[LauraGPT]

Final Publication Acceptance

GitHub Release and PyPI 1.4.20 are both published. Both distributions were downloaded from official PyPI and match the accepted GitHub assets by SHA-256 and size:

• funasr-1.4.20-py3-none-any.whl: 1014313 bytes; SHA-256 664a1fbf36bc177a5aec3230b2b1023c160daf544da540bea60c7f3acd6393c3.
• funasr-1.4.20.tar.gz: 1063112 bytes; SHA-256 906f1a496acbc7235b8739cafd487a8e6d47b9785f466c37c445590ad4531a85.

FunASR 1.4.20

Source: f122f6c7bc6176ed307cc1659b2bcc031d54791f / tag v1.4.20.

Fix

Deliver #3819 (integrating #3788): optional speaker imports no longer disappear behind an unrelated NameError. When spk_model is requested and an optional speaker dependency failed to import, the constructor reports an actionable ImportError chained to that original failure. ASR-only construction remains available.

This changes error reporting, not the underlying missing/incompatible dependency on the reporter's Windows machine. It does not claim Windows hardware acceptance, pretrained ASR/diarization quality or resolution of every speaker issue.

The separately released funasr-onnx==0.4.5 has its own tag, packages and verification. This root-package release does not republish it.

Verification Scope

• New candidate release/README checks: 45 cases, 0 skips. Related docs/package/release checks: 104 cases, 2 explicitly recorded skips; no failures or errors.
• Actual installed candidate wheel: 10 fresh-process CPU optional-import cases, including both kwargs/config entry points, missing-module/runtime-initialization errors and healthy controls. Controlled model constructors are not pretrained ASR acceptance. All 439 package files and 13 entrypoints match the accepted source.
• Reused fix(auto): preserve optional speaker dependency errors #3819's five real L20Z FP32 synthetic constructor/import-boundary results by rehashing all 172 evidence entries and binding all four recorded source modules to the installed wheel. Those cases include native reduced SANM/CIF CUDA inference and healthy native CAM++ embeddings, not new GPU execution, BF16, Windows or pretrained-quality validation.
• Exact release-head and merged-main CI were accepted with actual logs/artifacts. The final-tag workflow and the newly downloaded Python wheel/sdist passed byte/metadata/archive/source checks, strict Twine, offline installation, actual import and CLI checks.
• The 10 unchanged runtime archives reuse complete cache bytes freshly hashed against new asset metadata and the existing runtime source release. They were not rebuilt or re-executed. SHA256SUMS covers all 12 payloads. This is not a new runtime-hardware claim.
• The historical fix(auto): preserve optional speaker dependency errors #3819 native BLAS performance diagnostic failure and later private-environment repair remain separate records. No full repository-suite, production-environment upgrade, Nano benchmark or adoption/growth claim is made.

GitHub Release and official PyPI downloads are both accepted; the plan below is retained history.

Historical Pre-Release Plan

Purpose

Deliver the already merged optional speaker dependency diagnostic fix #3819 (integrating #3788) to users of the root PyPI package. Official 1.4.19 still contains the old auto_model.py; all other 438 installed package files already match the accepted main source, before the version bump.

When spk_model is requested after an optional speaker import fails, the constructor now raises an actionable ImportError chained to the original exception. ASR-only construction remains available. This release does not repair the underlying dependency on the reporter's Windows machine or close unresolved reporter issues.

Scope

• Exactly eight version/README/repository-role/test paths, based on 6c20419b0b547e57703e16540d24dd4472abfe6e.
• Verified candidate tree c3d3b1c69dc8e0570c2419ca1df389e6ca7dbdb1, all 4,175 tracked blobs and modes checked.
• No new inference code, dependency change, model download, runtime rebuild, Nano work or shared/production environment change.
• The separately published funasr-onnx==0.4.5 is not republished by this root release.

Verification Before PR

• Version-contract red check: 8 expected failures and 4 historical controls passing, no errors/skips.
• Updated release/README checks: 45 passed.
• Related docs/package/release checks: 102 passed, 2 explicitly skipped because private Chromium/Playwright are unavailable. These are not counted as passed; new exact-head hosted browser/site checks remain mandatory.
• Built wheel/sdist, strict Twine validation, offline target installation, actual installed import/CLI checks: passed. All 439 package files and 13 entrypoints matched accepted source; 23 unchanged source-only model YAML templates remain outside the established wheel contract.
• Actual installed wheel: 10 fresh-process CPU optional-import cases passed, covering kwargs/config, missing-module/runtime-initialization failures and healthy controls. Controlled constructors are not pretrained ASR acceptance.
• Existing fix(auto): preserve optional speaker dependency errors #3819 evidence: all 172 manifest entries rehashed, four source modules matched to the installed wheel, all five real L20Z FP32 synthetic constructor/import-boundary receipts verified. No repeated GPU execution. The historical BLAS performance diagnostic failure and private-environment repair remain separate evidence.

This is not a full repository-suite, pretrained-quality, BF16, Windows hardware, production-upgrade or benchmark/adoption claim. Review of this release was performed by the implementer and is not represented as an independent GitHub approval.

Publication Gates

Normal signed+DCO merge only after current identity/permissions/full protection/rules/head/main/fork/discussion checks and exact-head CI acceptance. Merged-main CI and final-tag workflow must each be accepted. GitHub Python artifacts will be newly downloaded and source/metadata checked; unchanged runtime archives will reuse complete cache bytes only after full SHA/size/archive and current asset verification, disclosed explicitly. Verify 12 payloads plus SHA256SUMS, then publish PyPI once and download both official payloads to compare their actual SHA/size against GitHub. No final publication is claimed by this candidate PR.