Iggie Wang
Community@hey-kong · Wuhan, China
Agent Skills by Iggie Wang
Showing 17 vetted skills indexed across 1 GitHub repositories.
hey-kongsglang-diffusion-modelopt-quant
Quantize diffusion transformers with NVIDIA ModelOpt to generate FP8 or NVFP4 checkpoints for SGLang Diffusion.
hey-kongsglang-diffusion-benchmark-profile
Benchmark denoise latency and profile hotspots in SGLang diffusion backends.
hey-kongsglang-diffusion-add-model
Guide developers to integrate diffusion models into SGLang pipelines.
hey-kongsglang-diffusion-ako4all-kernel
Optimize SGLang diffusion kernels with AKO4ALL microbenchmarks and denoise validation.
hey-kongsglang-diffusion-performance
Benchmark SGLang and Diffusers backends to optimize diffusion performance on multi-GPU setups.
hey-kongclean-startup-log
Clean noisy SGLang server startup logs by classifying and suppressing non-actionable lines.
hey-kongwrite-sglang-test
Generate SGLang unit and E2E tests with CustomTestCase and CI registration.
hey-kongci-workflow-guide
Design and maintain CI pipelines for SGLang projects with stage ordering and gating.
hey-kongadd-jit-kernel
Add a JIT CUDA kernel to SGLang's jit_kernel module with Python wrapper and tests.
hey-kongdebug-cuda-crash
Diagnoses CUDA crashes in SGLang via level-based kernel API logging and crash dumps.
hey-kongllm-serving-auto-benchmark
Generate and validate deployment commands for comparing LLM serving frameworks.
hey-kongsglang-prod-incident-triage
Replay failing SGLang requests on a clean target for incident triage.
hey-kongadd-sgl-kernel
Implement and integrate CUDA/C++ kernels with tests and benchmarks.
hey-kongllm-torch-profiler-analysis
Analyze torch.profiler traces to generate kernel, overlap, and fuse-pattern tables.
hey-kongsglang-bisect-ci-regression
Extract failure signatures and bisect CI regression commit ranges in SGLang workflows.
hey-kongsglang-sota-performance
Coordinate benchmarking, profiling, patching, and revalidation across SGLang, vLLM, and TensorRT-LLM.
hey-kongdebug-distributed-hang
Trace per-rank state and NCCL logs to diagnose multi-GPU SGLang inference hangs.