refactor: migrate pl.at(optimization=) to optimizations=[pl.auto_chunk]

[lyfne123]

Summary

pypto#1504 removed the deprecated pl.at(optimization=, split=) kwargs and the chunked_loop_optimizer sentinel. This migrates every callsite in pypto-lib to the supported optimizations=[pl.auto_chunk] form and refreshes stale comments.

• 18 files: examples + qwen3/deepseek/kimi/milm kernels
• No split= usage existed, so all become a plain optimizations=[pl.auto_chunk]
• pl.auto_chunk is itself deprecation-warned but still functional; kept to keep examples runnable

Test plan

• lint: English-only + headers pass
• golden harness on examples/models compiles & runs

[coderabbitai]

Warning

Review limit reached

@lyfne123, we couldn't start this review because you've reached your PR review rate limit.

More reviews will be available in 50 minutes and 19 seconds. Learn how PR review limits work.

Your organization has run out of usage credits. Purchase more in the billing tab.

⌛ How to resolve this issue?

After more reviews become available, a review can be triggered using the @coderabbitai review command as a PR comment. Alternatively, push new commits to this PR.

We recommend that you space out your commits to avoid hitting the rate limit.

🚦 How do rate limits work?

CodeRabbit enforces hourly rate limits for each developer per organization.

Our paid plans include higher PR review limits than trial, open-source, and free plans. In all cases, reviews become available again over time. During sustained high-volume PR review activity, CodeRabbit may temporarily slow when the next review becomes available.

Please see our Fair Usage Limits Policy for further information.

ℹ️ Review info

⚙️ Run configuration

Configuration used: Organization UI

Review profile: CHILL

Plan: Pro

Run ID: e33f0916-ef65-4cd1-ab28-410ab04ddf9c

📥 Commits

Reviewing files that changed from the base of the PR and between 09a33f0 and 5a2b095.

📒 Files selected for processing (15)

• examples/advanced/gemm_eltwise.py
• examples/beginner/hello_world.py
• examples/beginner/matmul.py
• examples/intermediate/gemm.py
• examples/intermediate/layer_norm.py
• examples/intermediate/rope.py
• examples/intermediate/softmax.py
• models/deepseek/v3_2/deepseek_v3_2_prefill_front_draft.py
• models/deepseek/v4/decode_indexer.py
• models/deepseek/v4/hc_post.py
• models/deepseek/v4/qkv_proj_rope.py
• models/kimi/kimi_k2_decode_draft.py
• models/milm/milm_decode_draft.py
• models/qwen3/14b/qwen3_14b_l3_generate.py
• models/qwen3/32b/qwen3_32b_prefill_draft.py

📝 Walkthrough

Walkthrough

This PR systematically replaces PyPTO's singular pl.at(..., optimization=pl.chunked_loop_optimizer) parameter with a list-based pl.at(..., optimizations=[pl.auto_chunk]) parameter across all examples and production models, unifying the loop optimization hint mechanism from one strategy to another.

Changes

Loop optimization hint unification

Layer / File(s)	Summary
Documentation and docstring updates examples/advanced/gemm_eltwise.py	Module-level comments and function docstrings describing the optimizer configuration are updated to reference auto_chunk instead of chunked_loop_optimizer.
Beginner and intermediate example programs examples/beginner/hello_world.py, examples/beginner/matmul.py, examples/intermediate/gemm.py, examples/intermediate/layer_norm.py, examples/intermediate/rope.py, examples/intermediate/softmax.py	Six example programs update their core-group pl.at contexts to use the new optimizer hint format via single-line parameter changes.
Single-region model kernel updates models/deepseek/v3_2/deepseek_v3_2_prefill_front_draft.py, models/deepseek/v4/decode_attention_hca.py, models/deepseek/v4/hc_post.py, models/kimi/kimi_k2_decode_draft.py, models/milm/milm_decode_draft.py	Deepseek v3_2, v4 (hca, hc_post), Kimi, and MiLM decode programs update their individual core-group scheduling regions via straightforward one-line parameter changes.
Multi-region models with updated comments models/deepseek/v4/decode_attention_swa.py, models/deepseek/v4/qkv_proj_rope.py, models/deepseek/v4/decode_indexer.py	Deepseek v4 attention SWA (three KV cache and assembly regions), qkv_proj_rope (RMSNorm partial-sum paths), and decode_indexer (score_quant path) each switch multiple pl.at contexts and update related inline documentation to reflect auto_chunk usage.
Large-scale Qwen3 model refactoring models/qwen3/14b/qwen3_14b_l3_generate.py, models/qwen3/32b/qwen3_32b_prefill_draft.py	Qwen3 14B and 32B models update all core-group pl.at directives across prefill and decode paths (Q/K/V projection, normalization, padding, RoPE, attention matmul/softmax, and MLP stages) to use the new optimizations=[pl.auto_chunk] format.

Estimated code review effort

🎯 2 (Simple) | ⏱️ ~12 minutes

Possibly related PRs

• hw-native-sys/pypto-lib#332: Both PRs modify models/deepseek/v4/qkv_proj_rope.py by replacing pl.at(..., optimization=pl.chunked_loop_optimizer) hints—this PR switches to optimizations=[pl.auto_chunk] while the related PR refactors the qkv_proj_rope scope structure.
• hw-native-sys/pypto-lib#350: Both PRs modify models/deepseek/v4/decode_indexer.py in the score_quant path—this PR updates the CORE_GROUP loop scheduling hint to pl.auto_chunk while the related PR refactors the scoring computation logic.
• hw-native-sys/pypto-lib#276: Both PRs update Qwen3-14B pl.at directives across the same core-group regions in prefill and decode paths; this PR switches the optimizer strategy while the related PR adds name_hint labeling.

Poem

🐰 With optimizer hints now crystalline clear,
From chunked loops to auto-chunks we steer,
Across all examples and models so grand,
A unified PyPTO, refactored so planned! ✨

🚥 Pre-merge checks | ✅ 4 | ❌ 1

❌ Failed checks (1 warning)

Check name	Status	Explanation	Resolution
Docstring Coverage	⚠️ Warning	Docstring coverage is 20.69% which is insufficient. The required threshold is 80.00%.	Write docstrings for the functions missing them to satisfy the coverage threshold.

✅ Passed checks (4 passed)

Check name	Status	Explanation
Title check	✅ Passed	The title clearly and concisely describes the main change: migrating deprecated pl.at(optimization=) syntax to the new optimizations=[pl.auto_chunk] form across the codebase.
Description check	✅ Passed	The description is directly related to the changeset, explaining the migration rationale, scope (18 files), technical details, and test plan.
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.

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

---

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.}

[gemini-code-assist]

Code Review

This pull request performs a widespread refactor to replace the chunked_loop_optimizer with auto_chunk across multiple example scripts and model implementations. The changes primarily involve updating the pl.at context manager to use the optimizations list parameter instead of the single optimization parameter, along with corresponding updates to docstrings and comments. I have no feedback to provide as there were no review comments.