642	12.2747
0.3581	3.0	1158	0.3187	83.6327	9.9116
0.2501	4.0	1544	0.2719	85.2170	8.7555
0.2501	5.0	1930	0.2431	86.2581	8.0075
0.1811	6.0	2316	0.2217	87.5235	7.1064
0.1395	7.0	2702	0.2080	81.1826	8.6705
0.1136	8.0	3088	0.1989	88.4843	6.4944
0.1136	9.0	3474	0.1878	88.9931	6.1034
0.0946	10.0	3860	0.1786	89.5849	5.8484
0.078	11.0	4246	0.1737	89.5253	5.6613
0.0683	12.0	4632	0.1713	89.5538	5.7293
0.0576	13.0	5018	0.1685	90.0542	5.4913
0.0576	14.0	5404	0.1656	90.3780	5.2703
0.0511	15.0	5790	0.1628	90.7216	5.1003
0.0439	16.0	6176	0.1621	90.5371	5.1513
0.0434	17.0	6562	0.1617	90.4864	5.2363
0.0434	18.0	6948	0.1605	90.6663	5.1853
0.0365	19.0	7334	0.1583	90.7443	5.1003
0.0347	20.0	7720	0.1575	90.6743	5.1003
0.0316	21.0	8106	0.1563	90.8012	5.1173
0.0316	22.0	8492	0.1563	90.8744	5.1173
0.0293	23.0	8878	0.1557	90.9605	5.0663
0.0276	24.0	9264	0.1559	91.3092	4.9643
0.0254	25.0	9650	0.1557	91.2291	4.9813
0.0249	26.0	10036	0.1559	91.2437	5.0153
0.0249	27.0	10422	0.1554	91.3500	4.9303
0.0236	28.0	10808	0.1554	91.4192	4.8963
0.0233	29.0	11194	0.1554	91.4589	4.8963
0.0226	30.0	11580	0.1553	91.4589	4.8963

Framework versions

• Transformers 4.44.2
• Pytorch 2.5.0+cu121
• Datasets 3.1.0
• Tokenizers 0.19.1