skills.rest
KappTech88
KappTech88COMMUNITY
Explore all skills by this author →
GitHub

model-pruning

Reduce LLM size and accelerate inference via Wanda, SparseGPT, N:M pruning methods.

Updated Mar 30, 2026
Search Tags:#llm#sparsity#pruning#model-compression#wanda#sparsegpt#inference-speed
One-click install
npx skills add https://github.com/KappTech88/AI-RESEARCH-SKILLS-MCP --skill model-pruning-kapptech88
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: model-pruning
Source: https://github.com/KappTech88/AI-RESEARCH-SKILLS-MCP/tree/main/skills/model-pruning
Command: npx skills add https://github.com/KappTech88/AI-RESEARCH-SKILLS-MCP --skill model-pruning-kapptech88

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires transformers, torch, and includes references (resource) components.

What problem does it solve?

Reduces model size and accelerates inference for large language models through pruning techniques, enabling deployment on constrained hardware and faster serving.

Core Features & Use Cases

  • Wanda pruning (weights × activations)
  • SparseGPT and N:M structured pruning options
  • One-shot and iterative pruning with optional fine-tuning
  • Use cases include compressing models without retraining and deploying on resource-constrained devices

Quick Start

Prune a pretrained model using Wanda to 50% sparsity and evaluate the impact on performance.

Frequently Asked Questions about model-pruning

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

FAQPage Schema
How do I prune a large language model to accelerate inference without retraining?▼

The Skill supports unstructured, structured, and N:M pruning methods. Unstructured targeting individual weights offers highest compression, while structured and N:M formats better align with hardware optimized for sparse inference and faster execution.

Can I use Wanda and SparseGPT to compress models for deployment on constrained hardware?▼

Yes, the Skill supports iterative pruning alongside optional fine-tuning steps. This allows you to gradually increase sparsity while recovering performance degradation, ensuring the compressed model retains accuracy for resource-constrained devices.

Does this pruning approach support both unstructured and N:M structured sparsity?▼

Yes, the Skill supports unstructured, structured, and N:M pruning methods. Unstructured removes individual weights, while structured and N:M formats better align with hardware optimized for sparse inference and faster execution.

Do I need calibration data to perform one-shot model pruning?▼

Yes, calibration data is used during the pruning process to evaluate the importance of weights. This ensures that techniques like Wanda and SparseGPT accurately assess weight activations to achieve optimal target sparsity without retraining.

Related Skills

Explore More Software Engineering Skills→
Clay-HHKClay-HHK
2

model-pruning

Prune transformer models with Wanda, SparseGPT, and N:M sparsity patterns.

Community
Advanced
informatico-madridinformatico-madrid
1

hqq-quantization

Quantize large language models to 1-8 bit precision without calibration.

Community
Intermediate
lxh755818-botlxh755818-bot

serving-llms-vllm

Serve LLMs with OpenAI API, quantization, and tensor parallelism.

Community
Advanced
zangjeicyzangjeicy

peft-fine-tuning

Fine-tune large language models with LoRA, QLoRA, and IA3.

Community
Advanced