[hephaex]

Summary

docs/vision-model-scope.md (added by #2508) lists the supported non-chat vision families but omits nemotron_parse, which already loads and serves in-tree. This adds its row beside florence2 and refreshes the "Last checked against main" commit.

Related issues

Refs #524 (already closed by #2508; this is a follow-up to that page, not a reopening).

Type of change

• feat — new user-visible feature
• fix — bug fix
• perf — performance improvement (include before/after numbers in the PR body)
• refactor — internal restructuring without behavior change
• chore — build, CI, dependencies, release infrastructure
• docs — documentation only
• test — tests only

Test plan

Documentation only; no Rust or checkpoint code changed.

• cargo fmt --all -- --check — N/A (no Rust changes)
• cargo clippy --workspace --all-targets --features metal,accelerate -- -D warnings — N/A (no Rust changes)
• cargo test --workspace --profile test-fast --features metal,accelerate --no-fail-fast -- --test-threads=1 — N/A (no Rust changes)
• cargo deny check — N/A (no dependency changes)
• Validated with a real checkpoint — N/A (docs only)
• Moves the numbers — N/A
• python3 scripts/ci/check_cross_repo_refs.py docs/vision-model-scope.md passes ("no bare 3+-digit '#NNN' added").
• Every path and symbol in the new row was checked against the tree at e160079e:
  • "nemotron_parse" => ModelType::NemotronParseVLM at src/models/detection.rs:1170
  • src/models/nemotron_parse/, src/loading/vlm_nemotron_parse.rs, src/commands/generate_nemotron_parse.rs, src/server/nemotron_parse_worker.rs
  • tests/nemotron_parse_real_model.rs — its two checkpoint tests are #[ignore]-gated, so the row describes them as gated and does not claim they were run.
  • feat(nemotron_parse): port Nemotron-Parse (C-RADIO ViT-H encoder + pre-norm mBART decoder) on the seq2seq worker #1369 is CLOSED and feat(nemotron_parse): port Nemotron-Parse on the seq2seq worker #2034 is MERGED.
• The only doc change between the old checked commit (9a6ea266) and e160079e is docs: record vision model scope for non-generative families #2508 itself, so the other rows' evidence still holds under the refreshed hash.

Notes for reviewers

• The row says explicitly that nemotron_parse is not one of the families tracking: non-generative vision models present in the VLM reference but out of mlxcel's LLM/VLM-generation scope #524 collected; it is listed under "Already supported" because the page is the lookup for non-chat vision coverage and a reader checking OCR/layout support would otherwise find no answer.
• Suggested labels (not settable from this account): type:docs.

Checklist

• PR title uses a Conventional Commits prefix (feat:, fix:, etc.)
• One logical change per PR (split unrelated changes)
• Updated docs/ if user-facing behavior or supported models changed
• Updated // Used by: ... comments on any shared function I modified — N/A
• No secrets, credentials, or .env files committed

[inureyes]

Hardening review for PR lablup/mlxcel#2525 (refs lablup/mlxcel#524, lablup/mlxcel#1369, lablup/mlxcel#2034).

• Purpose & correctness: Adds the nemotron_parse entry to the Already supported table in docs/vision-model-scope.md and refreshes the checked-against commit hash. Every cited path and symbol was verified against the live tree ("nemotron_parse" => Ok(ModelType::NemotronParseVLM) in src/models/detection.rs:1170, src/models/nemotron_parse/, src/loading/vlm_nemotron_parse.rs, src/commands/generate_nemotron_parse.rs, src/server/nemotron_parse_worker.rs, and #[ignore]-gated tests/nemotron_parse_real_model.rs).
• Security & performance: Documentation-only change; no runtime, security, or performance impact.
• Validation: python3 scripts/ci/check_cross_repo_refs.py docs/vision-model-scope.md and all 13 GitHub Actions CI checks passed.