Merge pull request #24 from lenianiva/experiments/minif2f

experiment: MiniF2F speedup

experiments/minif2f/README.md

@@ -1,10 +1,17 @@
 # MiniF2F
 
 This is an experiment on running a LLM prover on miniF2F data. Build the project
-`MiniF2F` with `lake build`, and run with
+`MiniF2F` with `lake build`. Check the environment and data with
+
+``` sh
+python3 experiments/minif2f/main.py check
+python3 experiments/minif2f/main.py list
+```
+
+and run experiments with
 
 ```sh
-python3 experiments/minif2f/main.py [--dry-run] [--use-llm]
+python3 experiments/minif2f/main.py eval [--use-llm] [--use-hammer]
 ```
 
 Read the help message carefully.

experiments/minif2f/main.py

@@ -29,10 +29,12 @@ def try_test_data(server, agent, entry: dict, max_steps: int, max_trials_per_goa
 
     goal_states = server.load_sorry(command)
 
-    if len(goal_states) == 0:
+    if len(goal_states) != 1:
         return None
 
     goal_state, = goal_states
+    if isinstance(goal_state, list):
+        return None
     try:
         return agent.search(
             server=server,
@@ -103,16 +105,28 @@ def run_eval(args):
             with open(file_name, 'w') as f:
                 json.dump({ 'id': datum['id'], 'success': result.success, 'steps': result.steps  }, f)
 
+def run_check(args):
+    project_path, lean_path = get_project_and_lean_path()
+    print(f"$PWD: {project_path}")
+    print(f"$LEAN_PATH: {lean_path}")
+    server = Server(
+        imports=["Mathlib", "Aesop"],
+        project_path=project_path,
+        lean_path=lean_path,
+        core_options=CORE_OPTIONS,
+    )
+
 if __name__ == '__main__':
     parser = argparse.ArgumentParser(
         prog='MiniF2F Search',
         description='Executes LLM on MiniF2F Search',
     )
-    parser.add_argument('--use-hammer', action='store_true')
     parser.add_argument(
-        '--dry-run',
+        'mode',
-        action='store_true',
+        help='Function',
-        help="List the data used, but don't run")
+        choices=['list', 'eval', 'check'],
+    )
+    parser.add_argument('--use-hammer', action='store_true')
     parser.add_argument('--validation', action='store_true')
     parser.add_argument('--use-llm', action='store_true')
     parser.add_argument('--max-steps', default=50)
@@ -120,7 +134,11 @@ if __name__ == '__main__':
     parser.add_argument('--feedback-turns', default=2)
     args = parser.parse_args()
 
-    if args.dry_run:
+    if args.mode == "list":
         dry_run(args)
-    else:
+    elif args.mode == "eval":
         run_eval(args)
+    elif args.mode == "check":
+        run_check(args)
+    else:
+        raise ValueError(f"Invalid mode: {args.mode}")

experiments/minif2f/model/llm_agent.py

@@ -1,5 +1,6 @@
 from typing import Optional
 import collections, unittest
+from termcolor import colored
 from pantograph.search import Agent
 from pantograph.server import Server, TacticFailure, ServerError
 from pantograph.expr import Expr, Tactic, GoalState
@@ -57,6 +58,7 @@ class LLMAgent(Agent):
             new_state = None
             for ii in range(self.n_trials):
                 print(f"===============trail {str(ii)}============")
+                try:
                     s = select_tactic.run(
                         server=self.server,
                         state=state,
@@ -70,7 +72,13 @@ class LLMAgent(Agent):
 
                     print("\n-- new state --\n", new_state)
                     if tactic:
+                        if not isinstance(tactic, Tactic):
+                            print(colored("[Tactic] Failed:", "red"), tactic)
+                            return None
                         return tactic
+                except Exception as e:
+                    print(colored(str(e), "red"))
+                    return None
             return None
         else:
             self.goal_tactic_id_map[key] = i + 1