mbart50_baseline

This model is a fine-tuned version of facebook/mbart-large-50-many-to-many-mmt on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 1.2537
• Sacrebleu: 34.65
• Chrf: 53.67
• Ter: 51.36

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
• optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• num_epochs: 2
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Sacrebleu	Chrf	Ter
1.3188	0.3000	5000	1.3907	32.32	51.86	53.73
1.2733	0.6001	10000	1.3395	32.72	51.74	53.11
1.2336	0.9001	15000	1.2857	33.53	52.74	52.66
1.0212	1.2001	20000	1.3058	33.75	52.81	52.47
0.9191	1.5002	25000	1.2757	34.22	53.04	51.74
0.9069	1.8002	30000	1.2600	34.54	53.59	51.47
0.9069	2.0	33330	1.2537	34.65	53.67	51.36

Framework versions

• Transformers 5.13.1
• Pytorch 2.11.0+cu128
• Datasets 4.0.0
• Tokenizers 0.22.2