Giovanni Puccetti

coca training doesn't work

2

I made a PR that should fix it #710

fix coca training

6

@rwightman I was too hasty with the `embed_cls` thing, I think now the model would not train. This should make it work with the tokenizer and train ok, fix #715

[WIP] Add mammut

6

Initial attempts at adding MaMMuT https://arxiv.org/abs/2303.16839 @mehdidc @JeniaJitsev @rom1504 I don´t know marianna github name can you tag her too? This is the PR, will add all of you as...

Add gen reg tests

9

This adds regression tests for generative models. @rwightman this should be almost done, however there seems to be a regression error for coca that I had not noticed

Make coca and HF work together

1

@ae86280 the HF support is indeed not good for coca yet this PR is to address this and also fix #445 @fedshyvana this will also remove the two unused args...

If you like a roc gif for the readme

Is the package still maintained?

3

Hi I would like to add a feature to the repo if it is still maintained. If it is can we talk about it?

@rwightman should add testing on eval with synthetic dataset, please have a look if you think it is not to hacky how it allows to load the synthetic val dataset

[Bug]: Different results in gsm8k when using chat

1

### Your current environment The output of `python collect_env.py` ```text Your output of `python collect_env.py` here ``` Collecting environment information... PyTorch version: 2.5.1+cu124 Is debug build: False CUDA used to...

@rom1504 this seems to make tests pass