parse_log_file() — pytorch Function Reference

Architecture documentation for the parse_log_file() function in analyze_templates.py from the pytorch codebase.

Function python DynamoBenchmarks Configuration called by 1

Entity Profile

DynamoBenchmarks→ Configuration→ parse_log_file() — pytorch Function Reference

Dependency Diagram

graph TD
  551bcc43_5dd8_fe79_e16b_3b670a030d46["parse_log_file()"]
  fc92d87b_ad85_a501_c5b5_6e3bd9880812["main()"]
  fc92d87b_ad85_a501_c5b5_6e3bd9880812 -->|calls| 551bcc43_5dd8_fe79_e16b_3b670a030d46
  style 551bcc43_5dd8_fe79_e16b_3b670a030d46 fill:#6366f1,stroke:#818cf8,color:#fff

Relationship Graph

Source Code

benchmarks/dynamo/microbenchmarks/analyze_templates.py lines 14–34

def parse_log_file(file_path):
    with open(file_path) as f:
        logs = json.load(f)

    occurrence_count = {}
    benchmark_logs = {}

    # Parse the logs
    for entry in logs:
        if "invoke" in entry:
            shape = entry["invoke"]
            if shape not in occurrence_count:
                occurrence_count[shape] = 0
            occurrence_count[shape] += 1
        else:
            for shape, timings in entry.items():
                if shape not in benchmark_logs:
                    benchmark_logs[shape] = []
                benchmark_logs[shape].extend(timings)

    return occurrence_count, benchmark_logs