m2m100-T4v2-es-zap-data

This model is a fine-tuned version of facebook/m2m100_418M on the None dataset. It achieves the following results on the evaluation set:

• Loss: 1.3573
• Bleu: 30.8447
• Ter: 52.0264
• Meteor: 0.5763
• Chrf: 56.8245

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: 5e-05
• train_batch_size: 4
• eval_batch_size: 4
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 16
• optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• num_epochs: 15
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Bleu	Ter	Meteor	Chrf
No log	1.0	90	1.9487	13.5871	71.1593	0.3703	38.8122
10.9829	2.0	180	1.5216	18.6256	64.8445	0.4659	45.8309
5.1399	3.0	270	1.3368	24.2671	61.3572	0.5057	49.7241
3.0537	4.0	360	1.2643	26.3137	56.8332	0.5261	52.3703
1.8906	5.0	450	1.2437	27.7429	55.3252	0.5407	53.5345
1.2006	6.0	540	1.2518	27.4709	56.1734	0.5452	54.8736
0.7365	7.0	630	1.2633	27.8943	54.6654	0.5634	55.4446
0.5015	8.0	720	1.3111	28.7957	54.0999	0.5564	55.3113
0.3393	9.0	810	1.2917	30.0222	52.1206	0.5699	55.9249
0.2421	10.0	900	1.3334	30.2220	53.2516	0.5645	56.6542
0.2421	11.0	990	1.3196	30.7501	53.5344	0.5653	56.1103
0.1677	12.0	1080	1.3200	30.0664	52.9689	0.5724	56.7284
0.1371	13.0	1170	1.3559	30.4747	52.4976	0.5721	57.1548
0.0937	14.0	1260	1.3600	30.1739	52.2149	0.5736	56.8540
0.0651	15.0	1350	1.3573	30.8447	52.0264	0.5763	56.8245

Framework versions

• Transformers 5.4.0
• Pytorch 2.10.0+cu128
• Datasets 4.8.4
• Tokenizers 0.22.2