MCPcopy Create free account
hub / github.com/Sphere-AI-Lab/FormalMATH-Bench / parse_args

Function parse_args

FoMA_Eval.py:9–62  ·  view source on GitHub ↗
()

Source from the content-addressed store, hash-verified

7from evaluate_results import monte_carlo_evaluate
8
9def parse_args():
10 parser = argparse.ArgumentParser(description="Pipeline for Lean theorem proof (generation, verification, and evaluation)")
11
12 # File paths
13 parser.add_argument("--input_file", default=None,
14 help="Path to the initial input file")
15 parser.add_argument("--generated_file", default=None,
16 help="Path to the output file containing generated answers")
17 parser.add_argument("--verification_file", default=None,
18 help="Path to the output file containing verification results")
19 parser.add_argument("--evaluation_file", default=None,
20 help="Path to the output file containing evaluation results")
21
22 # Task control
23 parser.add_argument("--auto_dl", action="store_true", default=True,
24 help="Automatically download dataset")
25 parser.add_argument("--generate", action="store_true", default=False,
26 help="Enable generation of answers")
27 parser.add_argument("--verify", action="store_true", default=False,
28 help="Enable verification of generated answers")
29 parser.add_argument("--evaluate", action="store_true", default=False,
30 help="Enable evaluation of verification results")
31 parser.add_argument("--datasets", default="FomaMATH-All",
32 help="Choose dataset version: FomaMATH-All or FomaMATH-Lite")
33
34 # Generation parameters - Add all parameters from the first script
35 parser.add_argument("--model", default=None,
36 help="Path to the model used for generating answers.")
37 parser.add_argument("--n", type=int, default=200,
38 help="Number of answers to generate per process via vllm.")
39 parser.add_argument("--nums_answer", type=int, default=3200,
40 help="Number of answers to generate per question.")
41
42 # Verification parameters
43 parser.add_argument("--repl_path", default="./repl",
44 help="Path to the Lean REPL used for verification")
45 parser.add_argument("--lean_env_path", default="./repl/test/Mathlib",
46 help="Path to the Lean environment used for verification")
47 parser.add_argument("--num_batches", default=32, type=int,
48 help="Number of parallel batches for verification")
49 parser.add_argument("--session_timeout", default=600, type=int,
50 help="Timeout for interactive sessions in seconds")
51 parser.add_argument("--expect_timeout", default=120, type=int,
52 help="Timeout for the expect command in seconds")
53
54 # Evaluation parameters
55 parser.add_argument("--n_simulations", default=50, type=int,
56 help="Number of Monte Carlo simulations")
57 parser.add_argument("--n_processes", default=50, type=int,
58 help="Number of parallel processes for Monte Carlo simulation")
59 parser.add_argument("--custom_sample_sizes", default=None, type=str,
60 help="Custom sampling sizes as a comma-separated list (e.g., '1,5,10,50,100')")
61
62 return parser.parse_args()
63def set_up_logging(level=logging.INFO):
64 """Set up logging with the specified level."""
65 logging.basicConfig(level=level, format='%(asctime)s - %(levelname)s - %(message)s')

Callers 1

mainFunction · 0.70

Calls

no outgoing calls

Tested by

no test coverage detected