Problems with training
MalatestaDavide opened this issue · comments
After downloading the videos we use the command "! Python -m backend.parser train --model / models" to train. The 1/5000 epoch starts giving us all the IDs of the videos on which to train but then it does not continue and goes in loop. I am attaching a photo to better clarify the doubt.
epoch starts giving us all the IDs of the videos on which to train but then it does not continue and goes in loop
@MalatestaDavide What you mean by "does not continue" / "goes in loop"? Does the program feel stuck after Epoch 1/1000 (and for how long) or what section is stuck in loop?
Can you please elaborate on what's happening a bit more? Possibly sharing the full log dump.
Basically, we tried to execute the network on Google Colab. However, when we try to execute the command "!python -m backend.parser train --model /models" the program feels stuck after epoch 1/5000. We let the program run for 6 hours but the training doesn't go on.
The only thing we changed on the code was the downloadVideo function on the videohandler.py file, because the website used for downloading videos from youtube was obsolete.
