torch-npu-comm-test

Community

Test and analyze NPU communication performance with PyTorch.

Authordongg622
Version1.0.0
Installs0

System Documentation

What problem does it solve?

It enables users to evaluate the performance and correctness of collective communication operators on Ascend NPUs using PyTorch's distributed API.

Core Features & Use Cases

  • Performance Benchmarking: Measure latency and bandwidth for operations like AllReduce, AllGather, ReduceScatter, and others.
  • Functional Validation: Ensure correctness of communication results via built-in validation routines.
  • Use Case: A researcher benchmarks the communication efficiency of a new cluster setup under various tensor shapes and data types, optimizing training throughput.

Quick Start

Use the tool to perform an AllReduce operation with shape 4096x12288 and dtype fp16 on 8 NPUs.

Dependency Matrix

Required Modules

torchtorch_npu

Components

scriptsreferences

💻 Claude Code Installation

Recommended: Let Claude install automatically. Simply copy and paste the text below to Claude Code.

Please help me install this Skill:
Name: torch-npu-comm-test
Download link: https://github.com/dongg622/china-ai-chip-skill/archive/main.zip#torch-npu-comm-test

Please download this .zip file, extract it, and install it in the .claude/skills/ directory.
View Source Repository

Agent Skills Search Helper

Install a tiny helper to your Agent, search and equip skill from 471,000+ vetted skills library on demand.