use/neural-net

Issues

• Disable these and see how it affects performance

  #3 opened 3 years ago by use
  0

• Try preallocating this with cudamalloc to see how it affects performance

  #4 opened 3 years ago by use
  0

• Try this out to see how it affects performance.

  #5 opened 3 years ago by use
  0

• Shouldn't this be total number of threads (i.e. batch size) rather than threads per block?

  #2 opened 3 years ago by use
  1

• Double check math. Shouldn’t this incorporate layer size 0?

  #1 opened 3 years ago by use
  1