81d3a4079ce7e480f57d39154cd51d85

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

• Loss: 2.1736
• Data Size: 1.0
• Epoch Runtime: 323.0230
• Bleu: 11.8442

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	3.2007	0	27.1672	7.4798
No log	1	1286	2.0052	0.0078	29.9548	7.7728
0.0408	2	2572	1.9307	0.0156	33.6060	8.3127
0.0523	3	3858	1.8820	0.0312	38.5140	8.5998
0.0791	4	5144	1.8436	0.0625	48.1833	9.6184
1.7606	5	6430	1.7948	0.125	66.3032	10.4245
1.6972	6	7716	1.7471	0.25	103.2815	9.4626
1.5887	7	9002	1.7021	0.5	180.0488	9.8835
1.47	8.0	10288	1.6586	1.0	326.3577	10.0040
1.1791	9.0	11574	1.7114	1.0	324.9995	26.0055
0.9299	10.0	12860	1.8308	1.0	326.3272	12.2330
0.7161	11.0	14146	1.9903	1.0	326.3843	10.3464
0.5721	12.0	15432	2.1736	1.0	323.0230	11.8442

Framework versions

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