search

facebookresearch / jepa

PyTorch code and models for V-JEPA self-supervised learning from video.
Other
2.63k stars 251 forks source link

Evaluation / Testing data #33

Open aakgun opened 7 months ago

aakgun commented 7 months ago

I want to evaluate a basic imagenet data, although i have downloaded config files and pretrained models from the github site provided, i get following error. I tried to use ImageNet-1K data, i could not find the imagenet_full_size/061417/ data. Can you please provide more infrmation about how to test / evaluate data using welknown imagenet dataset. Or can your provide any support file / info for imagenet_full_size/061417/

File "/home/ayhan/jepa/evals/image_classification_frozen/eval.py", line 363, in load_pretrained pretrained_dict = checkpoint['encoder'] KeyError: 'encoder'

Thank you.

HaoranZhuExplorer commented 6 months ago

Same problem here, what does "image_folder: imagenet_full_size/061417/" mean?

HaoranZhuExplorer commented 6 months ago

I found the solution. Please see how they manage their data in readme file: they create a directory named "061417" and put imagenet data set into "061417" folder. So the solution is simply to mange the data in the same way as they do, although I don't know what's the connection between "061417" and imagenet...