search

hrzhou2 / AdaptConv-master

Adaptive Graph Convolution for Point Cloud Analysis
110 stars 12 forks source link

Error in training of part segmentation #6

Open rabbiahassan opened 2 years ago

rabbiahassan commented 2 years ago

Hello ! Thanks for the interesting work. When I put the part segmentation model on training,it just stucks here and doesn't proceed further,it doesn't show any error either. Please tell me what should I do? Thanks for your time. image

hrzhou2 commented 2 years ago

You should check first if the gpu was working. Of course it takes a few time to process one epoch.

rabbiahassan commented 2 years ago

GPU is working absolutely fine,It stucks there for more than a day and din't move further an inch.

hrzhou2 commented 2 years ago

It looks still fine for me here.

I think you can check if the code is working as expected within one epoch, by specifying a record interval. For example, set --interval 1 in training. This is expected to give some outputs very quickly.

rabbiahassan commented 2 years ago

Thanks a lot for your reply. Even for one epoch it shows the same behaviour i.e. it gets stuck there. image

hrzhou2 commented 2 years ago

Some possible reasons that have came out of my mind:

  1. Check first if the dataset is correct. And you could use python ShapeNetPart.py to see if it is working fine.
  2. Check the gpu usage. It should be occupied (not continuously due to data loading).
  3. Can you debug further to see it stops in which step. Since it happens within an epoch (only a few lines).
rabbiahassan commented 2 years ago

Thanks for reply,I will check all of these things.

UestcJay commented 2 years ago

Thanks for reply,I will check all of these things.

hi, the same error occurs, did you solve it?