fix(deepseek): follow pypto-lib kernel dir rename v4 -> v4-flash

[lwDavid]

Summary

Follows the pypto-lib kernel-directory rename models/deepseek/v4 → models/deepseek/v4-flash (pypto-lib splits the variant into v4-flash + v4-pro; v4-flash is the serving/W8A8 variant). Without this, serving's DeepSeekV4 startup fails with FileNotFoundError: Cannot locate DeepSeekV4 kernels.

Changes

pypto_serving/model/deepseek/npu_executor.py (3 lines, path component only):

• _find_pypto_lib_deepseek_v4_dir — both candidate paths (PYPTO_ROOT/models/deepseek/v4 and <checkout>/pypto-lib/models/deepseek/v4) → .../v4-flash.
• _is_deepseek_v4_module_file — the ("models", "deepseek", "v4") parts check → ("models", "deepseek", "v4-flash").

Function names keep the deepseek_v4 model-family identifier — only the filesystem path component moves, matching how the qwen discovery (models/qwen3/14b) is hardcoded.

Why it's safe

The unit tests monkeypatch _find_pypto_lib_deepseek_v4_dir (tests/test_deepseek_v4.py), so they never exercise the real path and need no changes. The real discovery only runs at device-serving startup.

Coordination / follow-up

• Depends on the pypto-lib rename (feat(dsv4): split v4 into v4-flash + v4-pro, add dedicated A5 daily CI pypto-lib#816); should land alongside/after it.
• Follow-up (separate): bump the pypto-lib submodule pin to a commit that carries models/deepseek/v4-flash, so a checked-out serving tree is self-consistent. Not needed for serving's unit CI (discovery is monkeypatched there).

[gemini-code-assist]

Caution

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

[coderabbitai]

Important

Review skipped

Auto incremental reviews are disabled on this repository.

Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

⚙️ Run configuration

Configuration used: Organization UI

Review profile: CHILL

Plan: Pro Plus

Run ID: c431c121-061b-4f66-abd6-1722e6ae3898

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

• 🔍 Trigger review

📝 Walkthrough

Walkthrough

DeepSeek V4 kernel directory discovery and module detection now use models/deepseek/v4-flash instead of models/deepseek/v4.

Changes

DeepSeek V4 Flash Kernel Path

Layer / File(s)	Summary
Kernel path discovery and detection pypto_serving/model/deepseek/npu_executor.py	Updates directory lookup and module-path checks to recognize DeepSeek V4 kernels under models/deepseek/v4-flash.

Estimated code review effort: 1 (Trivial) | ~5 minutes

Possibly related PRs

• hw-native-sys/pypto-serving#93: Updates DeepSeek V4 LM-head kernel imports and validation for the v4-flash kernel directory.

Poem

A bunny hops where kernels gleam,
Through v4-flash flows the stream.
Paths align and modules load,
Along the freshly marked code road.
=^.^=/**

🚥 Pre-merge checks | ✅ 5

✅ Passed checks (5 passed)

Check name	Status	Explanation
Title check	✅ Passed	The title clearly summarizes the kernel directory rename from v4 to v4-flash in DeepSeek serving.
Description check	✅ Passed	The description directly explains the DeepSeek kernel path rename and the matching code updates.
Docstring Coverage	✅ Passed	No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check.
Linked Issues check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check	✅ Passed	Check skipped because no linked issues were found for this pull request.

---

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.}

[coderabbitai]

Actionable comments posted: 1

🤖 Prompt for all review comments with AI agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

Inline comments:
In `@pypto_serving/model/deepseek/npu_executor.py`:
- Around line 131-139: Update the kernel-directory lookup around the candidate
path in the executor initialization flow to remain compatible with checkouts
containing the existing models/deepseek/v4 location, or ensure this change is
gated and released atomically with the pypto-lib rename and pin update. Preserve
v4-flash resolution for updated checkouts while preventing premature
FileNotFoundError failures for older compatible layouts.

🪄 Autofix (Beta)

Fix all unresolved CodeRabbit comments on this PR:

• Push a commit to this branch (recommended)
• Create a new PR with the fixes

---

ℹ️ Review info

⚙️ Run configuration

Configuration used: Organization UI

Review profile: CHILL

Plan: Pro

Run ID: 7f7b01db-1146-477d-bd9e-9d00ffe5e7ee

📥 Commits

Reviewing files that changed from the base of the PR and between 8773989 and c10942d.

📒 Files selected for processing (1)

• pypto_serving/model/deepseek/npu_executor.py