a0b509d04caacd2bb4d6ecd6e801855d

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

• Loss: 1.4995
• Data Size: 1.0
• Epoch Runtime: 805.4644
• Bleu: 13.9721

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	3.1313	0	65.5918	7.9757
No log	1	3177	1.8461	0.0078	71.3801	14.9232
0.0301	2	6354	1.7517	0.0156	78.1510	15.4778
1.7062	3	9531	1.6580	0.0312	89.6609	20.4990
1.612	4	12708	1.5721	0.0625	113.7205	20.3604
1.4837	5	15885	1.4794	0.125	160.5438	15.9832
1.3806	6	19062	1.3938	0.25	249.6234	15.3109
1.2365	7	22239	1.3109	0.5	434.6573	19.4031
1.1171	8.0	25416	1.2416	1.0	805.8557	15.3693
0.9203	9.0	28593	1.2525	1.0	808.9224	14.7216
0.8008	10.0	31770	1.3020	1.0	808.8734	13.9900
0.6541	11.0	34947	1.3781	1.0	806.7826	13.6572
0.5121	12.0	38124	1.4995	1.0	805.4644	13.9721

Framework versions

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