# FlashInfer-Bench

## Docs

- [Definition](https://bench.flashinfer.ai/docs/flashinfer-trace/definition.md)
- [FlashInfer Trace Schema](https://bench.flashinfer.ai/docs/flashinfer-trace/index.md)
- [Solution](https://bench.flashinfer.ai/docs/flashinfer-trace/solution.md)
- [Trace](https://bench.flashinfer.ai/docs/flashinfer-trace/trace.md)
- [Workload](https://bench.flashinfer.ai/docs/flashinfer-trace/workload.md)
- [Introduction](https://bench.flashinfer.ai/docs/index.md): Welcome to FlashInfer-Bench, the AI-for-AI infrastructure that makes LLM serving self-improving.
- [Model Coverage](https://bench.flashinfer.ai/docs/model_coverage.md)
- [DSA Paged](https://bench.flashinfer.ai/docs/op-types/dsa-paged.md)
- [GDN](https://bench.flashinfer.ai/docs/op-types/gdn.md)
- [GEMM](https://bench.flashinfer.ai/docs/op-types/gemm.md)
- [GQA Paged](https://bench.flashinfer.ai/docs/op-types/gqa-paged.md)
- [GQA Ragged](https://bench.flashinfer.ai/docs/op-types/gqa-ragged.md)
- [MLA Paged](https://bench.flashinfer.ai/docs/op-types/mla-paged.md)
- [MoE](https://bench.flashinfer.ai/docs/op-types/moe.md)
- [RMSNorm](https://bench.flashinfer.ai/docs/op-types/rmsnorm.md)
- [RoPE](https://bench.flashinfer.ai/docs/op-types/rope.md)
- [Sampling](https://bench.flashinfer.ai/docs/op-types/sampling.md)
- [Installation](https://bench.flashinfer.ai/docs/start/installation.md): Set up FlashInfer-Bench from PyPI or source and verify your environment.
- [Quick Start](https://bench.flashinfer.ai/docs/start/quickstart.md): This guide shows you how to use FlashInfer-Bench python module with the FlashInfer-Trace dataset.
- [Bring Your Own Kernel to FlashInfer-Bench](https://bench.flashinfer.ai/docs/tutorials/bring-your-own-kernel.md): Add definitions, solutions, workloads, and evaluations to integrate custom kernels with FlashInfer-Bench.
- [CLI](https://bench.flashinfer.ai/docs/tutorials/cli.md): Use FlashInfer-Bench from the command line to run benchmarks and inspect results.
- [Run Benchmark](https://bench.flashinfer.ai/docs/tutorials/run-benchmark.md): Configure and run benchmarks with per-op-type and per-definition eval parameters.
- [Benchmark Server API](https://bench.flashinfer.ai/docs/tutorials/server-api.md): Start `flashinfer-bench serve` and use its HTTP API to inspect definitions, submit solutions, and poll evaluation tasks.

## Optional

- [FlashInfer-Bench](https://bench.flashinfer.ai/)
- [GitHub](https://github.com/flashinfer-ai/flashinfer-bench)
- [API Reference](https://bench.flashinfer.ai/docs/api/python)