search

facebookresearch / fairseq

Facebook AI Research Sequence-to-Sequence Toolkit written in Python.
MIT License
30.61k stars 6.42k forks source link

Fine tuning NLLB #4870

Closed ibtiRaj closed 1 year ago

ibtiRaj commented 2 years ago

❓ Questions and Help

Before asking:

  1. search the issues.
  2. search the docs.

    My question?

    I am trying to fine-tune the NLLB model on the Moroccan Arabic and English languages. I prepared my data using the "Filtering and Preparing the Data" pipeline as described in the ReadMe. The result of the Prepare Data step is the following directory: image

The second step now is to fine tune the model, I used this command:

DROP=0.1 python examples/nllb/modeling/train/train_script.py cfg=nllb200_dense3.3B_finetune_on_fbseed cfg/dataset=bilingual cfg.dataset.lang_pairs="ary_Arab-eng_Latn" cfg.fairseq_root=/home/docker/notebooks/khadija/Translation_with_NLLB-200/fairseq cfg.output_dir=model_fine_tuned cfg.dropout=$DROP cfg.warmup=10 cfg.finetune_from_model=/home/docker/notebooks/khadija/Translation_with_NLLB-200/checkpoint.pt

The bilingual.yaml file is as follows:

image

But I get this error when I run the command:

image

image

Can you help me please? Is there anything I've missed?

What's your environment?

qunash commented 1 year ago

Hi, could you share your insights on this issue? Were you able to fine-tune the model?