chore(ci): fix model path using mounted pvc location(wip)

[key4ng]

Description

Problem

Solution

Changes

Test Plan

Checklist

• cargo +nightly fmt passes
• cargo clippy --all-targets --all-features -- -D warnings passes
• (Optional) Documentation updated
• (Optional) Please join us on Slack #sig-smg to discuss, review, and merge PRs

Summary by CodeRabbit

• Chores
  • Updated continuous integration and benchmark infrastructure configuration for improved model loading and caching efficiency across test environments.

[gemini-code-assist]

Note

Gemini is unable to generate a summary for this pull request due to the file types involved not being currently supported.

[coderabbitai]

Caution

Review failed

The pull request is closed.

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: Organization UI

Review profile: ASSERTIVE

Plan: Pro

Run ID: 0c4d3a02-6094-4db4-a6bb-394be6d815d3

📥 Commits

Reviewing files that changed from the base of the PR and between e231a73 and f3ae25d.

📒 Files selected for processing (2)

• .github/workflows/nightly-benchmark.yml
• .github/workflows/pr-test-rust.yml

---

📝 Walkthrough

Walkthrough

Updated GitHub Actions workflows to replace ROUTER_LOCAL_MODEL_PATH environment variable with HF_HOME across benchmark job invocations in nightly and PR test workflows. This standardizes model loading configuration to use Hugging Face's standard environment variable across multiple CI job variants.

Changes

Cohort / File(s)	Summary
Nightly Benchmark Workflows .github/workflows/nightly-benchmark.yml	Replaced ROUTER_LOCAL_MODEL_PATH="/raid/models" with HF_HOME="/models" in single-worker, multi-worker, and single-worker-h200 job pytest invocations.
PR Test Rust Workflows .github/workflows/pr-test-rust.yml	Replaced ROUTER_LOCAL_MODEL_PATH="/home/ubuntu/models" with HF_HOME="/models" in benchmarks and go-bindings-e2e job pytest invocations.

Estimated code review effort

🎯 2 (Simple) | ⏱️ ~8 minutes

Possibly related PRs

• fix(ci): fix H200 nightly benchmark model path, worker logs and CUDA errors  #411: Modifies CI workflow environment variables and ROUTER_LOCAL_MODEL_PATH configuration, directly related to the model path standardization effort.

Suggested labels

ci

Suggested reviewers

• CatherineSue
• slin1237
• XinyueZhang369

Poem

🐰 Hop, hop, through the workflows we go,
ROUTER_LOCAL swaps for HF_HOME's glow,
Models now load with a standard care,
Benchmarks will run with variables fair! ✨

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch fix-model-path

📝 Coding Plan

• Generate coding plan for human review comments

---

_{Comment @coderabbitai help to get the list of available commands and usage tips.}

[mergify]

Closing this PR because the branch name fix-model-path does not follow the naming convention.

To fix this, please rename your branch locally, push the new branch, and open a new PR:

git branch -m <new-branch-name>
git push origin -u <new-branch-name>

[mergify]

Hi @key4ng, the branch fix-model-path does not follow our naming convention.

Please use one of the following formats:

• <type>/<description> — e.g. feat/add-auth, fix/null-pointer, dependabot/cargo/pyo3-0.28.1
• <username>/<description> — e.g. changsu/fix-routing

Allowed types: feat, fix, chore, docs, refactor, test, ci, perf

Note: PRs with non-conforming branch names will be auto-closed. Please follow the naming convention for all branches.

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: f3ae25d073

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[chatgpt-codex-connector]

Preserve ROUTER_LOCAL_MODEL_PATH in benchmark invocation

Switching this env var to HF_HOME breaks the local-model path plumbing used by the benchmark fixtures: get_model_spec() only checks ROUTER_LOCAL_MODEL_PATH when resolving model paths (e2e_test/infra/model_specs.py), and benchmark command construction only mounts/uses local tokenizer directories when ROUTER_LOCAL_MODEL_PATH is set (e2e_test/benchmarks/conftest.py). In this configuration, tests fall back to Hub model IDs instead of the mounted PVC model tree, which can cause nightly/CI benchmark runs to download from Hugging Face (or fail on gated/offline environments) instead of using local models.

Useful? React with 👍 / 👎.