Skip to content

Feature/retrieval optimization - #51

Merged
farshad68 merged 16 commits into
feature/optimizationfrom
feature/retrieval_optimization
Jul 27, 2026
Merged

Feature/retrieval optimization#51
farshad68 merged 16 commits into
feature/optimizationfrom
feature/retrieval_optimization

Conversation

@farshad68

@farshad68 farshad68 commented Jul 14, 2026

Copy link
Copy Markdown
Member

Hey, I've resolved the merge conflicts between the optimization and the retrieval_optimization branch. Could you review the changes and verify your feature functionality remains intact?

Here is a summary of how the conflicts were handled:

Pipeline Execution (Main): I kept your old_loaders.json_loader ingestion logic, but I rejected the commented-out search and generation blocks. The active asyncio search and generation logic from the baseline is retained.

Search Logging & Dependency Injection (search.py):

I kept your profilingLogger additions and the user_config/opt_config dependency injections in search_authorized (critical for the optimizer).

Ingestion ETA (ingestion.py): I completely accepted your progress bar, batch tracking, and perf_counter logic.

Dataset Loading (tools.py): I rejected the custom pandas CSVDataset wrapper because it hardcoded paths. Instead, I kept the native HuggingFace datasets.load_dataset logic, but I added your sep=';' argument to ensure your specific European CSV formats still parse correctly.

SMAC Optimizer Fix (retrieval_optimization.py): Your optimization script was crashing with a RuntimeError because the tight ForbiddenAndConjunction rules caused SMAC to discard all random initial configurations. I added an initial_design injection to force SMAC to evaluate the default config so it can always start successfully.

Please pull the latest branch, run a test optimization, and let me know if the logging or CSV ingestion behaves as you expect.

@farshad68
farshad68 requested review from carowa292 and hannred July 14, 2026 15:05
@carowa292

Copy link
Copy Markdown
Contributor

Hi,

I went through your points and ran the optimization successfully.

Dataset Loading (tools.py):
Adopted your approach — we stay with datasets.load_dataset (no
pandas), but added sep=';' so our semicolon-delimited CSVs parse
correctly.

main.py → debug.py:
I renamed main.py to debug.py. We only use it as a scratch script to
manually test ingestion/search/generate.

old_loaders.py:
Deleted — no longer used anywhere.

pyproject.toml:
Added bert-score and scikit-learn (were missing after
the merge).

Logging:
Works

@farshad68
farshad68 merged commit a4cfa3a into feature/optimization Jul 27, 2026
1 check passed
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.

3 participants