mindspeed-llm-weight-prep

Community

Simplify model weight conversion for Huawei Ascend NPUs.

Authordongg622
Version1.0.0
Installs0

System Documentation

What problem does it solve?

This Skill streamlines the process of converting and managing large AI model weights between HuggingFace and Megatron formats, reducing manual effort and errors.

Core Features & Use Cases

  • Weight format conversion: Easily switch model weights between HF and Megatron formats for training or inference.
  • Parallel configuration setup: Automatically configure tensor and pipeline parallelism settings for optimal performance.
  • Use Case: When deploying a large language model on Ascend NPUs, use this Skill to convert model weights and optimize parallel execution for efficient inference.

Quick Start

Use this Skill to convert your HuggingFace model weights to Megatron format for deployment on Huawei Ascend hardware.

Dependency Matrix

Required Modules

None required

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: mindspeed-llm-weight-prep
Download link: https://github.com/dongg622/china-ai-chip-skill/archive/main.zip#mindspeed-llm-weight-prep

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.