Skip to content

fix(fish-speech): import relocated source - #11580

Open
localai-org-maint-bot wants to merge 1 commit into
masterfrom
cron/fix-11579-fish-speech-import-20260818
Open

fix(fish-speech): import relocated source#11580
localai-org-maint-bot wants to merge 1 commit into
masterfrom
cron/fix-11579-fish-speech-import-20260818

Conversation

@localai-org-maint-bot

Copy link
Copy Markdown
Collaborator

Description

This PR fixes #11579.

The fish-speech editable install records its build-stage source path. That path no longer exists after LocalAI unpacks the scratch image under the runtime backend directory, so imports such as fish_speech.inference_engine fail.

The launcher now prepends the packaged, runtime-relative fish-speech-src directory to PYTHONPATH. Existing PYTHONPATH entries remain available, and CPU, CUDA, and ROCm dependency selection is unchanged.

A relocation regression test builds a minimal packaged backend layout and imports fish_speech.inference_engine through the real launcher.

Notes for Reviewers

Verified with:

  • make -C backend/python/fish-speech test-source-preparation test-runtime-import
  • bash -n backend/python/fish-speech/{install.sh,package.sh,prepare-source.sh,prepare-source_test.sh,run.sh,run_test.sh,test.sh}
  • git diff --check HEAD^

A full backend image build was not run because it downloads the fish-speech model stack. Backend CI runs the regression through the normal fish-speech build target.

Signed commits

  • Yes, I signed my commits.
  • Documentation updated (docs/content/) for user-facing changes, or not applicable

The editable install records the backend build path, which does not exist after LocalAI relocates the packaged backend. Add the runtime source directory to PYTHONPATH so inference modules remain importable.

Assisted-by: Codex:gpt-5
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.

Fish Speech backend using ROCM doesn't work

1 participant