12000samples

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: 0.1451
• Bleu Msl: 92.1051
• Bleu Asl: 0
• Ter Msl: 4.7161
• 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: 1e-05
• train_batch_size: 32
• eval_batch_size: 64
• seed: 42
• optimizer: Use 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	386	0.9869	20.9686	82.0457	66.6038	9.1959
1.5187	2.0	772	0.4370	48.4639	86.5183	36.0377	6.8058
0.5406	3.0	1158	0.2964	60.2032	88.7852	26.3208	5.6917
0.2888	4.0	1544	0.2504	65.3635	90.1910	22.3585	4.9828
0.2888	5.0	1930	0.2023	72.9532	91.6531	17.4528	4.2941
0.2019	6.0	2316	0.1680	62.1974	92.4105	19.8113	3.9093
0.1391	7.0	2702	0.1532	75.8711	92.9684	15.2830	3.6864
0.1058	8.0	3088	0.1407	47.4649	93.0796	26.0377	3.5852
0.1058	9.0	3474	0.1361	77.0535	93.1579	14.2453	3.5042
0.0868	10.0	3860	0.1288	59.0665	93.5742	18.4906	3.3421
0.0729	11.0	4246	0.1239	80.7391	93.4642	12.6415	3.3826
0.0629	12.0	4632	0.1206	49.3562	93.5802	23.6792	3.2408
0.054	13.0	5018	0.1185	77.7535	94.0327	12.8302	3.0788
0.054	14.0	5404	0.1170	79.9193	93.7494	12.4528	3.1193
0.0481	15.0	5790	0.1133	79.8567	93.7407	11.4151	2.9978
0.0425	16.0	6176	0.1131	83.0857	93.6535	10.7547	2.8560
0.0385	17.0	6562	0.1121	82.8013	94.0053	10.7547	2.6737
0.0385	18.0	6948	0.1112	84.5147	93.9244	10.1887	2.6939
0.0359	19.0	7334	0.1108	83.7256	94.1014	10.0943	2.6939
0.032	20.0	7720	0.1103	82.0662	94.2459	10.4717	2.7952
0.0298	21.0	8106	0.1099	84.3243	94.5135	10.1887	2.6939
0.0298	22.0	8492	0.1105	83.9736	94.5280	10.1887	2.7345
0.0275	23.0	8878	0.1091	84.6834	94.4765	10.0	2.7547
0.0261	24.0	9264	0.1100	84.3382	94.6091	10.1887	2.6737
0.0247	25.0	9650	0.1099	83.8395	94.5048	10.0943	2.7750
0.024	26.0	10036	0.1104	83.1135	94.5483	10.1887	2.7547
0.024	27.0	10422	0.1100	83.5175	94.5483	10.2830	2.7547
0.0227	28.0	10808	0.1101	84.2731	94.5440	10.0943	2.7750
0.022	29.0	11194	0.1100	84.3292	94.5440	10.0943	2.7750
0.0224	30.0	11580	0.1101	83.8359	94.5440	10.1887	2.7750

Framework versions

• Transformers 4.46.2
• Pytorch 2.5.1+cu121
• Datasets 3.1.0
• Tokenizers 0.20.3