b00d0be34d68b85e7a2a19f4682e69c9

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

• Loss: 3.6699
• Data Size: 1.0
• Epoch Runtime: 12.2692
• Bleu: 6.9891

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	7.6633	0	1.2762	0.7306
No log	1	29	6.8516	0.0078	1.6449	0.6988
No log	2	58	6.4197	0.0156	3.2571	0.8566
No log	3	87	6.1265	0.0312	4.1474	1.1336
No log	4	116	5.7254	0.0625	5.3474	1.3364
No log	5	145	5.1175	0.125	7.7061	1.9111
0.5244	6	174	4.4346	0.25	9.1575	3.2009
0.5244	7	203	3.8842	0.5	10.4727	4.0336
0.5244	8.0	232	3.4406	1.0	12.8134	5.1309
2.132	9.0	261	3.2886	1.0	11.7808	6.1192
2.132	10.0	290	3.3468	1.0	11.9443	6.8745
1.9583	11.0	319	3.4332	1.0	13.1868	6.8699
1.9583	12.0	348	3.5305	1.0	12.1705	7.4683
1.2707	13.0	377	3.6699	1.0	12.2692	6.9891

Framework versions

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