pruning-distillation guidance or docs for Multimodal model

[xduzhangjiayu]

Hi，
Is this project has some guidance or docs about how to pruning-distillation a multimodal model like qwen2.5vl-7B Qwen2.5-Omni-7B? Currently I only find the docs for Llama.
Looking forward to the reply, thanks!

[sharathts]

Hi @xduzhangjiayu - we currently support only pruning and distillation of LLMs. The same can be applied to compressing the LLM part of VLMs as they tend to be the bottleneck.

[xduzhangjiayu]

@sharathts
Hi,
Thanks for the reply, one more question is, how can I transfer the multimodal model (like qwen2vl) type from huggingface model to .nemo format, is there any script in this repo? Thanks!