d7ce87e738c963c83a04a1bd5dba8c77

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

• Loss: 2.8507
• Data Size: 1.0
• Epoch Runtime: 14.3309
• Bleu: 10.5215

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	7.1829	0	1.3389	1.0443
No log	1	29	4.7763	0.0078	1.9206	2.3895
No log	2	58	3.8623	0.0156	3.1141	4.3031
No log	3	87	3.4634	0.0312	4.3287	5.0885
No log	4	116	3.1043	0.0625	5.1385	5.2172
No log	5	145	2.8907	0.125	6.6665	6.5922
0.3218	6	174	2.7131	0.25	8.2420	14.5445
0.3218	7	203	2.5073	0.5	10.0555	13.8589
0.3218	8.0	232	2.4139	1.0	13.3983	9.8310
1.2527	9.0	261	2.4392	1.0	12.0660	10.3404
1.2527	10.0	290	2.5787	1.0	12.9503	8.8579
1.0153	11.0	319	2.7293	1.0	13.2793	9.4749
1.0153	12.0	348	2.8507	1.0	14.3309	10.5215

Framework versions

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