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, 00:45 UTC
[ci/build] fix gh200 test (#12681)
Signed-off-by: youkaichao <youkaichao@gmail.com>
youkaichao committed
Feb 3, 2025 at 15:59 UTC
1298a400e8e8496b6e9ce3847ada5ec9f6e6cb48
3 files changed
+5
-5
.buildkite/check-wheel-size.py
+2
-2
@@ -4,11 +4,11 @@ import os
4
import sys
5
import zipfile
6
7
-# Read the VLLM_MAX_SIZE_MB environment variable, defaulting to 300 MiB
7
+# Read the VLLM_MAX_SIZE_MB environment variable, defaulting to 400 MiB
8
# Note that we have 400 MiB quota, please use it wisely.
9
# See https://github.com/pypi/support/issues/3792 .
10
# Please also sync the value with the one in Dockerfile.
11
-VLLM_MAX_SIZE_MB = int(os.environ.get('VLLM_MAX_SIZE_MB', 300))
11
+VLLM_MAX_SIZE_MB = int(os.environ.get('VLLM_MAX_SIZE_MB', 400))
12
13
14
def print_top_10_largest_files(zip_file):
.buildkite/run-gh200-test.sh
+2
-2
@@ -23,6 +23,6 @@ trap remove_docker_container EXIT
23
remove_docker_container
24
25
# Run the image and test offline inference
26
-docker run --name gh200-test --gpus=all --entrypoint="" gh200-test bash -c '
27
- python3 examples/offline_inference/basic.py
26
+docker run -e HF_TOKEN -v /root/.cache/huggingface:/root/.cache/huggingface --name gh200-test --gpus=all --entrypoint="" gh200-test bash -c '
27
+ python3 examples/offline_inference/cli.py --model meta-llama/Llama-3.2-1B
28
'
Dockerfile
+1
-1
@@ -127,7 +127,7 @@ RUN --mount=type=cache,target=/root/.cache/ccache \
127
# Check the size of the wheel if RUN_WHEEL_CHECK is true
128
COPY .buildkite/check-wheel-size.py check-wheel-size.py
129
# sync the default value with .buildkite/check-wheel-size.py
130
-ARG VLLM_MAX_SIZE_MB=300
130
+ARG VLLM_MAX_SIZE_MB=400
131
ENV VLLM_MAX_SIZE_MB=$VLLM_MAX_SIZE_MB
132
ARG RUN_WHEEL_CHECK=true
133
RUN if [ "$RUN_WHEEL_CHECK" = "true" ]; then \