Upload tools/vocab_sweep.py with huggingface_hub

tools/vocab_sweep.py

@@ -68,6 +68,9 @@ def decode_piece(piece: str, byte_decoder: dict[str, int]) -> str | None:
 def main() -> int:
     ap = argparse.ArgumentParser()
     ap.add_argument("--tokenizer", required=True)
+    ap.add_argument("--name", default="muse-glimmer-30b",
+                    help="basename of the sweep shard under "
+                         "pool/vocab_sweep/synthetic/ (one per target model)")
     ap.add_argument("--include-special", action="store_true", default=True,
                     help="also sweep the 2,048 added/special ids")
     ap.add_argument("--no-include-special", dest="include_special", action="store_false")
@@ -155,9 +158,9 @@ def main() -> int:
                                 "function of the tokenizer, not of the corpus"},
         ))
     out = os.path.join(P.POOL_ROOT, "vocab_sweep", "synthetic",
-                       "muse-glimmer-30b.jsonl")
+                       f"{args.name}.jsonl")
     n = P.write_jsonl(out, records)
-    print(f"  pool/vocab_sweep/synthetic/muse-glimmer-30b.jsonl  {n} documents")
+    print(f"  pool/vocab_sweep/synthetic/{args.name}.jsonl  {n} documents")
 
     stats = {
         "tokenizer": os.path.abspath(args.tokenizer),