m2m100-es-zap-v3

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: 1.4560
• Bleu: 28.9525
• Ter: 55.6298
• Meteor: 0.5493
• Chrf: 54.9062

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: 8
• eval_batch_size: 8
• seed: 42
• 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: 50
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Bleu	Ter	Meteor	Chrf
2.6709	1.0	179	1.8407	14.6410	79.9618	0.3943	39.4821
1.2445	2.0	358	1.4534	23.1869	58.3969	0.4839	48.6435
0.7389	3.0	537	1.3348	24.1887	57.3473	0.5146	50.8358
0.4475	4.0	716	1.3474	25.4721	57.0611	0.5448	53.2475
0.2856	5.0	895	1.3585	29.3531	53.3397	0.5628	55.7674
0.1964	6.0	1074	1.4172	28.9094	55.8206	0.5490	54.8719
0.1524	7.0	1253	1.4145	27.3552	54.4847	0.5401	54.3912
0.1136	8.0	1432	1.4560	28.9525	55.6298	0.5493	54.9062

Framework versions

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