147f213713b70db1bbfe9ac82467db67

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

• Loss: 2.9535
• Data Size: 1.0
• Epoch Runtime: 12.5719
• Bleu: 9.9988

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	7.5027	0	1.1821	0.4099
No log	1	27	6.3237	0.0078	1.8547	0.5171
No log	2	54	5.3185	0.0156	2.1646	0.8354
No log	3	81	4.5542	0.0312	3.1090	1.7122
No log	4	108	3.9624	0.0625	4.9654	2.3570
No log	5	135	3.4286	0.125	6.4562	3.4324
No log	6	162	3.0683	0.25	7.8861	4.5550
No log	7	189	2.7720	0.5	9.1378	5.5939
0.6477	8.0	216	2.5398	1.0	12.5518	6.9088
0.6477	9.0	243	2.5423	1.0	11.1810	8.5070
1.8161	10.0	270	2.6157	1.0	11.4520	10.1818
1.8161	11.0	297	2.7531	1.0	11.7966	9.9224
1.0311	12.0	324	2.9535	1.0	12.5719	9.9988

Framework versions

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