119f1a07eca6f050f0d2c7cfdc8fcaf6

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

• Loss: 2.6363
• Data Size: 1.0
• Epoch Runtime: 97.9216
• Bleu: 7.5389

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	6.8964	0	8.4234	0.3135
No log	1	367	3.4586	0.0078	9.2410	3.9162
No log	2	734	2.9534	0.0156	10.6761	4.4737
No log	3	1101	2.8112	0.0312	12.5653	4.9819
No log	4	1468	2.6778	0.0625	16.4546	5.3239
0.1475	5	1835	2.5812	0.125	23.0314	5.6594
2.4934	6	2202	2.4550	0.25	35.1775	6.2832
2.2634	7	2569	2.3242	0.5	54.2946	7.2971
1.9688	8.0	2936	2.2277	1.0	98.3470	8.4064
1.631	9.0	3303	2.2411	1.0	96.9782	8.2178
1.3035	10.0	3670	2.3332	1.0	96.5266	8.1844
1.0268	11.0	4037	2.4907	1.0	97.1212	8.7896
0.8207	12.0	4404	2.6363	1.0	97.9216	7.5389

Framework versions

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