koja_mbartLarge_55p_run1

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: 0.9008
• Bleu: 8.9861
• Gen Len: 17.9149

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: 4
• eval_batch_size: 4
• seed: 42
• distributed_type: multi-GPU
• num_devices: 4
• gradient_accumulation_steps: 2
• total_train_batch_size: 32
• total_eval_batch_size: 16
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 10

Training results

Training Loss	Epoch	Step	Validation Loss	Bleu	Gen Len
1.0254	0.48	8000	1.0065	7.1252	18.5403
0.9374	0.97	16000	0.9200	8.2047	18.2091
0.7429	1.45	24000	0.9086	8.4622	17.9606
0.7039	1.93	32000	0.9008	8.9198	17.8841
0.5616	2.42	40000	0.9210	8.7938	17.78
0.5355	2.9	48000	0.9400	8.6506	17.9146
0.4194	3.38	56000	1.0054	8.4235	17.6759
0.3852	3.86	64000	1.0154	8.5886	17.7885

Framework versions

• Transformers 4.34.1
• Pytorch 2.1.0+cu121
• Datasets 2.14.6
• Tokenizers 0.14.1