Issue Encountered When Loading the Model: "pretrain_videomae_base_patch16_224"

[bbbdbbb]

Dear [Support Team],

I hope this message finds you well. First of all, I would like to express my appreciation for your work. I have encountered a problem when trying to reproduce your code and load a model using the following line: model = get_model(args).

The default model specified in args is pretrain_videomae_base_patch16_224. However, when running the code, I received an error stating that the model could not be found: RuntimeError: Unknown model (pretrain_videomae_base_patch16_224).

I am currently using timm version 0.4.12. When I attempt to search for the model using model_vit = timm.list_models('*videomae*'), it does not appear in the results.

I kindly request your assistance in resolving this issue. Any guidance or suggestions would be greatly appreciated.

Thank you very much for your attention and support.

[otmalom]

Hello, I have encountered the same issue as you. Indeed, the model pretrain_videomae_base_patch16_224 is not available in timm. However, I found it in the file modeling_pretrain.py. I'm curious to know if the author imported the model from there, and why the default function create_model from the timmlibrary is used in the file run_mae_pretraining.py.

[TuuSiwei]

Hello，do you solve the problem？

[TuuSiwei]

https://github.com/MCG-NJU/VideoMAE/issues/9

[bbbdbbb]

It has been quite some time, and I apologize for the delay in my response. I recently came across your question on GitHub, and after reviewing my previous work, I have found some information that might be helpful to you.

Based on my recollection, it seems that I made modifications to the code responsible for loading the model in the "run_mae_pretraining.py" file. Specifically, I commented out the line model = get_model(args) and instead used the following code:

# model = get_model(args)  
from modeling_pretrain import pretrain_videomae_large_patch16_224  
model = pretrain_videomae_large_patch16_224()  

These adjustments allowed me to successfully run the VideoMAE code, and I achieved promising experimental results. I hope this information proves useful to you in your own work.

[fusu80]

9

这个貌似解决不了，请问还有其他的办法吗？