Spaces:

miyuiu
/

microbe-model

Running

App Files Files Community

microbe-model / scripts

98.7 kB

Ctrl+K

2 contributors

History: 22 commits

Miyu Horiuchi

Add unified strain catalog (100K rows w/ provenance) + selective weak supervision for pH

4c18dfd 21 days ago

01_fetch_bacdive.py

2.59 kB
Rewrite BacDive client for v2 public API (no auth required) about 1 month ago
02_fetch_and_featurize.py

3.1 kB
Streaming fetch+featurize pipeline + 6× pyrodigal speedup + GCA version resolution 30 days ago
03_train_baseline.py

4.53 kB
Add MediaDive-derived features (medium pH, NaCl, n_media) — all 4 targets improve 21 days ago
04_eval.py

947 Bytes
Eval report enhancements: TL;DR + per-strain predictions + per-family error 30 days ago
05_overnight_summary.py

7.45 kB
Final cleanup: sync OVERNIGHT_SUMMARY.md + fix size display for small files 29 days ago
06_fetch_gtdb_candidates.py

5.19 kB
Fix GTDB column names + accession resolution for v226 metadata schema 29 days ago
07_predict_uncultured.py

5.99 kB
Phase C scaffolding: GTDB candidate selection + uncultured prediction 29 days ago
08_extract_strain_media.py

1.31 kB
Phase E scaffolding: MediaDive integration + strain↔medium links 29 days ago
09_fetch_media_recipes.py

2.89 kB
Phase E scaffolding: MediaDive integration + strain↔medium links 29 days ago
10_train_media_recommender.py

2.72 kB
Phase E #2: scripts/recommend.py — single-genome → ranked media + phenotype CLI 29 days ago
11_extract_embeddings.py

5.58 kB
v2 scaffolding: ESM-2 embedding extraction + GPU runner doc 29 days ago
12_train_with_embeddings.py

2.16 kB
v2 scaffolding: ESM-2 embedding extraction + GPU runner doc 29 days ago
13_compare_v1_v2.py

3.28 kB
v2 results: ESM-2 t6 (8M, 20-protein sample) loses to v1 hand-crafted features on all 4 phenotype targets 27 days ago
14_train_combined.py

3.72 kB
Expand training corpus to 46K strains: species-name → NCBI genome + isolation features 21 days ago
15_train_phenotype_heads.py

4.63 kB
Add unified strain catalog (100K rows w/ provenance) + selective weak supervision for pH 21 days ago
16_score_uncultured_media.py

3.44 kB
UI prep: pre-train phenotype heads + pre-score uncultured media 27 days ago
17_reextract_phenotypes.py

2.25 kB
Fix _derive_salt to pick optimum entries — salt MAE 2.52 → 2.17 (-13.7%) 21 days ago
18_resolve_species_to_genome.py

4.22 kB
Expand training corpus to 46K strains: species-name → NCBI genome + isolation features 21 days ago
19_featurize_resolved.py

4.18 kB
Expand training corpus to 46K strains: species-name → NCBI genome + isolation features 21 days ago
20_build_mediadive_features.py

3.51 kB
Add MediaDive-derived features (medium pH, NaCl, n_media) — all 4 targets improve 21 days ago
21_build_strain_catalog.py

4.32 kB
Add unified strain catalog (100K rows w/ provenance) + selective weak supervision for pH 21 days ago
22_train_with_weak_labels.py

3.34 kB
Add unified strain catalog (100K rows w/ provenance) + selective weak supervision for pH 21 days ago
23_weak_label_apples_to_apples.py

5.19 kB
Add unified strain catalog (100K rows w/ provenance) + selective weak supervision for pH 21 days ago
recommend.py

11.2 kB
UI prep: pre-train phenotype heads + pre-score uncultured media 27 days ago
run_train_and_eval.sh

1.06 kB
Harden post-featurize chain: each phase runs even if previous fails 30 days ago