Skip to content

Remove Gemma 3-4B multimodal redundant variable and optimize vllm_decode hbm_utilization of Lama3.1-70b to prevent OOM - #4662

Open
chiajunglien wants to merge 1 commit into
AI-Hypercomputer:mainfrom
CIeNET-International:emma/fix-redundant
Open

Remove Gemma 3-4B multimodal redundant variable and optimize vllm_decode hbm_utilization of Lama3.1-70b to prevent OOM#4662
chiajunglien wants to merge 1 commit into
AI-Hypercomputer:mainfrom
CIeNET-International:emma/fix-redundant

Conversation

@chiajunglien

Copy link
Copy Markdown
Contributor

Description

This PR removes the redundant variable in tests/end_to_end/tpu/gemma3/4b/test_gemma3_multimodal_sft.sh and adjust the hbm_utilization_vllm of vllm_decode in tests/end_to_end/tpu/llama3.1/70b/test_llama3.1_70b_rl.sh and tests/end_to_end/tpu/llama3.1/70b/test_llama3.1_70b_sft.sh

Tests

xpk workload create-pathways \
  --cluster=v5p-128-bodaborg-europe-west4-b \
  --workload=llama-povllm-$(date +%m%d%H%M) \
  --tpu-type=v5p-128 \
  --num-slices=1 \
  --priority=very-high \
  --project=cloud-tpu-multipod-dev \
  --zone=europe-west4-b \
  --skip-validation \
  --docker-image=[gcr.io/tpu-prod-env-multipod/maxtext_post_training_nightly:30411448246](http://gcr.io/tpu-prod-env-multipod/maxtext_post_training_nightly:30411448246) \
  --command="set -xue; \
    export TPU_MIN_LOG_LEVEL=0; \
    export TF_CPP_MIN_LOG_LEVEL=0; \
    export TPU_STDERR_LOG_LEVEL=0; \
    export JAX_PLATFORMS='proxy,cpu'; \
    export JAX_BACKEND_TARGET='grpc://127.0.0.1:29000'; \
    export ENABLE_PATHWAYS_PERSISTENCE='1'; \
    python3 -m maxtext.inference.vllm_decode \
      model_name=llama3.1-70b \
      tokenizer_path='meta-llama/Llama-3.1-70B-Instruct' \
      load_parameters_path='gs://cloud-tpu-multipod-dev-emma-data/llama3.1-70b-0729/rl/rl_test_0630/checkpoints/actor/5/model_params' \
      vllm_hf_overrides='{\"architectures\": [\"MaxTextForCausalLM\"]}' \
      hbm_utilization_vllm=0.6 \
      prompt='Suggest some famous landmarks in London.' \
      scan_layers=True \
      enable_single_controller=true \
      ici_tensor_parallelism=8"

Checklist

Before submitting this PR, please make sure (put X in square brackets):

  • I have performed a self-review of my code. For an optional AI review, add the gemini-review label.
  • I have necessary comments in my code, particularly in hard-to-understand areas.
  • I have run end-to-end tests tests and provided workload links above if applicable.
  • I have made or will make corresponding changes to the doc if needed, including adding new documentation pages to the relevant Table of Contents (toctree directive) as explained in our documentation.

@gemini-code-assist

Copy link
Copy Markdown

Caution

The consumer version of Gemini Code Assist on GitHub has been sunset. All code review activity has officially ceased.

@codecov

codecov Bot commented Jul 29, 2026

Copy link
Copy Markdown

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant