| |
| from __future__ import annotations |
|
|
| import argparse |
| import sys |
| from pathlib import Path |
|
|
| ROOT = Path(__file__).resolve().parents[1] |
| if str(ROOT) not in sys.path: |
| sys.path.insert(0, str(ROOT)) |
|
|
| from spec_rag.embeddings import SMILESEmbedder |
| from spec_rag.faiss_index import build_hnsw_index, save_index |
| from spec_rag.io import ensure_dir, load_smiles, save_embeddings, save_smiles |
|
|
|
|
| def parse_args() -> argparse.Namespace: |
| parser = argparse.ArgumentParser(description="Build FAISS HNSW index from SMILES.") |
| parser.add_argument("--smiles-path", required=True) |
| parser.add_argument("--model-name", default="seyonec/ChemBERTa-zinc-base-v1") |
| parser.add_argument("--out-dir", required=True) |
| parser.add_argument("--batch-size", type=int, default=64) |
| parser.add_argument("--max-length", type=int, default=256) |
| parser.add_argument("--pooling", choices=["cls", "mean"], default="cls") |
| parser.add_argument("--no-normalize", action="store_true") |
| parser.add_argument("--metric", choices=["cosine", "l2"], default="cosine") |
| parser.add_argument("--m", type=int, default=32) |
| parser.add_argument("--ef-construction", type=int, default=200) |
| parser.add_argument("--ef-search", type=int, default=128) |
| parser.add_argument("--dedupe", action="store_true") |
| parser.add_argument("--device", default="cuda") |
| return parser.parse_args() |
|
|
|
|
| def main() -> None: |
| args = parse_args() |
| if args.device == "cuda": |
| try: |
| import torch |
| except Exception: |
| torch = None |
| if torch is None or not torch.cuda.is_available(): |
| print("CUDA not available; falling back to CPU.") |
| args.device = "cpu" |
| out_dir = ensure_dir(args.out_dir) |
| smiles = load_smiles(args.smiles_path) |
| if args.dedupe: |
| seen = set() |
| smiles = [s for s in smiles if not (s in seen or seen.add(s))] |
|
|
| embedder = SMILESEmbedder( |
| model_name=args.model_name, |
| device=args.device, |
| pooling=args.pooling, |
| batch_size=args.batch_size, |
| max_length=args.max_length, |
| normalize=not args.no_normalize, |
| ) |
| embeddings = embedder.encode(smiles) |
|
|
| save_embeddings(out_dir / "smiles_embeddings.npy", embeddings) |
| save_smiles(out_dir / "smiles.txt", smiles) |
|
|
| index = build_hnsw_index( |
| embeddings, |
| m=args.m, |
| ef_construction=args.ef_construction, |
| ef_search=args.ef_search, |
| metric=args.metric, |
| ) |
| save_index(index, out_dir / "smiles.index") |
| print(f"Saved index and embeddings to {out_dir}") |
|
|
|
|
| if __name__ == "__main__": |
| main() |
|
|
