alexxu-amd comments

Results 11 comments of


                                            alexxu-amd

[Issue]: Stable diffusion, Pytorch conv2d breaks in rocm 6.0

Yea, I am able to reproduce this error using the installation step from https://github.com/AUTOMATIC1111/stable-diffusion-webui/wiki/Install-and-Run-on-AMD-GPUs#user-content-install-on-amd-and-arch-linux. Can you guys try reinstalling `torch` and `torchvision` using `pip3 install --pre torch torchvision torchaudio --index-url...

Blender Cycles engine causes "Memory access fault by GPU node-1"

Hi all, can you check if this issue can still be reproduced with the latest ROCm 6.1?

Blender Cycles engine causes "Memory access fault by GPU node-1"

@GZGavinZhao Thanks for the detailed description. We will look into it

[Issue]: HIP error on rocm/pytorch docker (gfx 1100) for float tensors

In the latest rocm 6.1 docker image, this error can be solved by upgrading torch 2.1 to 2.3. Can you check if it works for you?@yash-s20

Obtain the physical address when allocating via `hipMalloc`

Hi @xuantengh, Exposing the physical address can potentially lead to security concerns. So unfortunately, it is not a supported feature.

[Issue]: ROCm/xformer won't compile due to missing ck/tensor/tensor_view.hpp

@lhl @hackey, Currently, xformers on ROCm only works with MI200/MI300. So unfortunately, 7900 XTX won't be able to run it at the moment. Apologize for the inconvenience.

[Issue]: Can't compile for debugging with -O0

Hi @cameronshinn , I was able to compile `example_gemm_xdl` with the debugging flag `-D CMAKE_CXX_FLAGS_DEBUG="-g -O0"` on ROCm 6.2. I'm using Clang 18.0.0 and cmake 3.22.1. Could you please try...

[Issue]: Can't compile for debugging with -O0

Closing the issue now as `example_gemm_xdl` does seem to compile with the debugging flag on ROCm 6.2. @cameronshinn Feel free to reopen the issue if you are still not able...

[Issue]: ld.lld: build fails with `version script assignment of 'ROCR_1' to symbol 'hsa_tools_scratch_event_alloc_start' failed: symbol not defined`

Hi @AngryLoki , Thanks for bringing this to our attention. We have confirmed the above symbols are unused and deleted them, expect the change to come in future updates.

[Issue]: Soft lockup when having high vram usage

Hi @hartmark , Thank you for your patience. For the past few days, we've been attempting to reproduce this issue with various configurations. Just an update that we are able...