Add LVFace library #1685
Add LVFace library #1685
Conversation
GitHup2016yjh
requested review from
SBrandeis,
gary149,
Wauplin,
julien-c,
pcuenca and
ngxson
as code owners
| @@ -1703,6 +1703,24 @@ export const vfimamba = (model: ModelData): string[] => [ | |||
| model = Model.from_pretrained("${model.id}")`, | |||
| ]; | |||
|
|
|||
| export const lvface = (model: ModelData): string[] => [ | |||
| `## Initialize the inferencer from inference_onnx.py | |||
There was a problem hiding this comment.
can you please add imports 🥹 feel free to remove comments to make it more minimal
There was a problem hiding this comment.
yes^ - let's make the snippets complete and minimal.
There was a problem hiding this comment.
Excuse me, would it be necessary for me to delete some of the comments to reduce the number of lines taken by them?
There was a problem hiding this comment.
I think in general we keep them minimal, so lesser the better
There was a problem hiding this comment.
I think in general we keep them minimal, so lesser the better
Thank you for the reminder. I have already compressed it to just one line.
There was a problem hiding this comment.
Apologies for the confusion here @GitHup2016yjh - what we meant by minimal snippet was to have something that enables people to test your model in say a Google Colab without looking at your GitHub repo or documentation for more details.
You can see couple of examples of this in the rest of the snippets in the same file.
An ideal snippet has three main parts:
- Setting up environment in the commments (# pip instal x y z)
- Load the model weights from hugging face (inferencer in your case)
- Run inference on a sample input
Let me know if this is not clear - happy to elaborate more.
Thanks again for your contribution!
There was a problem hiding this comment.
Apologies for the confusion here @GitHup2016yjh - what we meant by minimal snippet was to have something that enables people to test your model in say a Google Colab without looking at your GitHub repo or documentation for more details.
You can see couple of examples of this in the rest of the snippets in the same file.
An ideal snippet has three main parts:
- Setting up environment in the commments (# pip instal x y z)
- Load the model weights from hugging face (inferencer in your case)
- Run inference on a sample input
Let me know if this is not clear - happy to elaborate more.
Thanks again for your contribution!
Thank you for the reminder. I have revised this part of the content. Is it correct now?
|
would be great if files could be linted! 🙏🏻💗 |
There was a problem hiding this comment.
Thanks for the PR! please see my comments.
More generally we recommend have one model checkpoint/ type per repo (the hub is optimised to work like that).
In your case LVFace has 4 different types/ flavours of weights which makes download tracking a bit difficult.
We recommend splitting them if possible. If not, then do make a note for next release!
Co-authored-by: vb <[email protected]>
There are quite a few of these models, and they are relatively fragmented, so I'm thinking of releasing them together. |
There was a problem hiding this comment.
There are quite a few of these models, and they are relatively fragmented, so I'm thinking of releasing them together.
Okay - fair enough.
Could you please tell me how I should make a note for the next release?
Ah, I just meant if/ when you release other family of models please make sure to split them in different repos (so more a note for yourself to keep in mind)
Alright, thank you for the reminder. |
This PR adds LVFace, a new ViT-based face model, as a library. Thanks.