mBART

This model is a fine-tuned version of facebook/mbart-large-50-many-to-many-mmt on the None dataset. It achieves the following results on the evaluation set:

• Loss: 2.8059
• Bleu: 12.5808

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: 16
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 32
• optimizer: Use adamw_torch_fused with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: cosine
• lr_scheduler_warmup_ratio: 0.06
• num_epochs: 30
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Bleu
1.8741	1.0	96	1.8915	13.3768
1.3859	2.0	192	1.9221	13.4097
1.0343	3.0	288	1.9922	13.6703
0.6956	4.0	384	2.2070	13.3118
0.4563	5.0	480	2.4922	12.2720
0.2949	6.0	576	2.8059	12.5808

Framework versions

• Transformers 4.55.4
• Pytorch 2.8.0+cu128
• Datasets 4.1.1
• Tokenizers 0.21.4