import argparse, os, joblib, numpy as np, pandas as pd, torch from features import * torch.set_num_threads(2) here = os.path.dirname(os.path.abspath(__file__)) ap = argparse.ArgumentParser() ap.add_argument("--input", required=True); ap.add_argument("--output", required=True) args = ap.parse_args() df = pd.read_parquet(args.input).reset_index(drop=True) df["sequence"] = df["sequence"].fillna("").astype(str) df["smiles"] = df["smiles"].fillna("").astype(str) ref = pd.read_parquet(os.path.join(here, "ref.parquet")) z = np.load(os.path.join(here, "ref_emb.npz"), allow_pickle=True) emb = {s: v for s, v in zip(z["seqs"], z["vecs"])} mu = z["mu"] emb.update(embed_unique(df["sequence"], os.path.join(here, "esm_local"))) X = np.hstack([base_X(df, emb), label_feats(ref, df, emb, mu)]) models = joblib.load(os.path.join(here, "models.pkl")) preds = [np.mean([m.predict(X) for m in models[t]], axis=0) for t in range(2)] out = pd.DataFrame({"id": df["id"], "log10_kcat": np.nan_to_num(np.clip(preds[0], -1e5, 1e5)), "log10_km": np.nan_to_num(np.clip(preds[1], -1e5, 1e5))}) os.makedirs(os.path.dirname(os.path.abspath(args.output)) or ".", exist_ok=True) out.to_csv(args.output, index=False)