vilbert_vqa

Submitting an experiment on beaker

1. Edit the training config ./vilbert_vqa_from_huggingface.jsonnet or create a new one.

2. Commit and push your changes to the training config.

3. Edit the beaker config ./beaker.yml to point to the commit of the training config.

4. Submit the beaker experiment:

   beaker experiment create \
       --workspace ai2/ViLBERT-VQA \  # change as needed
       --file beaker.yml \
       --name try-030  # change as needed