Tony Wu
If you look at the issue I mentioned, it seems that you stated the opposite. Here's what you wrote for context: > I'll update the documentation to make it less...
Hi @venki-lfc, thanks for the kind message! 👋🏼 Could you clarify what exactly did you put in your `./models--vidore--colqwen2-base/snapshots/c722b912b50b14e404b91679db710fa2e1c6a762` folder please? Is `c722b912b50b14e404b91679db710fa2e1c6a762` the renamed folder obtained after running `git...
Loading from [`vidore/colqwen2-base`](https://huggingface.co/vidore/colqwen2-base) gives you the untrained version of ColQwen2. The reason you get two folders is because [`vidore/colpali-v1.2`](https://huggingface.co/vidore/colpali-v1.2) is a LoRA adapter trained on top of `vidore/colqwen2-base`. Thus, in...
Hi @AlexAndrei98! You're right, the rearranging will indeed differ. From a quick look at the code, [this code](https://github.com/illuin-tech/vidore-benchmark/blob/c217f5285c3edc7efedc8221db00e7160537c70b/src/vidore_benchmark/interpretability/gen_similarity_maps.py#L75C1-L78C44) will have to be modified: ```python # Rearrange the output image tensor...
@AlexAndrei98 FYI I've just merged https://github.com/illuin-tech/colpali/pull/96 to add similarity map support for ColQwen2 👍🏼
@ManuelFay I managed to make the interpretability work with both ColPali and ColQwen while keeping the API as simple as possible. Wdyt?
> @tonywu71 any reason this is not merged ? Yup, was running a few extra tests yesterday. Gonna merge it tonight or tomorrow morning!
Hi @ashokrajab, This feature will be supported thanks to https://github.com/illuin-tech/vidore-benchmark/pull/68. Sorry for the delay, we are going a big repository restructuration at the moment! I'll link your issue to my...
@QuentinJGMace Could you look this PR up when you'll have some time please? It'd be great to be able to evaluate ColQwen2.5 straight from `vidore-benchmark` 🙏🏼
Recap from our conversation 👋🏼 Let's: - remove the legacy evaluation code - add optional training arg `run_vidore_evalutor`: if `False`, do not add the custom callback - add optional training...