Error while giving the command for training about unsupervised-videos

[jbboin]

It seems that you didn't compile the cudamat library, or the process didn't complete correctly.

Otherwise I agree with the second question: I would be really interested in getting access to some pretrained models used in the paper as well. In particular, any of the following would be super useful for me:

• composite LSTM model trained on the 300 hours of YouTube data
• composite LSTM classifier trained on the 300 hours of YouTube data and finetuned on UCF-101 RGB
• baseline LSTM classifier trained on UCF-101 RGB

Is there any possibility that these could be made available in the near future?

[mansimov]

Hey

We have been working on this project a bit more than 3 years ago and since then all of us moved to other projects and endeavors. We should have released trained models back then and I am afraid it is too late to find them

I will try finding something in very old repos but I can't promise anything.

Thanks for your understanding.

[jbboin]

Hi,

I completely understand. Thanks a lot for looking into it, anything would be helpful but I understand if it's been too long to find these models again.

[Ty-Won]

Hi,

Is it possible to complete the guide you wrote in the read-me without this file?

Thanks!

[DavidSUN2]

It seems that you didn't compile the cudamat library, or the process didn't complete correctly.

Hi Guys. I'm sorry for asking this simple questions: how to compile the cudamat library?
I have searched https://github.com/cudamat/cudamat and follow the install instruction, but it doesn't work. When I run on Bouncing (Moving) MNIST dataset, it always says that the specified module could not be found.
I have tried to use the similar way to the above website to install, but the unsupervised-videos-master folder doesn't have setup.py.
So, what should I do? How to compile this cudamat library? Thanks!

[DavidSUN2]

I use Ubuntu OS and make Makefile successfully.