Skip to content

Fix vLLM Gemma benchmark and PyTorch language modeling tests #37874

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
Amar3tto merged 1 commit into from
Mar 18, 2026
Merged

Fix vLLM Gemma benchmark and PyTorch language modeling tests #37874

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
1 change: 1 addition & 0 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 @@ -32,5 +32,6 @@
--metrics_table=gemma_vllm_batch
--influx_measurement=gemma_vllm_batch
--model_gcs_path=gs://apache-beam-ml/models/gemma-2b-it
--requirements_file=apache_beam/ml/inference/vllm_tests_requirements.txt
--dataflow_service_options=worker_accelerator=type:nvidia-tesla-t4;count:1;install-nvidia-driver
--experiments=use_runner_v2
3 changes: 1 addition & 2 deletions sdks/python/apache_beam/examples/inference/pytorch_language_modeling.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -52,8 +52,7 @@ def tokenize_sentence(
text_and_mask: tuple[str, str],
bert_tokenizer: BertTokenizer) -> tuple[str, dict[str, torch.Tensor]]:
text, masked_text = text_and_mask
tokenized_sentence = bert_tokenizer.encode_plus(
masked_text, return_tensors="pt")
tokenized_sentence = bert_tokenizer(masked_text, return_tensors="pt")

# Workaround to manually remove batch dim until we have the feature to
# add optional batching flag.
Expand Down
1 change: 0 additions & 1 deletion sdks/python/apache_beam/examples/inference/vllm_gemma_batch.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -103,7 +103,6 @@ def run(argv=None, save_main_session=True, test_pipeline=None):

gem = opts.view_as(GemmaVLLMOptions)
opts.view_as(SetupOptions).save_main_session = save_main_session

logging.info("Pipeline starting with model path: %s", gem.model_gcs_path)
handler = GcsVLLMCompletionsModelHandler(
model_name=gem.model_gcs_path,
Expand Down
3 changes: 2 additions & 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 @@ -17,6 +17,7 @@
torch>=1.7.1
torchvision>=0.8.2
pillow>=8.0.0
transformers>=4.18.0
transformers==4.57.1
sentencepiece==0.2.1
google-cloud-monitoring>=2.27.0
openai>=1.52.2
Loading