onnxruntime/orttraining/tools/amdgpu
Weixing Zhang fc614ad050 revert the code change which was based on b4869926
The change b4869926 which was to remove per-thread allocator would cause seg fault for
distributed training.

In addition, add dockerfile for ROCm3.9
2020-11-15 00:24:32 -08:00
..
model
script Add GPU kernels for ROCm EP (#5655) 2020-11-06 16:11:06 -08:00
Dockerfile.rocm3.7 revert the code change which was based on b4869926 2020-11-15 00:24:32 -08:00
Dockerfile.rocm3.7.pytorch revert the code change which was based on b4869926 2020-11-15 00:24:32 -08:00
Dockerfile.rocm3.8.pytorch revert the code change which was based on b4869926 2020-11-15 00:24:32 -08:00
Dockerfile.rocm3.9 revert the code change which was based on b4869926 2020-11-15 00:24:32 -08:00
Dockerfile.rocm3.9.pytorch revert the code change which was based on b4869926 2020-11-15 00:24:32 -08:00