search

42Shawn / LLaVA-PruMerge

LLaVA-PruMerge: Adaptive Token Reduction for Efficient Large Multimodal Models
Apache License 2.0
85 stars 4 forks source link

some wrong about the checkpoints you provided #2

Open XpracticeYSKM opened 4 months ago

XpracticeYSKM commented 4 months ago

When I try torch.load("llava-v1.5-7b-lora-prunemerg/adapter_model.bin")，I meet RuntimeError: PytorchStreamReader failed reading zip archive: failed finding central directory.

I am sure that I download the checkpoint correctly. And I want to know where the checkpoint of PruMerge is save?

niiickZ commented 4 months ago

@XpracticeYSKM I faced same issue. Downloading the checkpoint from liuhaotian/llava-v1.5-7b-lora works for me.

42Shawn commented 4 months ago

Please put the lora model in your local dir, and change this line of code in the eval scripts
--model-path /data/shangyuzhang/LLaVA/checkpoints/llava-v1.5-7b-lora-prunemerge \

change /data/shangyuzhang/LLaVA/checkpoints/llava-v1.5-7b-lora-prunemerge to your downloaded dir.

cnxupupup commented 4 months ago

@42Shawn I faced same issue, I am sure that I download the checkpoint correctly.

but PytorchStreamReader failed reading zip archive: failed finding central directory.

Yogurt2019 commented 4 months ago

Same problem here. I downloaded the lora weights from README.md at the google drive, unzip it. The prumerge code gives the RuntimeError: PytorchStreamReader failed reading zip archive: failed finding central directory.

And I found that if I run torch.load("adapter_model.bin"), it gives the following error: RuntimeError: PytorchStreamReader failed reading zip archive: failed finding central directory.

Seems the adapter_model.bin you provided in the google drive is broken.

42Shawn commented 4 months ago

I have uploaded our Lora checkpoint through huggingface (https://huggingface.co/yuzhang), please check it out.