CLI tool, valdo.pipeline#45
Open
minhuanli wants to merge 20 commits into
Open
Conversation
Converts the pipeline.ipynb workflow into 9 independently-runnable CLI stages (standardize, reindex, scale, preprocess, train, reconstruct, rescale, add_phases_and_blobs, tag_blobs). Each stage is configured via a YAML or JSON file; `valdo.pipeline init <stage>` prints a commented template. Includes docs/pipeline_cli.md with a full tutorial, stage reference, and guidance on when the reindex stage can be skipped. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Generates CC difference histogram and scatter plot after reindexing. If reindex_record.pkl already exists in output_folder, reindexing is skipped and only the plots are regenerated. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
…ariant Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Generates end_corr histogram and start_LS vs end_LS scatter after scaling. If scaling_metrics.pkl already exists, scaling is skipped and only plots are regenerated. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- Add PTP1B_pipeline/PIPELINE_RUN.md documenting steps 1-4 - Add configs for reindex, scale, preprocess, and train stages - Add scaled_filtered_files.txt (1,651 paths after Rf filtering) - Fix VAE training defaults to match notebook: stdof=128, activation=relu, ml_recon=true, sigdF_pct=95.0, extrapolate_factors=[2,4,6,8,16], cutoff=3.5, radius_in_A=4.0 - Add activation config field to train stage (relu/tanh/sigmoid) - Add training loss curve plot (vae_loss_curves.png) saved after training Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- Drop all-NaN input rows in run_train (dataset 0110_1 had all-NaN F-obs-scaled due to silent scaling divergence) - Add logvar clamping hook [-10,10] on encoder to prevent KL overflow - Add gradient clipping (max_norm=1.0) as additional safeguard - Add post-scaling check in run_scale that warns about all-NaN output files - Remove 0110_1.mtz from scaled_filtered_files.txt (now 1,650 datasets) - Fix train default activation to tanh (matches notebook) - Add vae_loss_curves.png from clean 1,650-sample training run - Update PIPELINE_RUN.md with step 5, NaN debugging notes, correct counts Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- Add configs/config_reconstruct.yaml and config_rescale.yaml - Update PIPELINE_RUN.md with steps 6 and 7 - Reconstruct: recons.npy shape (1650, 77821), all finite - Rescale: 1,650 MTZ files with recons and diff columns Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- run_tag_blobs: add bound column from optional bound_models_folder, create temp flat PDB dir for tag_lig_blobs, skip ligand tagging if bound_models_folder is omitted (bound=0, ligand=0) - config.py: make bound_models_folder optional in tag_blobs schema - config_tag_blobs.yaml: PTP1B config with Cys215 focal residue - plot_auc.py: standalone script to plot ROC/AUC from filtered blob stats - PIPELINE_RUN.md: document steps 8 and 9 with output counts (9318 blobs, 8316 filtered, AUC=0.9399) Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- PIPELINE_RUN.md: full rewrite as a how-to guide — Prerequisites section, per-step explanations, Troubleshooting callouts, apo refinement instructions, Step 10 evaluation section; inline Python replaced with script references - parse_refine_logs.py: parse PHENIX log files into refine_summary.csv - filter_datasets.py: filter scaled MTZ files by R-free, CC, and all-NaN F-obs-scaled; resolves reindexing ambiguity; writes filtered file list Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
…ode quality
pipeline.py bug fixes:
- run_scale: Scaler_pool does not support when_opt; fall back to single-process
Scaler whenever when_opt != "never" (pool silently ignored it before)
- run_scale: restore skip-if-exists with --force override (was accidentally removed)
- run_add_phases_and_blobs: drop NaN rows after extrapolation (mirrors pipeline_doeke.py)
- run_train: add torch.manual_seed for reproducible weight initialisation
pipeline.py new features:
- --force flag to override skip-if-exists on any stage
- Skip-if-exists guards added to all stages (reindex already had one)
pipeline.py code quality:
- Move all imports to function top (no imports inside loops or mid-function)
- Remove import glob as _glob alias (module-level glob already imported)
- Move matplotlib.use("Agg") before pyplot import in run_train
- Remove duplicate import tempfile in run_tag_blobs
- Update _TEMPLATES defaults to match Doeke's settings (when_opt, random_seed, activation)
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
… results
- Integrate filter_datasets.py as valdo.pipeline filter stage (remove standalone script)
- Add AUC-vs-N-blobs plot to plot_auc.py (sort by peakz, evaluate at N=500..6000)
- Fix bound model naming: standardize to {id}.pdb for tag_lig_blobs lookup
- Add configs for filter and add_phases_and_blobs stages
- Update PIPELINE_RUN.md with new run counts, AUC=0.9748, bound model naming guide,
and stale-file cleanup warnings for reruns
- Add CLAUDE.md to .gitignore
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- valdo/helper.py: fix find_phase_file to prefer *_001.mtz over *_data.mtz when multiple glob matches exist, fixing silent phasing failures for ~833 datasets - valdo/commandline/config.py, pipeline.py: filter stage additions - PTP1B_pipeline/compute_valdo_metrics.py: apo peak and heavy atom WDF metric script - notebooks/vae_metric_apo_peak_value.py, vae_metric_heavy_atom_peak_value.py: converted from Doeke's notebooks for reference Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- compute_valdo_metrics.py: add --recons-phased argparse, expose compute_metrics() for import, move mapping txt to local path - collect_ablation_metrics.py: loop all ablation settings, print comparison table, write ablation/ablation_metrics.csv - ligand_cif_to_dataset_mapping.txt: copy from notebooks/ so PTP1B_pipeline/ is self-contained Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
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.
No description provided.