14661

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: 0.5018
• Bleu Msl: 0.0
• Bleu 1 Msl: 0.6475
• Bleu 2 Msl: 0.0148
• Bleu 3 Msl: 0.0045
• Bleu 4 Msl: 0.0023
• Ter Msl: 100
• Bleu Asl: 0
• Bleu 1 Asl: 0
• Bleu 2 Asl: 0
• Bleu 3 Asl: 0
• Bleu 4 Asl: 0
• 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 1 Msl	Bleu 2 Msl	Bleu 3 Msl	Bleu 4 Msl	Ter Msl	Bleu Asl	Bleu 1 Asl	Bleu 2 Asl	Bleu 3 Asl	Bleu 4 Asl	Ter Asl
No log	1.0	367	0.1799	8.9666	0.4818	0.0094	0.0027	0.0013	100	0.0	0.4901	0.0045	0.0010	0.0004	100
0.4236	2.0	734	0.1523	0.0	0.3855	0.0084	0.0025	0.0012	100	0.0	0.5128	0.0046	0.0010	0.0004	100
0.1025	3.0	1101	0.1202	13.9892	0.5545	0.0101	0.0028	0.0014	100	0.0	0.5132	0.0046	0.0010	0.0004	100
0.1025	4.0	1468	0.1315	63.8943	0.5818	0.0103	0.0028	0.0014	100	0.0	0.4780	0.0045	0.0010	0.0004	100
0.0587	5.0	1835	0.1235	84.0896	0.54	0.0099	0.0028	0.0013	100	0.0	0.5166	0.0047	0.0010	0.0004	100
0.0373	6.0	2202	0.1300	50.0000	0.5818	0.0103	0.0028	0.0014	100	0.0	0.5166	0.0047	0.0010	0.0004	100
0.0294	7.0	2569	0.1226	50.0000	0.6018	0.0105	0.0029	0.0014	100	0.0	0.5195	0.0047	0.0010	0.0004	100
0.0294	8.0	2936	0.1256	63.8943	0.6055	0.0105	0.0029	0.0014	100	0.0	0.4969	0.0046	0.0010	0.0004	100
0.022	9.0	3303	0.1258	63.8943	0.6273	0.0107	0.0029	0.0014	100	0.0	0.5136	0.0046	0.0010	0.0004	100
0.0173	10.0	3670	0.1221	63.8943	0.6218	0.0106	0.0029	0.0014	100	0.0	0.5283	0.0047	0.0010	0.0004	100
0.014	11.0	4037	0.1283	63.8943	0.6036	0.0105	0.0029	0.0014	100	0.0	0.5241	0.0047	0.0010	0.0004	100
0.014	12.0	4404	0.1263	86.9442	0.5655	0.0101	0.0028	0.0014	100	0.0	0.5187	0.0047	0.0010	0.0004	100
0.0123	13.0	4771	0.1313	0.0	0.58	0.0103	0.0028	0.0014	100	0.0	0.5279	0.0047	0.0010	0.0004	100
0.0112	14.0	5138	0.1323	37.9918	0.6127	0.0106	0.0029	0.0014	100	0.0	0.5157	0.0047	0.0010	0.0004	100
0.0081	15.0	5505	0.1294	63.8943	0.6273	0.0107	0.0029	0.0014	100	0.0	0.5275	0.0047	0.0010	0.0004	100
0.0081	16.0	5872	0.1336	63.8943	0.6091	0.0105	0.0029	0.0014	100	0.0	0.5191	0.0047	0.0010	0.0004	100
0.0073	17.0	6239	0.1300	63.8943	0.6055	0.0105	0.0029	0.0014	100	0.0	0.5052	0.0046	0.0010	0.0004	100
0.0061	18.0	6606	0.1353	70.7107	0.5945	0.0104	0.0029	0.0014	100	0.0	0.5132	0.0046	0.0010	0.0004	100
0.0061	19.0	6973	0.1271	63.8943	0.62	0.0106	0.0029	0.0014	100	0.0	0.5141	0.0046	0.0010	0.0004	100
0.0056	20.0	7340	0.1289	48.1098	0.5691	0.0102	0.0028	0.0014	100	0.0	0.5237	0.0047	0.0010	0.0004	100
0.0043	21.0	7707	0.1285	37.9918	0.6455	0.0108	0.0029	0.0014	100	0.0	0.5237	0.0047	0.0010	0.0004	100
0.004	22.0	8074	0.1308	46.7138	0.5218	0.0097	0.0027	0.0013	100	0.0	0.5208	0.0047	0.0010	0.0004	100
0.004	23.0	8441	0.1334	54.1082	0.5309	0.0098	0.0028	0.0013	100	0.0	0.5195	0.0047	0.0010	0.0004	100
0.0036	24.0	8808	0.1341	86.9442	0.5636	0.0101	0.0028	0.0014	100	0.0	0.5329	0.0047	0.0010	0.0004	100
0.0027	25.0	9175	0.1326	86.9442	0.5382	0.0099	0.0028	0.0013	100	0.0	0.5199	0.0047	0.0010	0.0004	100
0.0028	26.0	9542	0.1334	77.3055	0.5582	0.0101	0.0028	0.0014	100	0.0	0.5229	0.0047	0.0010	0.0004	100
0.0028	27.0	9909	0.1338	54.1082	0.5855	0.0103	0.0029	0.0014	100	0.0	0.5266	0.0047	0.0010	0.0004	100
0.0021	28.0	10276	0.1349	54.1082	0.5927	0.0104	0.0029	0.0014	100	0.0	0.5313	0.0047	0.0010	0.0004	100
0.0021	29.0	10643	0.1355	54.1082	0.5691	0.0102	0.0028	0.0014	100	0.0	0.5283	0.0047	0.0010	0.0004	100
0.0014	30.0	11010	0.1351	54.1082	0.5655	0.0101	0.0028	0.0014	100	0.0	0.5287	0.0047	0.0010	0.0004	100

Framework versions

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