Unified LLM Torch Profiler Analysis

Overview

Use this skill for torch.profiler analysis across:

• sglang, including existing traces from SGLang Omni (sglang-omni) workloads
• vllm
• TensorRT-LLM
• TokenSpeed

SGLang Omni is not a separate framework. It serves through the SGLang runtime, so use --framework sglang and analyze its traces with --input or with --mapping-input and --formal-input. The SGLang live-capture path has not been validated against Omni servers.

There is only one public workflow:

• triage