software-engineering/spark-vllm-docker
2
0
Fork 0
Code Issues Actions Packages Projects Releases Wiki Activity

bumped up MXFP4 base image version

Browse Source
This commit is contained in:
Eugene Rakhmatulin
2026-01-31 16:12:33 -08:00
parent 9691eed1b0
commit c81edce091
1 changed files with 5 additions and 2 deletions
Download Patch File Download Diff File Expand all files Collapse all files

7
Dockerfile.mxfp4
View File

@@ -6,7 +6,7 @@ ARG BUILD_JOBS=16
# =========================================================
# STAGE 1: Base Image (Installs Dependencies)
# =========================================================
FROM nvcr.io/nvidia/pytorch:25.12-py3 AS base
FROM nvcr.io/nvidia/pytorch:26.01-py3 AS base
# Build parallemism
ARG BUILD_JOBS
@@ -225,7 +225,7 @@ RUN --mount=type=cache,id=ccache,target=/root/.ccache \
# =========================================================
# STAGE 4: Runner (Transfers only necessary artifacts)
# =========================================================
FROM nvcr.io/nvidia/pytorch:25.12-py3 AS runner
FROM nvcr.io/nvidia/pytorch:26.01-py3 AS runner
ENV DEBIAN_FRONTEND=noninteractive
ENV PIP_BREAK_SYSTEM_PACKAGES=1
@@ -275,3 +275,6 @@ RUN chmod +x $VLLM_BASE_DIR/run-cluster-node.sh
RUN --mount=type=cache,id=uv-cache,target=/root/.cache/uv \
uv pip install ray[default] fastsafetensors
# If not compiling Triton
# remove triton-kernels as they are not compatible with this vLLM version yet
RUN uv pip uninstall triton-kernels