yesj1234/mbart-mmt_15p_ko-ja

tst-translation-output

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: 0.8968
• Bleu: 9.457
• Gen Len: 17.4895

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: 2
• eval_batch_size: 2
• seed: 42
• distributed_type: multi-GPU
• num_devices: 4
• gradient_accumulation_steps: 4
• total_train_batch_size: 32
• total_eval_batch_size: 8
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 1500
• num_epochs: 30

Training results

Training Loss	Epoch	Step	Bleu	Gen Len	Validation Loss
1.1751	0.47	3000	7.6766	17.4644	1.1289
1.0268	0.93	6000	9.2277	17.7668	0.9895
0.8075	1.4	9000	9.1197	17.6811	0.9457
0.8082	1.87	12000	8.4837	17.4826	0.9053
0.5841	2.33	15000	9.8887	17.5166	0.9303
0.6142	2.8	18000	9.547	17.426	0.9142
0.4119	3.26	21000	9.5055	17.3378	0.9879
0.2837	7.46	24000	11.0549	17.2982	1.0063
0.1792	8.4	27000	8.9031	17.2801	1.0856
0.1204	9.33	30000	1.1643	11.3498	17.2986
0.0826	10.26	33000	1.2319	10.796	17.3627
0.0617	11.19	36000	1.2785	10.6211	17.3748
0.0523	12.13	39000	1.3217	9.8848	17.3358

Framework versions

• Transformers 4.34.0
• Pytorch 2.1.0+cu121
• Datasets 2.14.5
• Tokenizers 0.14.1