custom-op icon indicating copy to clipboard operation
custom-op copied to clipboard

Guide for building custom op for TensorFlow

Results 44 custom-op issues
Sort by recently updated
recently updated
newest added

With TensorFlow 2.5, there have been some code changes which require CUDA 11. You can read more about the issues when trying to build TF 2.5 against CUDA 10 here:...

It'd be helpful to add an example use of ResourceMgr for third party C++ objects used. I put together a minimal example for my own purposes https://gist.github.com/maedoc/37202e3e1b3dbff1b77dc6106e11917a but it'd be...

I am using a Lambda workstation which has tf 2.4.1 installed and cuda 11.2. What docker container should I use? I only see one for 2.4.0, if I build against...

I want to create a Windows wheel, what should I do?

Setup: - running bazel 3.1.0 on powershell (msys2 in path for bash) (powershell is recommended [here](https://docs.bazel.build/versions/master/windows.html#running-bazel-msys2-shell-vs-command-prompt-vs-powershell)) - cuda 10.1 with cudNN 7.6 files copied into the respective folders (bin, lib,...

Fixes issue described in #87. Putting together the copy command for tf headers in tf_configure.bzl using " && " causes a stack overflow in msys2 bash.exe on windows. Executing each...

I execute ``./configure.sh` with "Y, Y, n" and then fail to run `bazel build build_pip_pkg` after ` using the image `tensorflow/tensorflow:custom-op-ubuntu16`. ``` ERROR: /temp/custom-op/tensorflow_zero_out/BUILD:10:1: //tensorflow_zero_out:python/ops/_zero_out_ops.so depends on @local_config_tf//:libtensorflow_framework in repository...

Can we add in the example something related to TPU. There was a FAQ about creating custom ops for TPU https://cloud.google.com/tpu/docs/faq

Hi Team, I'm trying to build a custom op for beam search and have a question on specifying the input type. My question is: can we take RNNCell as input?...

Excuse me ,I have downloaded the related code packages within pyro-nn, pyro-nn-layers and custom-op links from GitHub, configured CUDA and CUDNn on Windows or Ubuntu, and installed Tensorflow-GPU through PyCharm....