Upload eval_lfm_finetuned.py with huggingface_hub

eval_lfm_finetuned.py

@@ -0,0 +1,29 @@
+# /// script
+# requires-python = ">=3.10"
+# dependencies = [
+#     "lighteval[accelerate]>=0.6.0",
+#     "torch>=2.0.0",
+#     "transformers>=4.40.0",
+#     "accelerate>=0.30.0",
+#     "peft>=0.7.0",
+# ]
+# ///
+
+"""Evaluate fine-tuned wheattoast11/agent-zero-lfm-1.2b-v1 on standard benchmarks."""
+
+import subprocess
+import sys
+
+tasks = "leaderboard|mmlu|5,leaderboard|hellaswag|0,leaderboard|arc_challenge|25"
+
+# PEFT adapter model — lighteval loads base + adapter automatically
+cmd = [
+    sys.executable, "-m", "lighteval", "accelerate",
+    "--model_args", "pretrained=wheattoast11/agent-zero-lfm-1.2b-v1,trust_remote_code=True",
+    "--tasks", tasks,
+    "--output_dir", "./eval_results_finetuned",
+]
+
+print(f"Running: {' '.join(cmd)}")
+result = subprocess.run(cmd, capture_output=False)
+sys.exit(result.returncode)