Spaces:
Sleeping
Sleeping
| """Precompute E5 embeddings for the seed profiles. | |
| Run after editing data/seed_profiles.json: | |
| .venv/bin/python scripts/seed.py | |
| """ | |
| from __future__ import annotations | |
| import sys | |
| from pathlib import Path | |
| import numpy as np | |
| from dotenv import load_dotenv | |
| sys.path.insert(0, str(Path(__file__).resolve().parent.parent)) | |
| load_dotenv() | |
| from core import matching, storage # noqa: E402 | |
| def main() -> None: | |
| profiles = storage.seed_profiles() | |
| queries = matching.embed([matching.query_text(p) for p in profiles]) | |
| passages = matching.embed([matching.passage_text(p) for p in profiles]) | |
| np.savez_compressed( | |
| matching.EMBED_FILE, | |
| ids=np.array([p.id for p in profiles]), | |
| query=queries.astype(np.float32), | |
| passage=passages.astype(np.float32), | |
| ) | |
| print(f"Saved {len(profiles)} profiles x {queries.shape[1]} dims to {matching.EMBED_FILE}") | |
| matching._seed_vectors.cache_clear() | |
| for pid in ("f-layla", "i-omar", "t-sara"): | |
| viewer = storage.get_profile(pid) | |
| ranked = matching.rank(viewer, k=5) | |
| print(f"\n{viewer.display_name}:") | |
| for p, s in ranked: | |
| print(f" {s:3d} {p.display_name:22s} {p.headline[:60]}") | |
| if __name__ == "__main__": | |
| main() | |