1a66e2c47e83a58738ea159c5c561413

This model is a fine-tuned version of facebook/mbart-large-50-many-to-one-mmt on the Helsinki-NLP/opus_books [fr-pt] dataset. It achieves the following results on the evaluation set:

• Loss: 3.4652
• Data Size: 1.0
• Epoch Runtime: 12.9832
• Bleu: 11.8420

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
• distributed_type: multi-GPU
• num_devices: 4
• total_train_batch_size: 32
• total_eval_batch_size: 32
• optimizer: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: constant
• num_epochs: 50

Training results

Training Loss	Epoch	Step	Validation Loss	Data Size	Epoch Runtime	Bleu
No log	0	0	7.4151	0	1.2985	0.5926
No log	1	31	6.8392	0.0078	1.9730	0.7211
No log	2	62	6.4163	0.0156	2.2556	0.4802
No log	3	93	6.1240	0.0312	3.3280	0.6339
No log	4	124	5.6468	0.0625	4.2883	1.3557
No log	5	155	5.0081	0.125	5.5391	1.8463
No log	6	186	4.2160	0.25	7.4184	2.9782
0.7414	7	217	3.6862	0.5	9.4439	3.9537
0.7414	8.0	248	3.2009	1.0	13.8121	5.9473
2.1973	9.0	279	3.0768	1.0	13.3382	6.2971
2.1478	10.0	310	3.1174	1.0	13.6390	11.6622
2.1478	11.0	341	3.2147	1.0	13.7926	13.6522
1.4338	12.0	372	3.3403	1.0	12.5182	13.9521
0.9045	13.0	403	3.4652	1.0	12.9832	11.8420

Framework versions

• Transformers 4.57.0
• Pytorch 2.8.0+cu128
• Datasets 4.2.0
• Tokenizers 0.22.1