Sergei Belousov
Sergei Belousov
Omg, I investigated this issue. So, the problem is related to the input shape of the tensor. [Here](https://github.com/facebookincubator/AITemplate/blob/main/python/aitemplate/compiler/ops/conv/conv_common.py#L44) we try to find the cuda kernel for Conv2d operation that corresponds...
- As I understood, nn.Conv2d doesn't work for small sized odd number of channels - It can be fixed using nn.Conv2dBiasFewChannels layer But there isn't any description for this behaviour...
> ```python > import logging > from aitemplate.compiler import ops > from aitemplate.frontend import nn > # AIT utils > from aitemplate.compiler import compile_model > from aitemplate.frontend import Tensor >...
@spinoza1791 we use gdrive to host our checkpoints, and we support autodownload of our checkpoints. But we don't support .ptl format
@spinoza1791 we use gdrive to host our checkpoints, and we support autodownload of our checkpoints. But we don't support .ptl format
@rotem154154 we can't support such Style-space directly, but i believe that it's possible
Hi, Do you use original code or modified version? What camera do you use? This project works only with MS Kinect camera (using OpenNI wrapper). Regards, BeS
Доброго времени суток. Увы, но картинки от той статьи были утеряны.
Не знаю. Вообще крутые библиотеки обычно используют для детектирования рук и пальцев методы оптимизации, а не контурный анализ.
@vc-zju и что ты предлагаешь с этим делать?