search

MCG-NJU / VideoMAE

[NeurIPS 2022 Spotlight] VideoMAE: Masked Autoencoders are Data-Efficient Learners for Self-Supervised Video Pre-Training
https://arxiv.org/abs/2203.12602
Other
1.32k stars 131 forks source link

KeyError: 'model' #119

Open fengjingchehu opened 5 months ago

fengjingchehu commented 5 months ago

VideoMAE-main$ bash vis.sh Namespace(decoder_depth=4, device='cuda:0', drop_path=0.0, imagenet_default_mean_and_std=True, img_path='./input/test1.mp4', input_size=224, mask_ratio=0.9, mask_type='tube', model='pretrain_videomae_base_patch16_224', model_path='./ckp/ucf101.pth', num_frames=16, sampling_rate=4, save_path='.demo')

Creating model: pretrain_videomae_base_patch16_224 Patch size = (16, 16) Traceback (most recent call last): File "run_videomae_vis.py", line 185, in main(opts) File "run_videomae_vis.py", line 102, in main model.load_state_dict(checkpoint['model']) KeyError: 'model'

i wonder how can i solve this question ?

fusu80 commented 4 months ago

你好，请问你解决这个问题了吗，我今天在运行.sh的时候也发生了这样的错误。期待你的回复。
Hello, did you fix this problem, I also got this error when I was running the .sh today. Looking forward to hearing from you.

censhallwe commented 1 month ago

你是从MODEL_ZOO上下载的权重吗？那里存的时候key用的是'module'，你可以试着把key改一下