skills.rest
PaddlePaddle
PaddlePaddleOFFICIAL
Explore all skills by this author →
WebsiteGitHub

benchmark

Benchmark FastDeploy versus SGLang with end-to-end tests and HTML reports.

3.7k|758|Updated Jun 27, 2022
Search Tags:#performance#metrics#reporting#benchmark#inference#sglang#fastdeploy
One-click install
npx skills add https://github.com/PaddlePaddle/FastDeploy --skill benchmark-paddlepaddle
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: benchmark
Source: https://github.com/PaddlePaddle/FastDeploy/tree/main/.claude/skills/benchmark-compare
Command: npx skills add https://github.com/PaddlePaddle/FastDeploy --skill benchmark-paddlepaddle

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes scripts (resource) and references (resource) components.

What problem does it solve?

Benchmark FastDeploy vs SGLang by orchestrating end-to-end tests, collecting performance metrics, and producing interactive HTML reports.

Core Features & Use Cases

  • End-to-end benchmark workflow: environment setup, service deployment, and metric reporting.
  • Supports single and multi-GPU deployments, multiple deployment modes (single, TP, PD), and a report-only path using existing logs.
  • Use case: teams evaluating model deployment frameworks for latency, throughput, and cost efficiency.

Quick Start

Run a full benchmark comparing FastDeploy and SGLang with default parameters.

Frequently Asked Questions about benchmark

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I benchmark FastDeploy vs SGLang inference performance?▼

You can benchmark FastDeploy vs SGLang by running end-to-end tests that collect performance metrics and produce interactive HTML reports, supporting both single- and multi-GPU deployments.

Can I generate a performance report from existing inference logs?▼

Yes, you can generate a performance report from existing inference logs by using the report-only path, which parses collected metrics and composes interactive HTML reports without redeploying services.

Does the benchmark support multi-GPU deployments and different modes?▼

Yes, the benchmark supports multi-GPU deployments and multiple modes including single, TP (tensor parallel), and PD (prefill-decode) to evaluate latency and throughput across different configurations.

What Python environment is required to run the FastDeploy and SGLang benchmark?▼

Running the FastDeploy and SGLang benchmark requires a Python 3.10 environment, utilizing provided environment setup scripts to deploy services and extract metrics for reporting.

What is the best way to evaluate model deployment frameworks for cost efficiency?▼

The best way to evaluate model deployment frameworks for cost efficiency is orchestrating end-to-end benchmark tests that collect latency and throughput metrics, producing interactive HTML reports for direct comparison.

Related Skills

Explore More Data & Analytics Skills→
dogekikidogekiki

vast-gpu

Automate vast.ai GPU instance provisioning, configuration, and cleanup via vastai CLI.

Community
Advanced
harshaldhadukharshaldhaduk
1

worker-benchmarks

Benchmark agentic-flow worker operations against configurable latency thresholds.

Community
Intermediate
Tyler-R-KendrickTyler-R-Kendrick
1

hf-cloud-sagemaker-deployment-planner

Plan machine learning model deployments to Amazon SageMaker with optimal instance configurations.

Community
Advanced
AMD-AGIAMD-AGI
29

performance-analysis

Analyze MaxText training runs to identify performance bottlenecks using TSDB, TraceLens, and IRLens.

Official
Advanced