vbatts/llama.cpp
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

Fix Docker ROCM builds, use AMDGPU_TARGETS instead of GPU_TARGETS (#9641)

Browse source 
* Fix Docker ROCM builds, use AMDGPU_TARGETS instead of GPU_TARGETS

* Set ROCM_DOCKER_ARCH as string due it incorrectly build and cause OOM exit code
This commit is contained in:
serhii-nakon 2024-09-30 21:57:12 +03:00 • committed by GitHub
parent 511636df0c
commit 6f1d9d71f4
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
4 changed files with 10 additions and 10 deletions
Download patch file Download diff file Expand all files Collapse all files

6
.devops/full-rocm.Dockerfile
View file

@ -11,7 +11,7 @@ FROM ${BASE_ROCM_DEV_CONTAINER} AS build
# Unless otherwise specified, we make a fat build.
# List from https://github.com/ggerganov/llama.cpp/pull/1087#issuecomment-1682807878
# This is mostly tied to rocBLAS supported archs.
ARG ROCM_DOCKER_ARCH=\
ARG ROCM_DOCKER_ARCH="\
gfx803 \
gfx900 \
gfx906 \
@ -21,7 +21,7 @@ ARG ROCM_DOCKER_ARCH=\
gfx1030 \
gfx1100 \
gfx1101 \
gfx1102
gfx1102"
COPY requirements.txt requirements.txt
COPY requirements requirements
@ -34,7 +34,7 @@ WORKDIR /app
COPY . .
# Set nvcc architecture
ENV GPU_TARGETS=${ROCM_DOCKER_ARCH}
ENV AMDGPU_TARGETS=${ROCM_DOCKER_ARCH}
# Enable ROCm
ENV GGML_HIPBLAS=1
ENV CC=/opt/rocm/llvm/bin/clang

6
.devops/llama-cli-rocm.Dockerfile
View file

@ -11,7 +11,7 @@ FROM ${BASE_ROCM_DEV_CONTAINER} AS build
# Unless otherwise specified, we make a fat build.
# List from https://github.com/ggerganov/llama.cpp/pull/1087#issuecomment-1682807878
# This is mostly tied to rocBLAS supported archs.
ARG ROCM_DOCKER_ARCH=\
ARG ROCM_DOCKER_ARCH="\
gfx803 \
gfx900 \
gfx906 \
@ -21,7 +21,7 @@ ARG ROCM_DOCKER_ARCH=\
gfx1030 \
gfx1100 \
gfx1101 \
gfx1102
gfx1102"
COPY requirements.txt requirements.txt
COPY requirements requirements
@ -34,7 +34,7 @@ WORKDIR /app
COPY . .
# Set nvcc architecture
ENV GPU_TARGETS=${ROCM_DOCKER_ARCH}
ENV AMDGPU_TARGETS=${ROCM_DOCKER_ARCH}
# Enable ROCm
ENV GGML_HIPBLAS=1
ENV CC=/opt/rocm/llvm/bin/clang

6
.devops/llama-server-rocm.Dockerfile
View file

@ -11,7 +11,7 @@ FROM ${BASE_ROCM_DEV_CONTAINER} AS build
# Unless otherwise specified, we make a fat build.
# List from https://github.com/ggerganov/llama.cpp/pull/1087#issuecomment-1682807878
# This is mostly tied to rocBLAS supported archs.
ARG ROCM_DOCKER_ARCH=\
ARG ROCM_DOCKER_ARCH="\
gfx803 \
gfx900 \
gfx906 \
@ -21,7 +21,7 @@ ARG ROCM_DOCKER_ARCH=\
gfx1030 \
gfx1100 \
gfx1101 \
gfx1102
gfx1102"
COPY requirements.txt requirements.txt
COPY requirements requirements
@ -34,7 +34,7 @@ WORKDIR /app
COPY . .
# Set nvcc architecture
ENV GPU_TARGETS=${ROCM_DOCKER_ARCH}
ENV AMDGPU_TARGETS=${ROCM_DOCKER_ARCH}
# Enable ROCm
ENV GGML_HIPBLAS=1
ENV CC=/opt/rocm/llvm/bin/clang