Remove SAM2 mask generation and SA-V support

[parkjinman98]

Summary

Removes SAM2HieraLarge promptable mask generation and SA-V dataset support from main. The full implementation is preserved on the jm/sam branch (a snapshot of main at f8d8a07).

Removed

• mblt_vision.mask_generation (SAM2HieraLarge, prompt-encoder port, graph contracts) and models/SAM2HieraLarge.yaml
• mask_generation from VISION_TASKS and the top-level exports
• SA-V support: datasets/sa-v.yaml, CustomSAV, organize_sav, readiness checks, eval_sav, benchmark/organize_sav.py
• CLI --point, encoder/decoder/prompt-weights path overrides, and SA-V val options (predict.py and _vision.py now match their pre-SAM versions)
• The mask_generation branch of Results
• SAM/SA-V tests and docs (root and Vision READMEs, benchmark/README.md, AGENTS.md, mblt-vision skill)

Kept (generic improvements that landed with the SAM PR)

• wrapper.download_hub_artifact
• _find_existing_source never returning the organized cache
• Letterbox/postprocess fixes, MANIFEST.in, sample assets/

Test plan

• No remaining SAM/SA-V references (grep)
• ruff check --select F . clean, git diff --check clean
• list_tasks() returns 8 tasks
• pytest: 708 passed, 5 failed. All 5 are tests/test_onnx_yolo.py Hugging Face download failures (no network in the test environment); not rerun with network access

Sets __version__ to 0.0.5: 0.0.5 was never released, so this removal ships as 0.0.5.

🤖 Generated with Claude Code

[chatgpt-codex-connector]

You have reached your Codex usage limits for code reviews. You can see your limits in the Codex usage dashboard.

[parkjinman98]

@codex review.

[chatgpt-codex-connector]

Codex Review Summary

This comment shows the latest Codex review activity on this pull request.

Review	Status	Commit	Review trigger
📝 Code Review	✅ Completed 2026-09-29T01:52:24.146209Z	4a6f8c1	Manual request
🔒 Security Review	✅ Completed 2026-09-29T01:52:04.571538Z	4a6f8c1	Manual request

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" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.

[chatgpt-codex-connector]

🛡️ Codex Security Review · Automatically triggered

Security review completed. No security issues were found in this pull request.

Reviewed commit: 4a6f8c1c66

View security finding report

Only the user who started this review can view the report in Codex.

ℹ️ About Codex security reviews in GitHub

This is an experimental Codex feature. Security reviews are triggered when:

• You comment "@codex security review"
• A regular code review gets triggered (for example, "@codex review" or when a PR is opened), and you’re opted in so security review runs alongside code review

Once complete, Codex will leave suggestions, or a comment if no findings are found.

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 4a6f8c1c66

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".

[parkjinman98]

@mobilint-review Please re-review at 0a0dccc.

Both earlier findings are addressed and resolved:

• NYU Depth rule dropped from AGENTS.md (P2): fixed. The rule is restored verbatim.
• Deprecate mask generation before removal (P1): closed as intentional. Please do not raise it again. The maintainer decided to remove SAM2 from main, and the code is kept on the jm/sam branch. The AGENTS.md deprecation rule covers established Model Zoo Vision behavior, and SAM2 was never part of mblt_model_zoo.vision. Because it did ship in 0.0.3/0.0.4, the removal is now documented as a breaking change in mblt_vision/README.md (pin <0.0.5). A stub or shim is deliberately not added.

Guide for this review:

• Scope: this PR only removes code. Every added line either removes SAM/SA-V wiring or documents the removal. A complete, accurate removal is the goal, not new behavior.
• Most useful checks:
  • Leftover references to mask_generation, SAM2HieraLarge, sa-v, eval_sav, CustomSAV, organize_sav, or --point / encoder / decoder / prompt-weights CLI options.
  • Imports or exports that now point at deleted modules.
  • Non-SAM rules or docs that were deleted by accident, like the NYU one.
• Kept on purpose, do not flag:
  • wrapper.download_hub_artifact, a generic helper that MBLT_Engine uses.
  • The _find_existing_source guard that never returns the organized cache (it still protects nyu-depth).
  • The letterbox/postprocess fixes, MANIFEST.in, and assets/.
  • The *.tar/*.tgz gitignore entries.
• Version: __version__ = "0.0.5" is intentional. 0.0.5 was never released, so going from 0.0.6 back to 0.0.5 is not a downgrade.
• Known test failures: 5 failures in tests/test_onnx_yolo.py are Hugging Face download failures in a sandbox without network. They are unrelated to this PR.