google-deepmind / google-deepmind/alphageometry
Consider hosting weights on huggingface.co :hugs:
- Dominant language
- Python
- Stars
- 4.9k
- Forks
- 572
- PR merge metrics
- No merged PRs in 30d
Description
Hey there :wave: I'm opening this issue to suggest moving the weights files from the GDrive to the [Hugging Face Hub](https://huggingface.co/). The weights files would be publicly available, discoverable under the google deepmind organization, versioned and stored along their model card. At the moment, I've uploaded the files to [Wauplin/alphageometry](https://huggingface.co/Wauplin/alphageometry) for a test but the idea would be to move the repo to a more appropriate organization (either [google](https://huggingface.co/google) or [deepmind](https://huggingface.co/deepmind)?).
Once hosted on the Hub, you would just need to update the [download.sh](https://github.com/google-deepmind/alphageometry/blob/b20b036787b50d166461c08b2dd4e502fe8e4d15/download.sh#L16) and [run.sh](https://github.com/google-deepmind/alphageometry/blob/b20b036787b50d166461c08b2dd4e502fe8e4d15/run.sh#L25) scripts to use the `huggingface-cli` tool.
Instead of:
```sh
gdown --folder https://bit.ly/alphageometry
DATA=ag_ckpt_vocab
```
you would have:
```sh
huggingface-cli download google-deepmind/alphageometry --local-dir ag_ckpt_vocab
DATA=ag_ckpt_vocab
```
`huggingface-cli` is the CLI tool shipped with the `huggingface_hub` Python package. I saw that this is already a dependency of the project. I would suggest to update the pinned version (currently `0.17.3`) to latest release (`0.20.2`) which include some bug fixes.
Please let me know what you think! :hugs:
**Disclaimer:** I am myself working at Hugging Face as `huggingface_hub`'s maintainer.
___
**Note:** not related, but I didn't manage to use `gdown` correctly (with `gdown 4.7.3`). I'm getting either
```
> gdown --folder https://bit.ly/alphageometry
Retrieving folder contents
Error:
file/folder name cannot be extracted from:
ag_ckpt_vocab – Google Drive
To report issues, please visit https://github.com/wkentaro/gdown/issues.
```
or
```
gdown --folder https://drive.google.com/drive/u/0/folders/1ZLaZ2ajtOcILDWa5ePPLX1bmaf_BNRZV
Retrieving folder contents
(...)
Failed to retrieve file url:
Too many users have viewed or downloaded this file recently. Please
try accessing the file again later. If the file you are trying to
access is particularly large or is shared with many people, it may
take up to 24 hours to be able to view or download the file. If you
still can't access a file after 24 hours, contact your domain
administrator.
You may still be able to access the file from the browser:
https://drive.google.com/uc?id=1qXkmmgoJ8oTYJdFV1xw0xGPpQj6SyOYA
but Gdown can't. Please check connections and permissions.
```
Contributor guide
Assessment
This issue has not been assessed yet.