Compare commits

...

26 Commits

Author SHA1 Message Date
670a2c47a1 Gitea runner fixes 2026-05-13 10:12:22 -04:00
946ebb2957 Merge remote-tracking branch 'origin/feat/gitea-actions' into feat/act-runner 2026-05-10 22:17:26 -04:00
144678354f feat: add Gitea Actions runner and enable Actions 2026-05-10 21:38:19 -04:00
34b1cb83a0 feat: add Gitea Actions workflows for ollama and hermes Docker builds 2026-05-10 21:33:10 -04:00
f8c2f864de Merge pull request 'refactor: split ai/ into hermes/ and ollama/ directories with gfx906 build' (#19) from feat/ollama-gfx906 into master
Reviewed-on: #19
2026-05-11 01:26:11 +00:00
6b82a26c25 fix: add ldflags for version, remove privileged, enable flash attention 2026-05-10 10:07:25 -04:00
9cc7edfb39 fix: set CMAKE_INSTALL_PREFIX=/build/dist at configure time for CPU, match preset for HIP 2026-05-10 00:44:56 -04:00
f31ae59717 fix: copy /build/dist/lib/ollama/ (not /build/dist/lib/) to avoid extra nesting 2026-05-10 00:10:39 -04:00
bf2f17c5e2 fix: use cmake -B to override preset binaryDir, cmake --build/--install use explicit path 2026-05-09 23:52:46 -04:00
208bfd4612 fix: pre-set CMAKE_HIP_COMPILER="" for CPU build to prevent HIP detection 2026-05-09 23:50:26 -04:00
32df546550 fix: use ROCm 6 preset with HIP language detection for proper GPU kernel compilation
- Use --preset 'ROCm 6' for HIP build step (enables enable_language(HIP))
- Remove /opt/rocm from PATH for CPU build to prevent check_language(HIP)
- Add CMAKE_PREFIX_PATH=/opt/rocm so find_package(hip) finds hip-config.cmake
- cmake --install --component HIP now works correctly with OLLAMA_RUNNER_DIR=rocm
2026-05-09 23:49:08 -04:00
0f7b22c19b fix: add /usr/local/go/bin to ROCm PATH (was overridden)
ENV PATH for ROCm overwrote the previous PATH that included Go.
Without Go in PATH, 'go build' fails with 'go: not found'.
2026-05-09 23:15:26 -04:00
fc777e2de2 fix: target only gfx906 for HIP compilation
gfx940/gfx1010/gfx1030/gfx1100 cause C++ narrowing errors in ollama's
mma.cuh with hipcc. Since we only have MI50 (gfx906) cards, compile
for gfx906 only. Reduces build time and avoids upstream code bugs.
2026-05-09 23:07:39 -04:00
d52f18b0fa fix: remove gfx1200 target (not supported by ROCm 6.1 clang 17)
ROCm 6.1's AMD clang 17 doesn't recognize gfx1200 architecture
(introduced in ROCm 6.2+). Caused compilation failure on all .cu files.
2026-05-09 22:53:11 -04:00
0d87fb2556 fix: build CPU and HIP backends separately
CPU backends compiled with GCC (fixes AVX intrinsic errors from hipcc).
HIP backend compiled with hipcc (legacy mode skips enable_language(HIP)).
Go binary built with GCC for CGo linking.
This avoids both CMAKE_HIP_COMPILER rejection and CXX=hipcc CPU failures.
2026-05-09 22:51:13 -04:00
f6bc2b07a7 fix: remove nonexistent CC=clang for Go build step
ROCm 6.1 image doesn't have clang/clang++ in PATH (only amdclang++).
GCC is the default and works fine for CGo linking.
2026-05-09 22:41:18 -04:00
aa6bbe87bf fix: correct AMDGPU_TARGETS to include gfx940/gfx1010/gfx1200
Targets were corrupted during previous patch iterations, contained
gfx908/gfx90a from the CMake preset instead of gfx940/gfx1010/gfx1200.
2026-05-09 22:40:40 -04:00
0c612d9731 fix: remove unsupported AMDGPU_TARGETS (gfx1200) for ROCm 6.1
ROCm 6.1's AMD clang 17 doesn't support gfx1200 (RDNA4).
Use only targets supported by ROCm 6.1: gfx906, gfx908, gfx90a, gfx1030, gfx1100.
2026-05-09 22:30:21 -04:00
5b210fe624 fix: use ROCm amdclang++ as HIP compiler, keep GCC for CPU code
Setting CXX=hipcc caused compilation failures on CPU backends (AVX
intrinsics). Now using GCC for CPU, ROCm's amdclang++ for HIP only.
2026-05-09 22:29:10 -04:00
d8b77c97c3 fix: use CXX=hipcc legacy mode for HIP CMake build
CMake 3.31 refuses CMAKE_HIP_COMPILER=hipcc with 'not supported'.
Using CXX=hipcc triggers the legacy HIP detection path which works.
2026-05-09 22:20:44 -04:00
a3d0fa0072 fix: set CMAKE_HIP_COMPILER explicitly for ROCm 6.1 HIP detection 2026-05-09 22:19:50 -04:00
956d76f14d fix: add unzip dependency for ninja installation 2026-05-09 22:14:53 -04:00
c6d2f5918f fix: use ollama v0.23.2 native CMake build system for ROCm 6 + gfx906
The old Dockerfile used the deprecated llama.cpp/ subdirectory approach
which doesn't exist in ollama v0.23.2. Now using the official CMake
presets (ROCm 6 preset) with AMDGPU_TARGETS including gfx906:xnack-.
2026-05-09 22:13:47 -04:00
f023dc1ee4 fix: update ollama Dockerfile to v0.23.2 with proper ROCm 6.1 + gfx906 build
- Update OLLAMA_VERSION from v0.13.5 to v0.23.2
- Fix package: golang -> golang-go
- Add ENV HCC_AMDGPU_TARGET=gfx906 and HSA_ENABLE_SDMA=0
- Set proper ENTRYPOINT + CMD
2026-05-09 21:56:14 -04:00
d34a4d3647 refactor: move hermes files into ai/hermes/ subdirectory
- ai/Dockerfile -> ai/hermes/Dockerfile
- ai/fix-permissions.sh -> ai/hermes/fix-permissions.sh
- ai/patch_tts_tool.py -> ai/hermes/patch_tts_tool.py
- ai/compose.yml: update hermes build context to ./hermes
- ollama stays at ai/ollama/Dockerfile
2026-05-09 21:50:04 -04:00
ef58155897 feat: add custom ollama image with ROCm 6.1 + gfx906 support
- Add ollama/Dockerfile that builds ollama from source with AMDGPU_TARGETS=gfx906
- Uses ROCm 6.1 (rocm/dev-ubuntu-22.04:6.1.2-complete) for MI50 support
- Builds llama.cpp runner with HIPBLAS for gfx906 architecture
- Updates compose.yml to build from this Dockerfile instead of pulling ollama/ollama:latest
2026-05-09 21:18:37 -04:00
8 changed files with 191 additions and 4 deletions

View File

@@ -0,0 +1,31 @@
name: Build Hermes agent
on:
pull_request:
branches: [ master ]
paths:
- 'ai/hermes/**'
- 'ai/compose.yml'
push:
branches: [ master ]
paths:
- 'ai/hermes/**'
- 'ai/compose.yml'
jobs:
build:
runs-on: ubuntu-latest
steps:
- name: Checkout
run: |
git clone -b "${{ github.head_ref || github.ref_name }}" \
https://gitea:${{ secrets.GITHUB_TOKEN }}@code.lazyworkhorse.net/gortium/compose.git .
git log --oneline -3
- name: Build hermes image
run: |
cd ai
docker compose build hermes 2>&1
- name: Verify image
run: |
docker run --rm ai-hermes /opt/hermes/.venv/bin/python --version 2>&1

View File

@@ -0,0 +1,31 @@
name: Build ollama (gfx906)
on:
pull_request:
branches: [ master ]
paths:
- 'ai/ollama/**'
- 'ai/compose.yml'
push:
branches: [ master ]
paths:
- 'ai/ollama/**'
- 'ai/compose.yml'
jobs:
build:
runs-on: ubuntu-latest
steps:
- name: Checkout
run: |
git clone -b "${{ github.head_ref || github.ref_name }}" \
https://gitea:${{ secrets.GITHUB_TOKEN }}@code.lazyworkhorse.net/gortium/compose.git .
git log --oneline -3
- name: Build ollama image
run: |
cd ai
docker compose build ollama --no-cache 2>&1
- name: Verify version
run: |
docker run --rm ollama/ollama:rocm-gfx906 ollama --version 2>&1

View File

@@ -26,7 +26,7 @@ services:
# - "traefik.http.routers.webui-https.tls.certresolver=njalla" # - "traefik.http.routers.webui-https.tls.certresolver=njalla"
hermes: hermes:
build: ./ build: ./hermes
container_name: hermes container_name: hermes
restart: always restart: always
# Gateway run enables the internal API server on port 8642 # Gateway run enables the internal API server on port 8642
@@ -58,9 +58,11 @@ services:
- ai_backend - ai_backend
ollama: ollama:
image: ollama/ollama:latest build:
context: ./ollama
dockerfile: Dockerfile
image: ollama/ollama:rocm-gfx906
container_name: ollama container_name: ollama
privileged: true
tty: true tty: true
restart: always restart: always
ports: ports:
@@ -78,7 +80,7 @@ services:
- HSA_ENABLE_SDMA=0 - HSA_ENABLE_SDMA=0
- OLLAMA_HOST=0.0.0.0 - OLLAMA_HOST=0.0.0.0
- OLLAMA_DEBUG=1 - OLLAMA_DEBUG=1
- OLLAMA_FLASH_ATTENTION=0 - OLLAMA_FLASH_ATTENTION=1
- OLLAMA_NUM_PARALLEL=2 - OLLAMA_NUM_PARALLEL=2
devices: devices:
# Map the render nodes and KFD for ROCm to work inside the container # Map the render nodes and KFD for ROCm to work inside the container

106
ai/ollama/Dockerfile Normal file
View File

@@ -0,0 +1,106 @@
# ollama-gfx906/Dockerfile
#
# Custom ollama image with ROCm 6.1 + gfx906 (MI50) support.
# The official ollama/rocm image ships ROCm 7.2 which dropped gfx906.
# This uses v0.23.2's native CMake build system with AMDGPU_TARGETS including gfx906.
#
# Build: docker build -t ollama/ollama:rocm-gfx906 ai/ollama
FROM rocm/dev-ubuntu-22.04:6.1.2-complete AS builder
# Build dependencies (CMake, Ninja, Go)
ARG CMAKEVERSION=3.31.2
ARG NINJAVERSION=1.12.1
ARG GOLANG_VERSION=1.22.0
RUN apt-get update && DEBIAN_FRONTEND=noninteractive apt-get install -y \
curl git ccache build-essential pkg-config unzip \
&& rm -rf /var/lib/apt/lists/*
# Install CMake from official binaries
RUN curl -fsSL https://github.com/Kitware/CMake/releases/download/v${CMAKEVERSION}/cmake-${CMAKEVERSION}-linux-x86_64.tar.gz \
| tar xz -C /usr/local --strip-components 1
# Install Ninja
RUN curl -fsSL -o /tmp/ninja.zip \
https://github.com/ninja-build/ninja/releases/download/v${NINJAVERSION}/ninja-linux.zip \
&& unzip /tmp/ninja.zip -d /usr/local/bin && rm /tmp/ninja.zip
# Install Go
RUN curl -fsSL https://go.dev/dl/go${GOLANG_VERSION}.linux-amd64.tar.gz \
| tar xz -C /usr/local
ENV PATH=/usr/local/go/bin:$PATH
ARG OLLAMA_VERSION=v0.23.2
RUN git clone --depth 1 --branch ${OLLAMA_VERSION} https://github.com/ollama/ollama.git /build
WORKDIR /build
# ROCm paths
ENV HIP_PATH=/opt/rocm
ENV ROCM_PATH=/opt/rocm
ENV CMAKE_GENERATOR=Ninja
ENV LDFLAGS=-s
# Step 1: Build CPU backends with GCC (no ROCm preset)
# Pre-set CMAKE_HIP_COMPILER="" to prevent check_language(HIP) from
# finding a HIP compiler (it searches /opt/rocm even without PATH).
# Remove /opt/rocm from PATH to prevent find_program from finding hipcc.
RUN mkdir -p build-cpu && \
PATH=/usr/local/go/bin:/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin \
cmake -B build-cpu -DCMAKE_BUILD_TYPE=Release \
-DCMAKE_HIP_COMPILER="" \
-DCMAKE_INSTALL_PREFIX=/build/dist && \
cmake --build build-cpu --target ggml-cpu -- -l $(nproc) && \
cmake --install build-cpu --component CPU --strip && \
echo "=== CPU install ===" && \
(find /build/dist/lib/ollama -type f -o -type l 2>&1 | head -20 || echo "empty")
# Step 2: Build HIP backend with ROCm preset + gfx906 target only
# The ROCm 6 preset enables HIP language detection (enable_language(HIP))
# which ensures GPU kernels are properly compiled for gfx906.
# OLLAMA_RUNNER_DIR=rocm from the preset, so HIP goes to lib/ollama/rocm/
# Need CMAKE_PREFIX_PATH so find_package(hip) finds hip-config.cmake
# at /opt/rocm/lib/cmake/hip/hip-config.cmake.
RUN mkdir -p build-hip && \
cmake -B build-hip \
--preset 'ROCm 6' \
-DAMDGPU_TARGETS="gfx906:xnack-" \
-DCMAKE_BUILD_TYPE=Release \
-DCMAKE_PREFIX_PATH="/opt/rocm" && \
cmake --build build-hip --target ggml-hip -- -l $(nproc) && \
cmake --install build-hip --component HIP --strip && \
echo "=== HIP install ===" && \
find /build/dist/lib/ollama -type f -o -type l | head -20
# Step 3: Build Go binary (GCC for CGo linking)
ENV CGO_ENABLED=1
RUN go build -trimpath -ldflags="-X=github.com/ollama/ollama/version.Version=${OLLAMA_VERSION}" -o /build/dist/ollama .
# ---------- Runtime image ----------
FROM ubuntu:24.04
RUN apt-get update && DEBIAN_FRONTEND=noninteractive apt-get install -y \
ca-certificates curl libstdc++6 libgomp1 libvulkan1 libopenblas0 \
&& rm -rf /var/lib/apt/lists/*
# Copy ROCm 6.1 runtime libraries
# These are needed at runtime by ggml-hip via LD_LIBRARY_PATH
COPY --from=builder /opt/rocm/lib/ /opt/rocm/lib/
COPY --from=builder /opt/rocm/share/ /opt/rocm/share/
# Copy ollama binary + all backends (CPU + HIP)
# CPU install: /build/dist/lib/ollama/libggml-*.so
# HIP install: /build/dist/lib/ollama/rocm/libggml-hip.so
COPY --from=builder /build/dist/ollama /usr/bin/ollama
COPY --from=builder /build/dist/lib/ollama/ /usr/lib/ollama/
RUN ldconfig
ENV LD_LIBRARY_PATH=/opt/rocm/lib:/usr/lib/ollama/rocm:/usr/lib/ollama
ENV HSA_OVERRIDE_GFX_VERSION=9.0.6
ENV HCC_AMDGPU_TARGET=gfx906
ENV HSA_ENABLE_SDMA=0
EXPOSE 11434
ENTRYPOINT ["/bin/ollama"]
CMD ["serve"]

View File

@@ -7,6 +7,7 @@ services:
- USER_UID=1000 - USER_UID=1000
- USER_GID=1000 - USER_GID=1000
- GITEA__server__ROOT_URL=https://code.lazyworkhorse.net - GITEA__server__ROOT_URL=https://code.lazyworkhorse.net
- GITEA__actions__ENABLED=true
- SSH_PORT=2222 - SSH_PORT=2222
- SSH_LISTEN_PORT=2222 - SSH_LISTEN_PORT=2222
volumes: volumes:
@@ -40,6 +41,22 @@ services:
# Internal Routing # Internal Routing
- "traefik.http.services.gitea.loadbalancer.server.port=3000" - "traefik.http.services.gitea.loadbalancer.server.port=3000"
act_runner:
image: gitea/act_runner:latest
container_name: act_runner
environment:
- GITEA_INSTANCE_URL=http://gitea:3000
- GITEA_RUNNER_REGISTRATION_TOKEN=${GITEA_RUNNER_TOKEN}
- GITEA_RUNNER_NAME=ai-host-runner
- GITEA_RUNNER_LABELS=ubuntu-latest:docker://catthehacker/ubuntu:full-22.04,nixos-builder:docker://nixos/nix
volumes:
- /var/run/docker.sock:/var/run/docker.sock
networks:
- vc_net
restart: always
depends_on:
- gitea
networks: networks:
vc_net: vc_net:
external: true external: true