search

TinyLLaVA / TinyLLaVA_Factory

A Framework of Small-scale Large Multimodal Models
https://arxiv.org/abs/2402.14289
Apache License 2.0
582 stars 53 forks source link

请问本代码能否在单卡4090上做微调呢？最少显存占用多少 #46

Open huiby23 opened 5 months ago

baichuanzhou commented 5 months ago

用LoRA可以微调TinyLLaVA-3.1B，但需要把gradient accumulation开大些，详情请参考这里：https://github.com/DLCV-BUAA/TinyLLaVABench/blob/main/scripts/tiny_llava/finetune/finetune_lora.sh ，文档描述在这里：https://github.com/DLCV-BUAA/TinyLLaVABench/blob/main/docs/CUTOM_FINETUNE.md