Skip to content

Fix vLLM Gemma, add vLLM extra #36451

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
damccorm merged 1 commit into from
Oct 24, 2025
Merged

Fix vLLM Gemma, add vLLM extra #36451

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
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
3 changes: 2 additions & 1 deletion ...erence_Python_Benchmarks_Dataflow_Pytorch_Sentiment_Streaming_DistilBert_Base_Uncased.txt
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -31,5 +31,6 @@
--device=CPU
--input_file=gs://apache-beam-ml/testing/inputs/sentences_50k.txt
--runner=DataflowRunner
--dataflow_service_options=worker_accelerator=type:nvidia-tesla-t4;count:1;install-nvidia-driver
--model_path=distilbert-base-uncased-finetuned-sst-2-english
--model_state_dict_path=gs://apache-beam-ml/models/huggingface.sentiment.distilbert-base-uncased.pth
--model_state_dict_path=gs://apache-beam-ml/models/huggingface.sentiment.distilbert-base-uncased.pth
4 changes: 2 additions & 2 deletions ...oad-tests-pipeline-options/beam_Inference_Python_Benchmarks_Dataflow_VLLM_Gemma_Batch.txt
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -20,7 +20,7 @@
--input=gs://apache-beam-ml/testing/inputs/sentences_50k.txt
--machine_type=n1-standard-8
--worker_zone=us-central1-b
--disk_size_gb=50
--disk_size_gb=200
--input_options={}
--num_workers=8
--max_num_workers=25
Expand All @@ -33,4 +33,4 @@
--influx_measurement=gemma_vllm_batch
--model_gcs_path=gs://apache-beam-ml/models/gemma-2b-it
--dataflow_service_options=worker_accelerator=type:nvidia-tesla-t4;count:1;install-nvidia-driver
--experiments=use_runner_v2
--experiments=use_runner_v2
6 changes: 0 additions & 6 deletions .github/workflows/refresh_looker_metrics.yml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -19,19 +19,13 @@ name: Refresh Looker Performance Metrics

on:
workflow_dispatch:
inputs:
READ_ONLY:
description: 'Run in read-only mode'
required: false
default: 'true'

env:
GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
LOOKERSDK_BASE_URL: ${{ secrets.LOOKERSDK_BASE_URL }}
LOOKERSDK_CLIENT_ID: ${{ secrets.LOOKERSDK_CLIENT_ID }}
LOOKERSDK_CLIENT_SECRET: ${{ secrets.LOOKERSDK_CLIENT_SECRET }}
GCS_BUCKET: 'public_looker_explores_us_a3853f40'
READ_ONLY: ${{ inputs.READ_ONLY }}
Copy link
Contributor

@damccorm damccorm Oct 24, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Is this related to the vllm pieces? Rather than removing it entirely, we could also just flip the default to false

Copy link
Collaborator Author

@Amar3tto Amar3tto Oct 24, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I found that this parameter was added by mistake and never used


jobs:
refresh_looker_metrics:
Expand Down
14 changes: 4 additions & 10 deletions sdks/python/apache_beam/ml/inference/test_resources/vllm.dockerfile
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -46,23 +46,17 @@ RUN curl -sS https://bootstrap.pypa.io/get-pip.py | python3 && \
python3 -m pip install --upgrade pip setuptools wheel

# 4) Copy the Beam SDK harness (for Dataflow workers)
COPY --from=gcr.io/apache-beam-testing/beam-sdk/beam_python3.10_sdk:2.68.0.dev \
COPY --from=gcr.io/apache-beam-testing/beam-sdk/beam_python3.10_sdk:latest \
/opt/apache/beam /opt/apache/beam

# 5) Make sure the harness is discovered first
ENV PYTHONPATH=/opt/apache/beam:$PYTHONPATH

# 6) Install the Beam dev SDK from the local source package.
# This .tar.gz file will be created by GitHub Actions workflow
# and copied into the build context.
# and copied into the build context. This will include vLLM dependencies
COPY ./sdks/python/build/apache-beam.tar.gz /tmp/beam.tar.gz
RUN python3 -m pip install --no-cache-dir "/tmp/beam.tar.gz[gcp]"

# 7) Install vLLM, and other dependencies
RUN python3 -m pip install --no-cache-dir \
openai>=1.52.2 \
vllm>=0.6.3 \
triton>=3.1.0
RUN python3 -m pip install --no-cache-dir "/tmp/beam.tar.gz[gcp,vllm]"

# 8) Use the Beam boot script as entrypoint
ENTRYPOINT ["/opt/apache/beam/boot"]
ENTRYPOINT ["/opt/apache/beam/boot"]
2 changes: 1 addition & 1 deletion sdks/python/apache_beam/ml/inference/vllm_tests_requirements.txt
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -19,4 +19,4 @@ torchvision>=0.8.2
pillow>=8.0.0
transformers>=4.18.0
google-cloud-monitoring>=2.27.0
openai>=1.52.2
openai>=1.52.2
2 changes: 1 addition & 1 deletion sdks/python/apache_beam/testing/benchmarks/inference/vllm_gemma_benchmarks.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -26,7 +26,7 @@ def __init__(self):
self.metrics_namespace = "BeamML_vLLM"
super().__init__(
metrics_namespace=self.metrics_namespace,
pcollection="WriteBQ.out0",
pcollection="FormatForBQ.out0",
)

def test(self):
Expand Down
3 changes: 2 additions & 1 deletion sdks/python/setup.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -610,7 +610,8 @@ def get_portability_package_data():
],
'xgboost': ['xgboost>=1.6.0,<2.1.3', 'datatable==1.0.0'],
'tensorflow-hub': ['tensorflow-hub>=0.14.0,<0.16.0'],
'milvus': milvus_dependency
'milvus': milvus_dependency,
'vllm': ['openai==1.107.1', 'vllm==0.10.1.1', 'triton==3.3.1']
},
zip_safe=False,
# PyPI package information.
Expand Down
10 changes: 5 additions & 5 deletions website/www/site/data/performance.yaml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -238,15 +238,15 @@ looks:
write:
folder: 86
cost:
- id: tJWFWW3cnF2CWpmK2zZdXGvWmtNnJgrC
- id: J5TtpRykjwPs4W6S88FnJ28Tr8sSHpqN
title: RunTime and EstimatedCost
date:
- id: J5TtpRykjwPs4W6S88FnJ28Tr8sSHpqN
- id: tJWFWW3cnF2CWpmK2zZdXGvWmtNnJgrC
title: AvgThroughputBytesPerSec by Date
- id: Jf6qGqN25Zf787DpkNDX5CBpGRvCGMXp
title: AvgThroughputElementsPerSec by Date
version:
- id: dKyJy5ZKhkBdSTXRY3wZR6fXzptSs2qm
title: AvgThroughputBytesPerSec by Version
- id: Qwxm27qY4fqT4CxXsFfKm2g3734TFJNN
title: AvgThroughputElementsPerSec by Version
title: AvgThroughputBytesPerSec by Version
- id: dKyJy5ZKhkBdSTXRY3wZR6fXzptSs2qm
title: AvgThroughputElementsPerSec by Version
Loading