Paper-matrix harness, Tunix backend, GCP scaffolding#1
Merged
Conversation
- scripts/run_paper_matrix.py drives B0-B5 x tasks x seeds with fairness fingerprint guard via cts.train.fairness.assert_consistent. - scripts/aggregate_results.py builds the Holm-corrected paper table using the existing cts.eval.stats primitives. - scripts/gen_all_teacher_data.sh fans out the teacher CLI across (gsm8k, apps) x (train, val, test). - src/cts/backends/tunix_adapter.py: M8 wiring (table-driven trainer resolution, real Orbax checkpoint/restore). Construction-time guard raises a clean ImportError without the tunix dep installed. - src/cts/train/main.py: route backend=tunix through the adapter; add --gcs-bucket flag that lazily imports deploy/gcs_sync.upload. - deploy/: Dockerfile.cpu, Dockerfile.tpu, cloudbuild.yaml, launch_tpu.sh (DRY_RUN=1 honored), launch_vertex.py, gcs_sync.py, README.md. Files in place; nothing executed yet. - Makefile, .pre-commit-config.yaml, .github/workflows/ci.yml for local dev QoL and CI. - pyproject.toml: gcp extra (google-cloud-storage, google-cloud-aiplatform). - .gitignore: build artifacts, IDE, coverage, GCP service-account keys.
- src/cts/train/loop.py: bind `step` as default arg to silence B023 (loop variable not bound in nested function); preserves behavior. - src/cts/rl/verifier.py, src/cts/eval/stats.py, src/cts/losses/sinkhorn.py: remove dead variables, replace try/except/pass with contextlib.suppress. - tests/unit/test_bug_fixes.py: rewrite
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
scripts/run_paper_matrix.pyruns the paper's B0–B5 × tasks × seeds matrix with a fairness-fingerprint guard;scripts/aggregate_results.pybuilds the Holm-corrected results table;scripts/gen_all_teacher_data.shfans out the teacher CLI over(gsm8k, apps) × (train, val, test).src/cts/backends/tunix_adapter.pynow has table-driven trainer resolution + real Orbax checkpoint/restore. CleanImportErrorwithout thetunixdep.cts.train.mainroutesbackend=tunixthrough it.deploy/{Dockerfile.cpu, Dockerfile.tpu, cloudbuild.yaml, launch_tpu.sh, launch_vertex.py, gcs_sync.py, README.md}.--gcs-bucketflag incts.train.mainlazily wiresgcs_sync.upload.Makefile,.pre-commit-config.yaml,.github/workflows/ci.yml.gcpextra added topyproject.toml..gitignoreexpanded (build, IDE, coverage, GCP service-account keys).