Closed danpovey closed 6 years ago
Here https://github.com/waldo-seg/waldo/blob/3566eaec5752796eb569e32d2f95b9e5effe8165/egs/dsb2018/v1/local/train.py#L231 I think autgrad.Variable() means we will need the derivative w.r.t. that thing, which we don't for the image. It would be slightly more efficient not to do this. I don't know how you tell pytorch you don't need the derivative though. Maybe just use the original numpy array?
Here https://github.com/waldo-seg/waldo/blob/3566eaec5752796eb569e32d2f95b9e5effe8165/egs/dsb2018/v1/local/train.py#L231 I think autgrad.Variable() means we will need the derivative w.r.t. that thing, which we don't for the image. It would be slightly more efficient not to do this. I don't know how you tell pytorch you don't need the derivative though. Maybe just use the original numpy array?