Authoritative Scope: package_scoped |
Distinct Packages: 10 |
Repository Package Instances: 10
euse -i rocmeuse -E rocmeuse -D rocm| Package | Source Kind | Source File | Description |
|---|---|---|---|
| dev-python/tilelang | metadata.xml |
dev-python/tilelang/metadata.xml |
Build the AMD ROCm backend instead of the CUDA backend |
| dev-python/vllm | metadata.xml |
dev-python/vllm/metadata.xml |
Build for AMD ROCm inference (VLLM_TARGET_DEVICE=rocm); pull HIP libs + torch{audio,vision} |
| media-gfx/comfyui | metadata.xml |
media-gfx/comfyui/metadata.xml |
Build the torch backend (<pkg>sci-ml/caffe2</pkg>) with AMD ROCm/HIP acceleration; installs the non-CUDA wheels of <pkg>dev-python/comfy-aimdo-bin</pkg> and <pkg>dev-python/comfy-kitchen-bin</pkg> and pulls <pkg>dev-python/triton-bin</pkg>, whose Triton backend runs comfy_kitchen.apply_rope on AMD. Mutually exclusive with cuda |
| sci-misc/llama-cpp | metadata.xml |
sci-misc/llama-cpp/metadata.xml |
Build a HIP (ROCm) backend |
| sci-misc/stable-diffusion-cpp | metadata.xml |
sci-misc/stable-diffusion-cpp/metadata.xml |
Build a HIP (ROCm) backend |
| sci-ml/bitsandbytes | metadata.xml |
sci-ml/bitsandbytes/metadata.xml |
Build the HIP/ROCm GPU backend (COMPUTE_BACKEND=hip) instead of the CPU backend |
| sci-ml/caffe2 | metadata.xml |
sci-ml/caffe2/metadata.xml |
Enable ROCm gpu computing support |
| sci-ml/ollama | metadata.xml |
sci-ml/ollama/metadata.xml |
Build the ROCm/HIP (rocm_v7_2) llama-server GPU backend |
| sci-ml/pytorch | metadata.xml |
sci-ml/pytorch/metadata.xml |
Enable ROCm gpu computing support |
| sci-ml/torchvision | metadata.xml |
sci-ml/torchvision/metadata.xml |
Enable ROCm gpu computing support |
| Package | Observations | Versions | Local Description (use.local.desc) | Metadata Description (metadata.xml) |
|---|---|---|---|---|
| dev-python/tilelang | IUSE + REQUIRED_USE | 0.1.10, 0.1.12, 0.1.13, 0.1.14, 0.1.15 | - | Build the AMD ROCm backend instead of the CUDA backend |
| dev-python/vllm | IUSE + REQUIRED_USE | 0.32.0, 0.33.0, 0.34.0 | - | Build for AMD ROCm inference (VLLM_TARGET_DEVICE=rocm); pull HIP libs + torch{audio,vision} |
| media-gfx/comfyui | IUSE + REQUIRED_USE | 0.45.6, 0.46.0, 0.46.1 | - | Build the torch backend (<pkg>sci-ml/caffe2</pkg>) with AMD ROCm/HIP acceleration; installs the non-CUDA wheels of <pkg>dev-python/comfy-aimdo-bin</pkg> and <pkg>dev-python/comfy-kitchen-bin</pkg> and pulls <pkg>dev-python/triton-bin</pkg>, whose Triton backend runs comfy_kitchen.apply_rope on AMD. Mutually exclusive with cuda |
| sci-misc/llama-cpp | IUSE + REQUIRED_USE | 0.8.0, 0.9.0, 1.0.0, 1.0.0_p12503, 1.0.0_p12532, 1.0.0_p12559, 9999 | - | Build a HIP (ROCm) backend |
| sci-misc/stable-diffusion-cpp | IUSE + REQUIRED_USE | 0_pre1091, 0_pre1104, 0_pre1106, 9999 | - | Build a HIP (ROCm) backend |
| sci-ml/bitsandbytes | IUSE + REQUIRED_USE | 0.50.0, 0.50.1, 0.50.2 | - | Build the HIP/ROCm GPU backend (COMPUTE_BACKEND=hip) instead of the CPU backend |
| sci-ml/caffe2 | IUSE + REQUIRED_USE | 2.13.0, 2.14.0, 2.14.1, 2.15.0 | - | Enable ROCm gpu computing support |
| sci-ml/ollama | IUSE | 0.40.13, 0.40.15, 0.40.16 | - | Build the ROCm/HIP (rocm_v7_2) llama-server GPU backend |
| sci-ml/pytorch | IUSE + REQUIRED_USE | 2.13.0, 2.14.0, 2.14.1, 2.15.0 | - | Enable ROCm gpu computing support |
| sci-ml/torchvision | IUSE + REQUIRED_USE | 0.28.0, 0.29.0, 0.29.2 | - | Enable ROCm gpu computing support |