Skip to content

docs(qwen3): remove stale docs and dead code#702

Open
Nyvo-io wants to merge 2 commits into
openinfer-project:mainfrom
Nyvo-io:chore/qwen3-doc-dead-code-cleanup-248
Open

docs(qwen3): remove stale docs and dead code#702
Nyvo-io wants to merge 2 commits into
openinfer-project:mainfrom
Nyvo-io:chore/qwen3-doc-dead-code-cleanup-248

Conversation

@Nyvo-io

@Nyvo-io Nyvo-io commented Jul 17, 2026

Copy link
Copy Markdown
Contributor

Summary

  • remove the unused Qwen3 probe_model/ModelInfo path and dead trace constants
  • replace stale Qwen3 crate and tensor-parallel bring-up records with current ownership/runtime documentation
  • move the durable KV pressure lesson into docs/lessons/, delete obsolete model records, and repair indexes/cross-references
  • correct the Qwen3 roadmap so landed real-adapter LoRA and cached-token propagation gates are not listed as open work

Maintainer Follow-up

Commit 2702ad6 aligns the roadmap and documentation index with the regression coverage already present on this PR's base:

  • openinfer-qwen3/tests/lora_golden_gate.rs covers a non-zero PEFT adapter against its golden reference for bs=1, mixed base/LoRA batches, and TP=2; it also rejects a silently inactive adapter
  • openinfer-qwen3/tests/cached_tokens_usage.rs and openinfer-vllm-frontend/src/bridge/tests.rs cover TokenEvent::Scheduled.cached_tokens propagation into frontend prefill and usage statistics, including first-output and terminal-output paths
  • the remaining roadmap gaps now distinguish TP numerical parity, aggregate prefix-cache telemetry, LoRA cache-salt isolation, and YaRN [Feature Request]: Support YaRN (Yet another RoPE extensioN) for context window extension #8

Verification

The current head is 2702ad6; both commits are Nyvo-authored, Nyvo-committed, and DCO-signed.

GitHub CI run 29891075821 passes all 15 reported checks, including:

  • Rust formatting, DCO, attribution, and locked Cargo metadata
  • CPU Clippy and all CPU unit-test jobs
  • simulated frontend E2E
  • Qwen3 CUDA compile and CUDA Clippy (sm_80)

A full review of the complete PR diff against origin/main, including the follow-up, found no qualifying regressions. Removed symbols and document paths have no remaining workspace references, active documentation links remain valid, and git diff --check passes.

Closes #248

@Nyvo-io

Nyvo-io commented Jul 20, 2026

Copy link
Copy Markdown
Contributor Author

Status update: PR #702 remains ready for maintainer review. The current head 8b0ad956 is clean, all reported CI checks are passing, and there are no new review comments or requested code changes. I have kept the scope limited to the Qwen3 documentation/dead-code cleanup and the durable KV-pressure lesson.

Signed-off-by: Nyvo <75425811+Nyvo-io@users.noreply.github.com>
@Nyvo-io
Nyvo-io force-pushed the chore/qwen3-doc-dead-code-cleanup-248 branch from 8b0ad95 to e1f9fb5 Compare July 21, 2026 18:29
@Nyvo-io

Nyvo-io commented Jul 21, 2026

Copy link
Copy Markdown
Contributor Author

Final update after rebasing PR #702 onto origin/main@afd7a64:

  • final head is e1f9fb5, a single Nyvo-authored, Nyvo-committed, DCO-signed commit
  • the only rebase conflict was the openinfer-qwen3/src/lib.rs import block; the resolution preserves the current upstream rustfmt/import layout and all new runtime modules while continuing to remove only the unused ModelInfo/probe_model path
  • the 12-file scope is unchanged, deleted symbols and document paths have no remaining repository references, and the PR is clean/mergeable
  • final CI run 29857427251 passes all 15 checks, including formatting, DCO, attribution, CPU Clippy/tests, simulated frontend E2E, Qwen3 CUDA compile, and Qwen3 CUDA Clippy

Please review when convenient.

@xiaguan xiaguan left a comment

Copy link
Copy Markdown
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This is not fully addressed yet: the refreshed roadmap still lists the real-adapter LoRA gate and cached-token propagation as open, although both already exist with regression coverage on this PR's base.

Signed-off-by: Nyvo <75425811+Nyvo-io@users.noreply.github.com>
@Nyvo-io

Nyvo-io commented Jul 22, 2026

Copy link
Copy Markdown
Contributor Author

Addressed the requested roadmap correction in 2702ad6:

  • the real-adapter LoRA gate is now recorded as landed, with non-zero PEFT golden coverage for bs=1, mixed base/LoRA batches, TP=2, and an inactive-adapter guard
  • cached-token propagation is now recorded as landed from TokenEvent::Scheduled.cached_tokens through frontend prefill/usage statistics, with Qwen3 and bridge regressions
  • the open items are narrowed to TP numerical parity, aggregate prefix-cache telemetry, LoRA cache-salt isolation, and YaRN [Feature Request]: Support YaRN (Yet another RoPE extensioN) for context window extension #8

I reviewed the complete PR diff again after the correction and found no remaining issues. The worktree is clean, the branch is synchronized, and CI run 29891075821 passes all 15 checks, including Qwen3 CUDA compile and CUDA Clippy. Please re-review when convenient.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

qwen3: docs/dead-code cleanup — stale records, deleted-tool references, unused consts

2 participants