On reproducing LLaVA alignment experiments.

[bhattg]

Hi! I've a couple of questions on the LLaVa alignment:

• Which LLaVA model was used for the experiment? Was it the 7B one or 13B one?
• What precision was used w/ LoRA based results? Was that bf16 of full fp32?

[kvablack]

I used the 13B variant of LLaVA, and all experiments were run with fp16 precision.

[bhattg]

Thanks, closing the issue now. :-)