Clémentine Fourrier

Replace constrained strings (model/task names...) with enums.

14

A number of strings across the project are actually constrained in value (task names, model names, ...). Using enums instead of strings would (on the user side): - constrain choice...

TokenGT

25

### Model description Adding the TokenGT graph transformer model with @Raman-Kumar (see [Graphormer issue](https://github.com/huggingface/transformers/issues/20962#issuecomment-1375361519)) @Raman-Kumar I'll create a PR with what I had ported of TokenGT at the end of...

Adding LRGB to the HuggingFace hub

27

Hi! @migalkin suggested [on Twitter](https://twitter.com/michael_galkin/status/1623049877122076673) adding your datasets to the HuggingFace hub, which I think is a super cool idea, so I'm opening this issue to see if you need...

Homogeneize logging system

14

Make it so both model types can run tensorboard evals. This should be good to go, but will need to wait for an update in `huggingface_hub` before merging, as the...

Allow to launch TGI models with the Inference Endpoints code

1

New cherrypicked branch to avoid the hassle of rebasing.

This PR needs #158 to be merged first. The main problem is that splits are no longer by size, but they are however now making sure that all batched generative...

At the moment, we have two ways of managing logging, depending on whether we come from nanotron or accelerate models. Nanotron use tensorboard for ex, which accelerate does not

Add `ligheval --list-tasks`

4

This command should list all the available tasks (in extended tasks and the task table)