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

Dockerfiles: use LLAMA_AVX2_ONLY.

Browse source 
Docker builds produced by GitHub actions seem to fail on Alder Lake
machines.

The purpose of the Docker builds is to provide an easy way to start
generating tokens. Targeting AVX2-only is a compromise between performance
and compatibility.
This commit is contained in:
Bryan Silverthorn 2023-05-05 10:49:03 -07:00
parent 2e3b2e7e09
commit e0ed30dffe
2 changed files with 2 additions and 2 deletions
Download patch file Download diff file Expand all files Collapse all files

2
.devops/full.Dockerfile
View file

@ -14,6 +14,6 @@ WORKDIR /app
COPY . .
RUN make
RUN LLAMA_AVX2_ONLY=1 make
ENTRYPOINT ["/app/.devops/tools.sh"]

2
.devops/main.Dockerfile
View file

@ -9,7 +9,7 @@ WORKDIR /app
COPY . .
RUN make
RUN LLAMA_AVX2_ONLY=1 make
FROM ubuntu:$UBUNTU_VERSION as runtime