Skip to content

fix(ingest): bound parent-chunk pools by the caller's process count - #549

Merged
akhileshh merged 2 commits into
mainfrom
pcgv2-pipeline
Oct 21, 2026
Merged

akhileshh merged 2 commits into
mainfrom
pcgv2-pipeline

Conversation

@akhileshh

Copy link
Copy Markdown
Contributor
  • add_layer passes n_threads down to every pool instead of sizing them from mp.cpu_count(), so the pipeline's PCG_N_PROCESSES is the pod's bound.
  • A caller passing no count still gets node-core pools; 1 stays serial.

akhileshh and others added 2 commits October 21, 2026 21:54
- add_layer passes n_threads down to every pool instead of sizing them from mp.cpu_count(), so the pipeline's PCG_N_PROCESSES is the pod's bound.
- A caller passing no count still gets node-core pools; 1 stays serial.

Co-Authored-By: Claude <noreply@anthropic.com>
- The conda stage runs `conda tos accept` for `pkgs/main` and `pkgs/r` after updating conda, as pcgv3's Dockerfile does; `nodefaults` no longer kept `conda env create` from stopping on `CondaToSNonInteractiveError`.
- `requirements.yml` drops `nodefaults`, so both lines list the same channels.

Co-Authored-By: Claude <noreply@anthropic.com>
@akhileshh
akhileshh merged commit c75ace9 into main Oct 21, 2026
1 check passed
@akhileshh
akhileshh deleted the pcgv2-pipeline branch October 21, 2026 22:46
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.

1 participant