Add TP2 prefill serving-path smoke experiment

runs/frontier-tp2-prefill-serving-v0/materialize_prefill_request.py

@@ -0,0 +1,50 @@
+#!/usr/bin/env python3
+"""Create one exact-length, prefix-disjoint prefill request."""
+
+from __future__ import annotations
+
+import argparse
+import json
+from pathlib import Path
+
+
+def parse_args() -> argparse.Namespace:
+    parser = argparse.ArgumentParser()
+    parser.add_argument("--model", type=Path, required=True)
+    parser.add_argument("--input-tokens", type=int, default=8192)
+    parser.add_argument("--output-tokens", type=int, default=2)
+    parser.add_argument("--output", type=Path, required=True)
+    return parser.parse_args()
+
+
+def main() -> None:
+    args = parse_args()
+    if min(args.input_tokens, args.output_tokens) <= 0:
+        raise ValueError("token counts must be positive")
+    if args.input_tokens + args.output_tokens > 40960:
+        raise ValueError("request exceeds the server max model length")
+
+    from transformers import AutoTokenizer
+
+    tokenizer = AutoTokenizer.from_pretrained(args.model, trust_remote_code=True)
+    special = set(tokenizer.all_special_ids)
+    candidates = [
+        token for token in range(tokenizer.vocab_size) if token not in special
+    ]
+    if len(candidates) < 2:
+        raise ValueError("tokenizer has too few non-special token IDs")
+
+    body = {
+        "model": "qwen30-prefill-profile",
+        "prompt": [candidates[1], *([candidates[0]] * (args.input_tokens - 1))],
+        "min_tokens": args.output_tokens,
+        "max_tokens": args.output_tokens,
+        "ignore_eos": True,
+        "temperature": 0,
+    }
+    args.output.parent.mkdir(parents=True, exist_ok=True)
+    args.output.write_text(json.dumps(body, separators=(",", ":")) + "\n")
+
+
+if __name__ == "__main__":
+    main()