Welcome to FlashInfer-Bench Python API documentation!

Blog Post | GitHub | Slack (Join channel #flashinfer-bench)

FlashInfer-Bench is a comprehensive benchmark and infrastructure designed to create a “virtuous cycle” where AI can automatically optimize and improve the core GPU kernels of the AI systems it runs on. It provides a systematic framework to identify performance bottlenecks, generate solutions, and deploy them immediately into production.

API Reference

• flashinfer_bench.apply
• flashinfer_bench.tracing
• FlashInfer Trace Schema Python API
  • Definition
  • Solution
  • Trace
  • TraceSet