opus-mt-en-bkm-10e

This model is a fine-tuned version of Helsinki-NLP/opus-mt-en-ro on the arrow dataset. It achieves the following results on the evaluation set:

• Loss: 1.5644
• Bleu: 7.1794
• Gen Len: 60.0222

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: 2e-05
• train_batch_size: 96
• eval_batch_size: 96
• seed: 42
• 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
No log	1.0	186	2.6868	0.299	76.178
No log	2.0	372	2.0190	2.2444	67.167
3.115	3.0	558	1.8364	4.5959	59.7357
3.115	4.0	744	1.7372	5.1827	61.7218
3.115	5.0	930	1.6732	5.8295	59.7346
1.8706	6.0	1116	1.6301	6.4389	60.9085
1.8706	7.0	1302	1.6002	6.6498	60.4191
1.8706	8.0	1488	1.5792	6.8315	60.2721
1.7133	9.0	1674	1.5680	7.1239	60.5609
1.7133	10.0	1860	1.5644	7.1794	60.0222

Framework versions

• Transformers 4.38.2
• Pytorch 2.2.1+cu121
• Datasets 2.18.0
• Tokenizers 0.15.2