Git-lfs issue

    state_dict = load_state_dict(resolved_archive_file)
  File "/srv/conda/envs/notebook/lib/python3.7/site-packages/transformers/modeling_utils.py", line 467, in load_state_dict
    "You seem to have cloned a repository without having git-lfs installed. Please install "
OSError: You seem to have cloned a repository without having git-lfs installed. Please install git-lfs and run `git lfs install` followed by `git lfs pull` in the folder you cloned.

Hi,
I uploaded my models to the correct path using lfs. but my submission still failed due to this error.
Please help…

submission_hash : 4b49d2ec2cb3bf387261b20b48de51e08b2561cf.

1 Like

@Erica : This was happening because in your repository in your .gitattributes file you were not tracking the *.bin files (model weights) which were hosted using git-lfs.

For now, I have sent a merge request to your submission repository, fixing your .gitattributes so that it tracks the git-lfs files correctly. You can find it here : AIcrowd

For future reference, if inside your code repository, you have a model weight file called as “XYZ.bin”, then you can track all the model weight files with the same extension by running :

git lfs track "*.bin"

which should automatically update your .gitattributes file. This is the preferred way of including your model weights using git-lfs to ensure the overall docker image sizes do not blow up.

Additionally, just for this challenge, we are updating our evaluators to always do a separate git lfs pull after cloning your repository to avoid this error in the future (irrespective of if your .gitattributes file is correctly specified or not.)