| |
| 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.dataset import build_examples |
| from spec_rag.faiss_index import load_index |
| from spec_rag.io import load_embeddings, load_smiles, save_jsonl |
| from spec_rag.retrieval import search_index |
|
|
|
|
| def parse_args() -> argparse.Namespace: |
| parser = argparse.ArgumentParser(description="Build RAG dataset for MolT5/Llama.") |
| parser.add_argument("--index-path", required=True) |
| parser.add_argument("--smiles-path", required=True) |
| parser.add_argument("--spec-embeddings", required=True) |
| parser.add_argument("--ground-truth-smiles", default=None) |
| parser.add_argument("--ground-truth-mgf", default=None) |
| parser.add_argument("--out-jsonl", required=True) |
| parser.add_argument("--k", type=int, default=10) |
| parser.add_argument("--format", choices=["molt5", "chat"], default="molt5") |
| parser.add_argument("--spectrum-token", default="<Spectrum_Token>") |
| return parser.parse_args() |
|
|
|
|
| def _load_smiles_from_mgf(mgf_path: str) -> list[str]: |
| try: |
| from pyteomics import mgf |
| except Exception as e: |
| raise ImportError(f"pyteomics is required to read MGF: {e}") |
| smiles = [] |
| with mgf.MGF(mgf_path) as reader: |
| for spec in reader: |
| params = spec.get("params", {}) |
| smi = params.get("SMILES") or params.get("smiles") or "" |
| smiles.append(str(smi).strip()) |
| return smiles |
|
|
|
|
| def main() -> None: |
| args = parse_args() |
| if bool(args.ground_truth_smiles) == bool(args.ground_truth_mgf): |
| raise ValueError("Provide exactly one of --ground-truth-smiles or --ground-truth-mgf.") |
| index = load_index(args.index_path) |
| smiles = load_smiles(args.smiles_path) |
| spec_embeddings = load_embeddings(args.spec_embeddings) |
| if args.ground_truth_mgf: |
| gt_smiles = _load_smiles_from_mgf(args.ground_truth_mgf) |
| else: |
| gt_smiles = load_smiles(args.ground_truth_smiles) |
|
|
| scores, indices = search_index(index, spec_embeddings, args.k) |
| contexts = [] |
| for row in indices: |
| contexts.append([smiles[j] if j >= 0 else "" for j in row.tolist()]) |
|
|
| examples = build_examples( |
| spectrum_ids=list(range(len(contexts))), |
| spectrum_token=args.spectrum_token, |
| contexts=contexts, |
| targets=gt_smiles, |
| ) |
| if args.format == "chat": |
| rows = [ex.as_chat() for ex in examples] |
| else: |
| rows = [ex.as_molt5() for ex in examples] |
|
|
| save_jsonl(args.out_jsonl, rows) |
| print(f"Wrote {len(rows)} examples to {args.out_jsonl}") |
|
|
|
|
| if __name__ == "__main__": |
| main() |
|
|
