use vllm 0.19.0 for torch 2.10.0

[winglian]

Summary by CodeRabbit

• Chores
  • Updated vllm dependency to require version 0.19.0 or later for enhanced compatibility.

[coderabbitai]

📝 Walkthrough

Walkthrough

Updated the vllm minimum version requirement from 0.17.1 to 0.19.0 in setup.py's parse_requirements() function for Torch ≥ 2.10, following removal of fbgemm-gpu constraints and conditional xformers requirement adjustments.

Changes

Cohort / File(s)	Summary
Dependency version update setup.py	Bumped vllm minimum version from 0.17.1 to 0.19.0 for Torch ≥ 2.10 branch.

Estimated code review effort

🎯 1 (Trivial) | ⏱️ ~2 minutes

Possibly related PRs

• upgrade vllm to v0.14.0 #3345: Also modifies vllm version constraints in setup.py's parse_requirements for PyTorch-specific branches.

Suggested reviewers

• salmanmohammadi

🚥 Pre-merge checks | ✅ 2 | ❌ 1

❌ Failed checks (1 warning)

Check name	Status	Explanation	Resolution
Docstring Coverage	⚠️ Warning	Docstring coverage is 0.00% which is insufficient. The required threshold is 80.00%.	Write docstrings for the functions missing them to satisfy the coverage threshold.

✅ Passed checks (2 passed)

Check name	Status	Explanation
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Title check	✅ Passed	The title accurately describes the main change: updating the vllm version requirement from 0.17.1 to 0.19.0 for the torch 2.10 branch.

_{✏️ Tip: You can configure your own custom pre-merge checks in the settings.}

✨ Finishing Touches

📝 Generate docstrings

• Create stacked PR
• Commit on current branch

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch vllm-0190

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

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

[coderabbitai]

🧹 Nitpick comments (1)

setup.py (1)

92-92: Constrain vllm to 0.19.x for reproducible installs.

Line 92 uses an open-ended lower bound (vllm>=0.19.0), which is inconsistent with all other PyTorch version branches in setup.py (which pin to exact versions: ==0.13.0, ==0.14.0, ==0.11.0, etc.). Combined with the version-sensitive vLLM API handling in src/axolotl/scripts/vllm_serve_lora.py (conditional imports for GuidedDecodingParams, StructuredOutputsParams, and multiple import fallbacks), this open range risks silently pulling future 0.x releases with breaking changes. Constrain to the 0.19 series for torch >=2.10.

Proposed change

-                extras_require_map["vllm"] = ["vllm>=0.19.0"]
+                extras_require_map["vllm"] = ["vllm>=0.19.0,<0.20.0"]

🤖 Prompt for AI Agents

Verify each finding against the current code and only fix it if needed.

In `@setup.py` at line 92, The extras_require entry extras_require_map["vllm"]
currently allows any vllm >=0.19.0; change it to constrain the package to the
0.19 series (e.g. "vllm>=0.19.0,<0.20.0") so installs for the torch >=2.10
branch remain reproducible and avoid pulling future 0.x releases that may break
the conditional imports and API handling in
src/axolotl/scripts/vllm_serve_lora.py (GuidedDecodingParams,
StructuredOutputsParams, etc.).

🤖 Prompt for all review comments with AI agents

Verify each finding against the current code and only fix it if needed.

Nitpick comments:
In `@setup.py`:
- Line 92: The extras_require entry extras_require_map["vllm"] currently allows
any vllm >=0.19.0; change it to constrain the package to the 0.19 series (e.g.
"vllm>=0.19.0,<0.20.0") so installs for the torch >=2.10 branch remain
reproducible and avoid pulling future 0.x releases that may break the
conditional imports and API handling in src/axolotl/scripts/vllm_serve_lora.py
(GuidedDecodingParams, StructuredOutputsParams, etc.).

---

ℹ️ Review info

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Pro

Run ID: 0a8ff5b6-5e83-4144-aad7-f1b5a63c94ff

📥 Commits

Reviewing files that changed from the base of the PR and between 149178d and a95b3e4.

📒 Files selected for processing (1)

• setup.py

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!