Alazar

@VictorSanh just an fyi - I am hoping this model is TF portable..

Thank you @Rocketknight1 for the detailed guide. I have 1,2 & 3 done already, i just updated the PR. I will continue to work on the rest. Regarding the GPT-4...

Thank you @Rocketknight1, yes that makes sense, taking a closer look, idefics is 2 pre-trained models combined together, so vision.py is a core part. I will take a look at...

Hello @Rocketknight1 - Sorry I went MIA, I have been occupied with my 9-5. I just made some progress. I have the tests running, I am running into a weird...

Thank you @Rocketknight1!

Hi @Rocketknight1, I have a few questions: 1. For processing_idefics.py, how do you suggest I handle both pytorch and tf? right now I just have it hacked in my view...

Thank you @Rocketknight1 for the detailed explanation and help! Moving on to other errors..

Sorry to bother you again @Rocketknight1, Do i need to worry about gradient checkpointing in the TF training tests? I am asking because for TFIdeficsForVisionText2Text there is `test_training` and `test_training_gradient_checkpointing`...

Hi @Rocketknight1, how do i run the integration tests? For example `IdeficsModelIntegrationTest`. `test_training` for tf is failing due to `model.train()`. i see it is defined in Trainer class in files...

Thanks @Rocketknight1. I have a few follow up questions. There are some test failures I don't understand. 1. For `image_processing_idefics.py`, I made changes in [this commit ](https://github.com/huggingface/transformers/pull/26870/commits/0a9fc014d431be78a5c4c2ab11e96e2ff41c47bf) to pass `return_tensors`...