Reuse Cua-S1 image preprocessing and vision features per request

[Levius-Fubuki]

Purpose

Reuse preprocessing and adapted vision features across questions within one request while preserving question-specific language execution and the single-question reference path.

Stacked on #12. Incremental runtime comparison. The PR diff remains limited to runtime Python source; validation helpers, tests and generated evidence are preserved separately. This optimizes the Python multimodal worker, not the native Rust/CUDA worker.

Test Plan

System1-Omni Version / Commit: 1a339e1b63631271c0a180cb3c848d8698d61bfa.

Fresh validation started 2026-09-30 on RTX 4090 24 GiB, driver 595.71.05, Torch 2.14.0+cu130, Transformers 5.17.0 and PEFT 0.21.0; BF16 base with unmerged adapter, without FLA/causal-conv1d.

• Validation checkout and pinned setup: PYTHONPATH=src:recipe/cua_s1 python -m pytest tests/cua_s1 -q.
• Exact GPU commands and exit statuses, using upstream-verified weights and a fresh output directory.
• Runtime/validation source mapping: git diff <runtime> <validation> -- src is empty. Source hashes and numerical/performance reports are independently verified.

Test Result

• CPU suite: 115 passed in 17.51s.
• Fresh full-checkpoint GPU validation passes 13 synthetic cases with exact prepared tensors, language embeddings/3D positions and complete responses. Cases include PNG/JPEG, changed images, structured/non-ASCII candidates, distinct questions and reordered questions.
• Runtime source is unchanged from the maintainer-reviewed head.

Full evidence and tradeoffs · Raw primary report · Independent verification summary.

This run validates correctness on synthetic inputs; it makes no new latency or throughput claim.

Self-review

Agent-assisted full-diff review and validation limits found no actionable correctness or architecture defect. Runtime scope, commands and claims were checked against the recorded sources/results. The contributor remains responsible for understanding the changes; this does not replace maintainer review.

[Copilot]

Copilot was unable to review this pull request because the user who requested the review has reached their quota limit.

[hsliuustc0106]

Reviewed original snapshot 95713a5 and the follow-up diff through 642d14c754785de5f2c2768fe537cef1d4d690ab. No new actionable findings.

The follow-up fixes the inherited extreme-aspect-ratio validation and lock-cleanup test race from #12. Reviewed the protocol/test changes and documentation updates. Protocol and HTTP tests on this head: 61 passed in 14.73s. No accelerator execution was performed.

Validation and scope of the original snapshot review:

No new findings in image-reuse delta. 15 passed, 1 skipped; tensor reuse test later passes on #18 in torch environment.

[hsliuustc0106]

are you serious? 70K+ LoC

[hsliuustc0106]

Review of 1a339e1b6363 (incremental stack changes).

Reviewed the request-local preprocessing/image-feature reuse, question-specific embeddings and positions, and token-limit handling. No actionable correctness defect found in the inspected paths. This is a reasonable optimization for the existing Python multimodal reference worker; I recommend accepting it.

Validation: 115 archived CPU tests passed against this head's runtime source. No full-checkpoint GPU parity or new latency measurement was performed. Local environment used Torch 2.13 and Transformers 5.14.1, rather than the recipe's Torch 2.14 / Transformers 5.17 pins.

The validation suites were retrieved from the immutable archived revisions linked in the PR descriptions; they are not retained in the current PR diffs.

[Levius-Fubuki]

@hsliuustc0106 Following up on your recommendation to accept 1a339e1b6363: the runtime head is unchanged, and full-checkpoint validation on the recipe-pinned RTX 4090 stack is now available. All 13 synthetic cases have exact prepared-tensor, language-input and complete-response parity. The archived CPU suite reports 115 passes.

Could you approve/merge #17 first, followed by #18? Both remain conflict-free against current main (566dec1); a merge-tree check confirms the changed runtime file is identical to the validated source.

Immutable GPU evidence and limitations · raw #17 report. The evidence/source verifier was rerun successfully on October 1; this follow-up adds no new performance claim.