Noah Kasmanoff

Hey, posting in here since I'm interested in multi-modal. I'm currently trying to convert [bakLlava](https://huggingface.co/llava-hf/bakLlava-v1-hf). The model doesn't matter too much to me, but this one worked out of the...

@gboduljak Thank you this is a lot of great info! Will try to catch myself up and help :-)

@gboduljak No problem, but after looking it over, not sure I can be extraordinarily helpful beyond some simpler tasks. This is a lot lower level coding than I'm used to...

@gboduljak I submitted a PR to your existing PR, which creates a local implementation of the CLIPImageProcessor. https://github.com/gboduljak/mlx-examples/pull/1 This should eliminate the dependency on transformers, aside from using it for...

Not my intent to re-open this PR, in here to say I'm currently testing on a raspberry pi. At the moment the "bottleneck" is that time to process those 729...

Hey great initiative! Excited to try out these new and improved LlaVA models. Agreed that using the standard transformers format makes it easier. I've been thinking a bit about the...

> I’m unfortunately busy and won’t be able to take a closer look again until later today, but re: fine-tuning demos, this video might be helpful: https://www.youtube.com/watch?v=eIziN2QUt8U It did, in...