Skip to content

[NVIDIA] Add SGL B200 FP4 docker script #38

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
functionstackx merged 9 commits into from
Sep 23, 2025
Merged

[NVIDIA] Add SGL B200 FP4 docker script #38

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
9 commits
Select commit Hold shift + click to select a range
192c3ab
add warmup for dsr1 fp4
kedarpotdar-nv Sep 22, 2025
057b082
update conc for bmk-b200 runner sweep
kedarpotdar-nv Sep 22, 2025
fbd73c2
improve warmup logic
kedarpotdar-nv Sep 22, 2025
ebbd381
revert test sweep logic
kedarpotdar-nv Sep 22, 2025
4b11fa8
add warmup update
kedarpotdar-nv Sep 22, 2025
d111cd1
add conc list
kedarpotdar-nv Sep 22, 2025
1df3fef
improve warmup
kedarpotdar-nv Sep 22, 2025
89ed378
revert conc list for runner-sweep-test
kedarpotdar-nv Sep 22, 2025
3beead2
reduce concurreny for b200 dsr1 fp4
kedarpotdar-nv Sep 22, 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
4 changes: 2 additions & 2 deletions .github/workflows/dsr1-tmpl.yml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -151,7 +151,7 @@ jobs:
uses: ./.github/workflows/benchmark-tmpl.yml
secrets: inherit
with:
runner: b200-trt
runner: b200
image: 'lmsysorg/sglang:v0.5.3rc1-cu129-b200'
model: 'nvidia/DeepSeek-R1-0528-FP4'
framework: 'sglang'
Expand All @@ -162,7 +162,7 @@ jobs:
max-model-len: ${{ inputs.max-model-len }}
random-range-ratio: ${{ inputs.random-range-ratio }}
tp-list: '[8]'
conc-list: '[4, 8, 16, 32, 64, 128, 256]' # Custom concurrency values for this job
conc-list: '[4, 8, 16, 32, 64, 128]' # Custom concurrency values for this job

bmk-b200-trt-fp4:
if: ${{ inputs.use_b200 }}
Expand Down
1 change: 1 addition & 0 deletions benchmarks/dsr1_fp4_b200_docker.sh
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -15,3 +15,4 @@ PYTHONNOUSERSITE=1 python3 -m sglang.launch_server --model-path $MODEL --host 0.
--chunked-prefill-size 16384 \
--enable-ep-moe --quantization modelopt_fp4 --enable-flashinfer-allreduce-fusion --scheduler-recv-interval $SCHEDULER_RECV_INTERVAL \
--enable-symm-mem --disable-radix-cache --attention-backend trtllm_mla --enable-flashinfer-trtllm-moe --stream-interval 10

19 changes: 19 additions & 0 deletions runners/launch_b200-nvd.sh
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -28,6 +28,25 @@ done < <(docker logs -f --tail=0 $server_name 2>&1)

git clone https://github.com/kimbochen/bench_serving.git

# warmup for JIT kernels - only for DeepSeek-R1-0528-FP4 model
if [[ "$MODEL" == "nvidia/DeepSeek-R1-0528-FP4" ]]; then
echo "Running warmup for DeepSeek-R1-0528-FP4 model..."
WARMUP_PROMPTS=$(( $CONC * 10 ))
echo "Warmup prompts: $WARMUP_PROMPTS"
docker run --rm --network host --name warmup-client \
-v $GITHUB_WORKSPACE:/workspace/ -w /workspace/ \
-e HF_TOKEN -e PYTHONPYCACHEPREFIX=/tmp/pycache/ \
--entrypoint=/bin/bash \
$(echo "$IMAGE" | sed 's/#/\//') \
-lc "pip install -q datasets pandas && \
python3 bench_serving/benchmark_serving.py \
--model $MODEL --backend vllm --base-url http://localhost:$PORT \
--dataset-name random \
--random-input-len $ISL --random-output-len $OSL --random-range-ratio $RANDOM_RANGE_RATIO \
--num-prompts $WARMUP_PROMPTS --max-concurrency $CONC \
--request-rate inf --ignore-eos"
fi

set -x
docker run --rm --network host --name $client_name \
-v $GITHUB_WORKSPACE:/workspace/ -w /workspace/ \
Expand Down