bbc-to-ind-nmt-v3

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: 1.1424
  • Sacrebleu: 37.0979
  • Gen Len: 37.248

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: 16
  • 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: 5
  • mixed_precision_training: Native AMP

Training results

Training Loss Epoch Step Validation Loss Sacrebleu Gen Len
4.2762 1.0 413 1.5344 30.2931 37.92
1.3119 2.0 826 1.1962 35.3764 37.3035
1.0696 3.0 1239 1.1529 36.5225 37.204
0.9763 4.0 1652 1.1420 36.8954 37.0565
0.9298 5.0 2065 1.1424 37.0979 37.248

Framework versions

  • Transformers 4.41.2
  • Pytorch 2.3.0+cu121
  • Datasets 2.14.6
  • Tokenizers 0.19.1
Downloads last month
15
Safetensors
Model size
615M params
Tensor type
F32
·
Inference Examples
This model does not have enough activity to be deployed to Inference API (serverless) yet. Increase its social visibility and check back later, or deploy to Inference Endpoints (dedicated) instead.

Model tree for kepinsam/bbc-to-ind-nmt-v3

Finetuned
(86)
this model

Evaluation results