b9e9f41e77374d9a6bbee21b3e9b1e75

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

• Loss: 2.3350
• Data Size: 1.0
• Epoch Runtime: 176.5877
• Bleu: 7.9937

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	6.5825	0	15.1943	0.6759
No log	1	684	2.7787	0.0078	16.4378	3.9950
No log	2	1368	2.5828	0.0156	18.8634	4.8094
No log	3	2052	2.4816	0.0312	21.9964	5.8867
No log	4	2736	2.3685	0.0625	26.7889	8.9320
2.3351	5	3420	2.2677	0.125	36.7512	8.2855
2.1996	6	4104	2.1496	0.25	57.0899	9.7158
1.9464	7	4788	2.0444	0.5	97.2348	12.4715
1.7288	8.0	5472	1.9489	1.0	174.4436	10.8355
1.4184	9.0	6156	1.9837	1.0	174.1284	12.1745
1.1338	10.0	6840	2.0624	1.0	174.1556	9.7148
0.9011	11.0	7524	2.1731	1.0	174.7115	9.5537
0.6888	12.0	8208	2.3350	1.0	176.5877	7.9937

Framework versions

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