Completely incapable of getting the installation requirements working

[kwende]

After following the steps for Ubuntu 16, verbatim, I consistently get this:

from caffe2.python import core
WARNING:root:This caffe2 python run does not have GPU support. Will run in CPU only mode.
CRITICAL:root:Cannot load caffe2.python. Error: No module named caffe2_pybind11_state

I would like some input.

Edit: Basically, ignore these instructions if you have trouble. Your source should be:

https://caffe2.ai/docs/getting-started.html?platform=ubuntu&configuration=compile

[daniel-j-h]

I had similar problems that's why I packages up the installation into a self-contained and reproducible Docker environment; see this pull request #94

[kwende]

Hah, @daniel-j-h that's good to hear. I didn't see that PR. I too have a docker container now (almost, anyway). Yours looks to be downloading their pre-existing models and allowing people to use them for inference, right? I'm trying to piece together one for training. Does yours do that? Or do you have one that does that? Maybe I missed it....

Regardless, it's amazing how messed up the instructions on this repo are.

Another question: downloading the data sets to train -- have you had any success? They link to to the ActivityNet from here, but running the scripts from the ActivityNet repo fail because the wonky youtube-dl program is failing to download stuff from YouTube. For example, following the instructions here result in nothing more than a complete list of "not found" from youtube-dl being spat out into the download_report.json file.

It'd be nice to at least get the thing operational and training something, even a test data set.

Edit: the problem was the standard Ubuntu youtube-dl image (apt-get install youtube-dl) is out of date. I downloaded the latest and it's downloading stuff from YouTube.

Also: Who stores stuff on YouTube?