sclarkson

Results 4 comments of sclarkson

Infinite recursion error when building cuda targets

@chsigg I also encounter this problem. My CUDA toolkit install at /usr/local/cuda was done with the runfile, not a package manager. ``` $ dpkg -S /usr/local/cuda dpkg-query: no path found...

Fail to compile with gloo: multiple definition of `gloo::rendezvous::Store::kDefaultTimeout

This change worked for me ``` --- a/torch/csrc/distributed/c10d/ProcessGroupGloo.hpp +++ b/torch/csrc/distributed/c10d/ProcessGroupGloo.hpp @@ -125,7 +125,7 @@ } void wait(const std::vector& keys) override { - store_->wait(keys, Store::kDefaultTimeout); + store_->wait(keys, ::c10d::Store::kDefaultTimeout); } void wait(...

Build TensorFlow for raspberry pi failed by ): relocation truncated to fit: R_AARCH64_CALL26 against symbol

The solution seems to be using clang instead of GCC.

ARM build failed with recent XNNPACK update: third_party/XNNPACK/src/reference/unary-elementwise.cc:125:14: error: invalid ‘static_cast’ from type ‘xnn_bfloat16’ to type ‘_Float16’

Also encountering this problem on linux ARM64, with the 2.6.0 release