-
Similar to the problem the cuda dma branch fixed for cuda, we need a fix for slow copies on AMD GPUs.
-
### Describe the bug
````
# RUN: at line 4
/__w/llvm/llvm/toolchain/bin//clang++ -Werror -Xsycl-target-backend=amdgcn-amd-amdhsa --offload-arch=gfx1031 -fsycl -fsycl-targets=amdgcn-amd-amdhsa /…
-
Compared to the hip-devel RPM packages (e.g. https://rpmfind.net/linux/RPM/fedora/updates/39/x86_64/Packages/h/hip-devel-5.7.1-1.fc39.noarch.html) the header files for the NVidia back-end are currentl…
-
- virtualization of different hardware driver (CUDA, HIP) where core targeting happens
- native drivers when found losing virtual overhead
- eventual singularity container based environments for HPC…
-
https://github.com/ROCm-Developer-Tools/HIP
HIP allows using AMD's gpu.
rocm/pytorch of Docker image provides libtorch_hip.so for HIP.
We can get the file by following scripts.
```
# pull doc…
-
### Checklist
- [X] The issue has not been resolved by following the [troubleshooting guide](https://github.com/lllyasviel/Fooocus/blob/main/troubleshoot.md)
- [X] The issue exists on a clean inst…
-
I'm trying to compile this for an AMD 6900XT. On an AMD CPU, arch linux. Pytorch works properly with ROCM.
Here are some of the link issues I'm getting:
```
/usr/bin/ld: CMakeFiles/ctranslate2.d…
-
Suggested by Johannes:
Ideally it should be `const` (The CUDA counterpart uses `const`), but the source in HIP [API](https://dec4.readthedocs.io/en/latest/ROCm_API_References/HIP_API/Memory-Manage…
KaoCC updated
4 weeks ago
-
定格を調査するにあたりG1のどの関節のトルクと角速度のパラメータを変更し調査するのかを決める。
Isaac Lab内の[unitree.py](https://github.com/unitreerobotics/unitree_rl_gym/blob/main/resources/robots/g1/urdf/g1.urdf)では一律effort_limit=300,velocity_limi…
-
Run into this error when in CMake Debug configuration. In Release configuration, fewer files fail, the ones that do fail, fail with 24 not 56.
Attached are the sources files
```
llvm-spirv -…