LLaVA-NeXT
LLaVA-NeXT copied to clipboard
How can tensor parallelism be implemented in the LLaVA framework when pretraining a 7B model on A40 GPUs?