Skip to content

[NVIDIA] Update vLLM version to latest v0.10.2 on H200. #11

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
kimbochen merged 5 commits into from
Sep 15, 2025
Merged

[NVIDIA] Update vLLM version to latest v0.10.2 on H200. #11

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
5 commits
Select commit Hold shift + click to select a range
a157a04
upgrade to vllm 0.10.2
kedarpotdar-nv Sep 15, 2025
6c4230b
update sqsh name to allow for new containers
kedarpotdar-nv Sep 15, 2025
2070359
ignore log in vllm 0.10.2
kedarpotdar-nv Sep 15, 2025
43aafa0
add export PYTHONNOUSERSITE=1
kedarpotdar-nv Sep 15, 2025
032510b
add PYTHONNOUSERSITE=1 only to vllm serve
kedarpotdar-nv Sep 15, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
2 changes: 1 addition & 1 deletion .github/workflows/70b-tmpl.yml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -62,7 +62,7 @@ jobs:
secrets: inherit
with:
runner: h200
image: 'kedarpotdar147/vllm0.1:latest'
image: 'vllm/vllm-openai:v0.10.2'
model: 'nvidia/Llama-3.3-70B-Instruct-FP8'
framework: 'vllm'
precision: 'fp8'
Expand Down
2 changes: 2 additions & 0 deletions .github/workflows/runner-sweep-test.yml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -24,6 +24,7 @@ on:
options:
- 'kedarpotdar147/vllm0.1:latest'
- 'kedarpotdar147/vllm:05'
- 'vllm/vllm-openai:v0.10.2'
- 'rocm/7.0-preview:rocm7.0_preview_ubuntu_22.04_vllm_0.10.1_instinct_rc1'
- 'rocm/vllm-dev:open-mi300-08052025'
- 'rocm/vllm-dev:open-mi355-08052025'
Expand All @@ -32,6 +33,7 @@ on:
- 'rocm/7.0-preview:rocm7.0_preview_ubuntu_22.04_sgl-dev-v0.5.2rc2-mi30x_rc1'
- 'nvcr.io#nvidia/tensorrt-llm/release:1.1.0rc2'
- 'nvcr.io#nvidia/tensorrt-llm/release:1.1.0rc2.post1'
- 'nvcr.io#nvidia/tensorrt-llm/release:1.1.0rc2.post2'

model:
description: 'Model'
Expand Down
24 changes: 17 additions & 7 deletions benchmarks/70b_fp8_h200_slurm.sh
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -18,21 +18,31 @@ echo "JOB $SLURM_JOB_ID running on $SLURMD_NODENAME"

set -x
hf download $MODEL
pip install datasets pandas

# Create config.yaml
cat > config.yaml << EOF
kv-cache-dtype: fp8
async-scheduling: true
no-enable-prefix-caching: true
max-num-batched-tokens: 8192
max-model-len: 10240
EOF

SERVER_LOG=$(mktemp /tmp/server-XXXXXX.log)
PORT=$(( 8888 + $PORT_OFFSET ))

export TORCH_CUDA_ARCH_LIST="9.0"
vllm serve $MODEL --host 0.0.0.0 --port $PORT \
--trust-remote-code --quantization modelopt --gpu-memory-utilization 0.9 \
--pipeline-parallel-size 1 --tensor-parallel-size $TP --max-num-seqs $CONC --max-num-batched-tokens 8192 --max-model-len $MAX_MODEL_LEN \
--enable-chunked-prefill --async-scheduling --no-enable-prefix-caching \
--compilation-config '{"pass_config": {"enable_fi_allreduce_fusion": true}, "custom_ops": ["+rms_norm"], "level": 3}' \
--disable-log-requests > $SERVER_LOG 2>&1 &

PYTHONNOUSERSITE=1 vllm serve $MODEL --host 0.0.0.0 --port $PORT --config config.yaml \
--gpu-memory-utilization 0.9 --tensor-parallel-size $TP --max-num-seqs $CONC \
--disable-log-requests > $SERVER_LOG 2>&1 &

set +x
while IFS= read -r line; do
printf '%s\n' "$line"
if [[ "$line" =~ [Ee][Rr][Rr][Oo][Rr] ]]; then
# Ignore intel_extension_for_pytorch import errors
if [[ "$line" =~ [Ee][Rr][Rr][Oo][Rr] ]] && [[ ! "$line" =~ "intel_extension_for_pytorch" ]]; then
sleep 5
tail -n100 $SERVER_LOG
echo "JOB $SLURM_JOB_ID ran on NODE $SLURMD_NODENAME"
Expand Down
2 changes: 1 addition & 1 deletion runners/launch_h200-cw.sh
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -7,7 +7,7 @@ MODEL_CODE="${EXP_NAME%%_*}"
FRAMEWORK_SUFFIX=$([[ "$FRAMEWORK" == "trt" ]] && printf '_trt' || printf '')

PARTITION="h200"
SQUASH_FILE="/mnt/vast/squash/image_${MODEL_CODE}_h200${FRAMEWORK_SUFFIX}.sqsh"
SQUASH_FILE="/mnt/vast/squash/image_${MODEL_CODE}_h200${FRAMEWORK_SUFFIX}-0915.sqsh"

salloc --partition=$PARTITION --gres=gpu:$TP --exclusive --time=180 --no-shell
JOB_ID=$(squeue -u $USER -h -o %A | head -n1)
Expand Down
2 changes: 1 addition & 1 deletion runners/launch_h200-nb.sh
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -7,7 +7,7 @@ MODEL_CODE="${EXP_NAME%%_*}"
FRAMEWORK_SUFFIX=$([[ "$FRAMEWORK" == "trt" ]] && printf '_trt' || printf '')

PARTITION="main"
SQUASH_FILE="/home/squash/image_${MODEL_CODE}_h200${FRAMEWORK_SUFFIX}.sqsh"
SQUASH_FILE="/home/squash/image_${MODEL_CODE}_h200${FRAMEWORK_SUFFIX}-0915.sqsh"

salloc --partition=$PARTITION --gres=gpu:$TP --exclusive --time=180 --no-shell
JOB_ID=$(squeue -u $USER -h -o %A | head -n1)
Expand Down
2 changes: 1 addition & 1 deletion runners/launch_h200-nv.sh
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -7,7 +7,7 @@ MODEL_CODE="${EXP_NAME%%_*}"
FRAMEWORK_SUFFIX=$([[ "$FRAMEWORK" == "trt" ]] && printf '_trt' || printf '')

PARTITION="dgx-h200"
SQUASH_FILE="/raid/image_${MODEL_CODE}_h200${FRAMEWORK_SUFFIX}.sqsh"
SQUASH_FILE="/raid/image_${MODEL_CODE}_h200${FRAMEWORK_SUFFIX}-0915.sqsh"

salloc --partition=$PARTITION --gres=gpu:$TP --exclusive --time=180 --no-shell
JOB_ID=$(squeue -u $USER -h -o %A | head -n1)
Expand Down