esp-to-lsm-model-split

This model is a fine-tuned version of Helsinki-NLP/opus-mt-es-es on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.5118
• Bleu: 85.2207
• Ter: 8.9524
• Rouge1: 0.9459
• Rouge2: 0.8835

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: 0.00015
• train_batch_size: 32
• eval_batch_size: 32
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 20
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Bleu	Ter	Rouge1	Rouge2
0.997	1.0	75	0.7490	39.0819	32.0952	0.8972	0.8107
0.4353	2.0	150	0.5363	78.0900	11.9048	0.9174	0.8421
0.2602	3.0	225	0.5100	47.9363	15.3333	0.9230	0.8407
0.2316	4.0	300	0.4974	79.0734	11.5238	0.9214	0.8392
0.1203	5.0	375	0.4946	80.3010	10.7619	0.9343	0.8695
0.1216	6.0	450	0.5038	80.6212	10.6667	0.9313	0.8571
0.0754	7.0	525	0.4650	69.5096	9.1429	0.9366	0.8773
0.0848	8.0	600	0.5086	83.6847	9.6190	0.9340	0.8599
0.0504	9.0	675	0.4904	82.1068	11.2381	0.9389	0.8777
0.0367	10.0	750	0.5040	84.9459	9.4286	0.9425	0.8787
0.0386	11.0	825	0.4964	84.3063	9.6190	0.9417	0.8770
0.0377	12.0	900	0.4992	84.2254	9.2381	0.9437	0.8809
0.0244	13.0	975	0.5052	84.7862	9.4286	0.9407	0.8762
0.0237	14.0	1050	0.5037	82.0504	9.8095	0.9396	0.8706
0.0173	15.0	1125	0.5105	85.2564	9.0476	0.9453	0.8821
0.0193	16.0	1200	0.5129	85.2911	8.9524	0.9460	0.8854
0.0184	17.0	1275	0.5121	84.8249	9.2381	0.9444	0.8808
0.0142	18.0	1350	0.5088	85.2207	8.9524	0.9459	0.8835
0.0131	19.0	1425	0.5127	85.2207	8.9524	0.9459	0.8835
0.0122	20.0	1500	0.5118	85.2207	8.9524	0.9459	0.8835

Framework versions

• Transformers 4.26.1
• Pytorch 2.5.0+cu121
• Datasets 3.0.2
• Tokenizers 0.13.3