Clément Pinard

Thanks for the info, the build/sdist distinction makes sense. Someone is apparently working on a patch for this package in fiftyone, so we're saved. I suppose I can close this...

Hi, just did a chekc on the pretrained networks. I renamed them to make it clear that you shouldn't try to unter them, and the flown from caffe is now...

Hello, See similar issue here : https://github.com/alexandrosstergiou/SoftPool/issues/6 Issue is that your input tensor is not contiguous, probably because you used some stride (something like `my_tensor[..., ::2]`) or some transpose without...

Hard to say without the code you used, but my guess would be that optical flow is normalized so that the ighest values is fully colored. As such, when there...

div_flow is purely conventional, it was deemed to ease training stability that the network had to output small values (hence the division of target optical flow by 20 by default)...

Hi, is this PR considered ? Helped me a lot for auto completion, and would be nice to have it included in the pip package (which could use an update...

Hi thanks @Chris-hughes10 don't hesistate to ping me if some changes are needed

/quack review

Hi, I do believe it's solved with the commits I just made, can you check it out ? Moreover, I added the feature of saving the generated the split in...

Actually, it should be 0, `n_iter` is only here for tensorflow purpose. I am surprised this does not make the code crash, since the variable `start_epoch` is not defined :thinking:...