Skip to content

v0.30.8

Latest

Choose a tag to compare

@likelovewant likelovewant released this 13 Jun 09:46
· 218 commits to main since this release
4698f6d

Quick Setup for Ollama on AMD (ROCm 7.1.1)

⚠️ Demo Release Notice
The releases here are for demo purposes and may not be up to date. For the latest version, please build it yourself or refer to the detailed wiki guide.

Tip

Ollama Official Support AMD in Vulkan * Ollama now natively supports AMD GPUs via Vulkan for most architectures. No need for the complex steps in this repo – just use the official installer.

Downloads

Supported GPUs
gfx1010; gfx1011; gfx1012; gfx1030; gfx1031; gfx1032; gfx1034; gfx1035; gfx1036; gfx1100; gfx1101; gfx1102; gfx1103; gfx1150; gfx1151; gfx1152; gfx1153; gfx1200; gfx1201


Setup Steps

  1. Install Ollama

    • Download and run OllamaSetup.exe (from this repo or the official site).
    • If using the official installer, replace the ggml-hip files in C:\Users\<username>\AppData\Local\Programs\Ollama\lib\ollama\rocm_v7_1 with those from ollama-windows-amd64-rocm.zip (skip if you used this repo's installer).
  2. Replace ROCm libraries

    • Go to C:\Users\<username>\AppData\Local\Programs\Ollama\lib\ollama\rocm_v7_1.
    • Delete the rocblas/library folder.
    • Copy the correct ROCm libs for your GPU from the ROCmLibs v7.1.1 release into the same location.
  3. Restart your PC (if you see internal errors), then launch Ollama.


For integrated GPUs (e.g., gfx1103, gfx1151…)

Set the environment variable OLLAMA_IGPU_ENABLE=1.
Learn how to set environment variables on Windows. Otherwise, your GPU may not be detected.