Add HF Jobs how-to for benchmarking a model

README.md

@@ -153,16 +153,19 @@ and prints accuracy, KL from gold, Brier and ECE:
 ```bash
 hf jobs run --flavor a100-large --timeout 2h \\
   --env MODEL_ID=your-org/your-model \\
+  --secrets HF_TOKEN \\
   --volume hf://datasets/LocalLLaMA/typed-decisions:/repo:ro \\
   pytorch/pytorch:2.9.1-cuda12.8-cudnn9-devel \\
   bash /repo/bench/score_with_jobs.sh
 ```
 
 - Pick the `--flavor` for the model's size: `l4x1` up to ~2B, `l40sx1` or
-  `a100-large` for ~27B, `h200` above. Gated models also need `--secrets HF_TOKEN`.
+  `a100-large` for ~27B, `h200` above.
 - `bench/score_with_jobs.sh` installs vLLM nightly, serves the model, scores all
   400 cases, and writes `results.json`.
-- Then open a discussion with the numbers and the mode (specialist or general).
+- The run finishes by creating the eval YAML and opening a PR on the model repo
+  with the four scores, so they show on the Hub leaderboard once merged. Add
+  `--env MODE=specialist` if you fitted the model on `train`.
 
 ### Notes on the rows
 

bench/score_with_jobs.sh

@@ -1,6 +1,7 @@
 #!/usr/bin/env bash
 # Bootstrap for `hf jobs run`: install deps, then score $MODEL_ID on the test split.
-# Reads $MODEL_ID (required). Optional: $SERVE_FLAGS (extra `vllm serve` flags).
+# Reads $MODEL_ID (required). Optional: $SERVE_FLAGS (extra `vllm serve` flags),
+# $MODE=general|specialist, $HF_TOKEN (opens the eval-results PR automatically).
 set -euo pipefail
 
 MODEL_ID=${MODEL_ID:?set --env MODEL_ID=org/model}
@@ -8,3 +9,9 @@ SCRIPT_DIR=$(cd "$(dirname "$0")" && pwd)
 pip install -q datasets requests
 pip install -q -U vllm --torch-backend=auto --extra-index-url https://wheels.vllm.ai/nightly
 python "$SCRIPT_DIR/score_with_jobs.py" "$MODEL_ID"
+
+if [ -n "${HF_TOKEN:-}" ]; then
+  python "$SCRIPT_DIR/submit_results.py" "$MODEL_ID" results.json --mode "${MODE:-general}"
+else
+  echo "[submit] pass --secrets HF_TOKEN to also open a PR with the eval results on the model repo"
+fi

bench/submit_results.py

@@ -0,0 +1,85 @@
+"""Create the Typed Decisions eval YAML and open a PR on the model repo.
+
+Usage:
+    python submit_results.py MODEL_ID [RESULTS_JSON] [--mode general|specialist] [--dry-run]
+
+Reads the summary produced by score_with_jobs.py (accuracy, kl_from_gold,
+brier, ece), writes the Hub eval-results YAML, and opens a pull request on
+the model repo (`.eval_results/<model>.yaml`). With --dry-run the YAML is
+printed instead of uploaded.
+"""
+import argparse
+import datetime
+import json
+
+BENCH = "LocalLLaMA/typed-decisions"
+TASKS = ["accuracy", "kl_from_gold", "brier", "ece"]
+NOTE_SPECIALIST = (
+    "Model was fitted/fine-tuned on this benchmark's train split; "
+    "not comparable with zero-shot rows."
+)
+
+
+def build_yaml(summary: dict, mode: str, date: str) -> str:
+    lines = []
+    for task_id in TASKS:
+        value = round(float(summary[task_id]), 4)
+        lines += [
+            "- dataset:",
+            f"    id: {BENCH}",
+            f"    task_id: {task_id}",
+            f"  value: {value}",
+            f'  date: "{date}"',
+            "  source:",
+            f"    url: https://huggingface.co/datasets/{BENCH}",
+            '    name: "Typed Decisions benchmark run on HF Jobs"',
+        ]
+        if mode == "specialist":
+            lines.append(f"  notes: {json.dumps(NOTE_SPECIALIST)}")
+    return "\n".join(lines) + "\n"
+
+
+def main():
+    ap = argparse.ArgumentParser(description=__doc__)
+    ap.add_argument("model_id", help="model repo the results belong to, e.g. org/model")
+    ap.add_argument("results", nargs="?", default="results.json", help="summary JSON from score_with_jobs.py")
+    ap.add_argument("--mode", choices=["general", "specialist"], default="general",
+                    help="specialist = fitted on the benchmark's train split")
+    ap.add_argument("--dry-run", action="store_true", help="print the YAML instead of opening a PR")
+    args = ap.parse_args()
+
+    with open(args.results) as f:
+        summary = json.load(f)
+    missing = [t for t in TASKS if t not in summary]
+    if missing:
+        raise SystemExit(f"results file is missing task keys: {missing}")
+    date = datetime.date.today().isoformat()
+    yaml_text = build_yaml(summary, args.mode, date)
+
+    file_name = args.model_id.replace("/", "__") + ".yaml"
+    if args.dry_run:
+        print(yaml_text)
+        return
+
+    from huggingface_hub import HfApi
+
+    api = HfApi()
+    commit = api.upload_file(
+        path_or_fileobj=yaml_text.encode(),
+        path_in_repo=f".eval_results/{file_name}",
+        repo_id=args.model_id,
+        repo_type="model",
+        create_pr=True,
+        commit_message="Add Typed Decisions eval results",
+        commit_description=(
+            f"{args.mode} run on {date}: accuracy {summary['accuracy']}, "
+            f"KL {summary['kl_from_gold']}, Brier {summary['brier']}, ECE {summary['ece']} "
+            f"({summary['n_decisions']} decisions, one request per case)."
+        ),
+    )
+    pr_url = getattr(commit, "pr_url", None) or getattr(commit, "pull_request_url", "")
+    print(f"[submit] eval YAML opened as a PR on {args.model_id}: {pr_url}", flush=True)
+
+
+if __name__ == "__main__":
+    main()