search

OpenGVLab / UniFormerV2

[ICCV2023] UniFormerV2: Spatiotemporal Learning by Arming Image ViTs with Video UniFormer
https://arxiv.org/abs/2211.09552
Apache License 2.0
294 stars 19 forks source link

加载clip的预训练模型 #2

Closed wntg closed 2 years ago

wntg commented 2 years ago

加载预训练模型的时候，clip的模型是pt后缀，为什么你用的是pth，想问下在哪下载的？ 我修改了后缀名，加载clip的pt文件但是后面还是会报错。 _MODELS = { "ViT-B/16": os.path.join(MODEL_PATH, "ViT-B-16.pth"), "ViT-L/14": os.path.join(MODEL_PATH, "vit_l14.pth"), "ViT-L/14_336": os.path.join(MODEL_PATH, "vit_l14_336.pth"), }

wntg commented 2 years ago

就是去哪里下载clip的图像encoder权重

Andy1621 commented 2 years ago

Please follow the step in INSTRUCTION. You have to use the code here to generate the weight.

image
Andy1621 commented 2 years ago

BTW, if you only want to use the pretrained weight of UniFormerV2, you can set pretrained=False in the code, thus it will skip loading CLIP pretrained weight.

wntg commented 2 years ago

Thanks for your response! Thanks for your work!

Andy1621 commented 2 years ago

Thanks for your try! As there is no more activity, I am closing the issue, don't hesitate to reopen it if necessary.

xwg0508 commented 6 months ago

请问在运行extract.ipynb文件时出现以下错误需要怎么解决呢

image

conda环境中clip已经安装了

image