ind-to-bbc-nmt-v8

This model is a fine-tuned version of facebook/nllb-200-distilled-600M on the nusatranslation_mt dataset. It achieves the following results on the evaluation set:

• Loss: 1.1714
• Sacrebleu: 31.404
• Gen Len: 45.259

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: 32
• eval_batch_size: 16
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 10
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Sacrebleu	Gen Len
5.9718	1.0	207	3.5105	24.1454	44.5145
2.5392	2.0	414	1.6246	28.1486	45.45
1.46	3.0	621	1.3187	30.5425	45.375
1.2013	4.0	828	1.2437	31.2443	45.2075
1.0869	5.0	1035	1.2084	31.0749	45.3445
1.0083	6.0	1242	1.1851	31.167	45.35
0.9563	7.0	1449	1.1811	31.2377	45.344
0.9149	8.0	1656	1.1719	31.2539	45.343
0.8881	9.0	1863	1.1738	31.5399	45.145
0.872	10.0	2070	1.1714	31.404	45.259

Framework versions

• Transformers 4.41.2
• Pytorch 2.3.0+cu121
• Datasets 2.14.6
• Tokenizers 0.19.1