@tduncan / Nvidia-Optimization-Stack-VLLM Mirror · read-only / commits / bd7efe95d0
Fork 0

A high-throughput and memory-efficient inference and serving engine for LLMs

Mirrored from https://github.com/tduncan-lum/Nvidia-Optimization-Stack---VLLM.git

Synced Nov 16, 23:15 UTC

Add ccache to amd (#5555)

Simon Mo committed Jun 14, 2024 at 19:18 UTC bd7efe95d03773c65fa7dc1e122f3ce0e079a542
2 files changed +5 -1
.buildkite/test-template-aws.j2
+1
@@ -30,6 +30,7 @@ steps:
30 command: bash .buildkite/run-amd-test.sh "cd {{ (step.working_dir or default_working_dir) | safe }} ; {{ step.command or (step.commands | join(" ; ")) | safe }}"
31 env:
32 DOCKER_BUILDKIT: "1"
33 + priority: 100
34 soft_fail: true
35 {% endif %}
36 {% endfor %}
Dockerfile.rocm
+4 -1
@@ -42,6 +42,7 @@ RUN apt-get update && apt-get install -y \
42 unzip \
43 nvidia-cuda-toolkit \
44 tmux \
45 + ccache \
46 && rm -rf /var/lib/apt/lists/*
47
48 ### Mount Point ###
@@ -102,7 +103,9 @@ ENV RAY_EXPERIMENTAL_NOSET_ROCR_VISIBLE_DEVICES=1
103
104 ENV VLLM_NCCL_SO_PATH=/opt/rocm/lib/librccl.so
105
105 -RUN --mount=type=cache,target=/root/.cache/pip \
106 +ENV CCACHE_DIR=/root/.cache/ccache
107 +RUN --mount=type=cache,target=/root/.cache/ccache \
108 + --mount=type=cache,target=/root/.cache/pip \
109 pip install -U -r requirements-rocm.txt \
110 && patch /opt/rocm/include/hip/amd_detail/amd_hip_bf16.h ./rocm_patch/rocm_bf16.patch \
111 && python3 setup.py install \