model-pruning
npx machina-cli add skill Orchestra-Research/AI-Research-SKILLs/model-pruning --openclawModel Pruning: Compressing LLMs
When to Use This Skill
Use Model Pruning when you need to:
- Reduce model size by 40-60% with <1% accuracy loss
- Accelerate inference using hardware-friendly sparsity (2-4× speedup)
- Deploy on constrained hardware (mobile, edge devices)
- Compress without retraining using one-shot methods
- Enable efficient serving with reduced memory footprint
Key Techniques: Wanda (weights × activations), SparseGPT (second-order), structured pruning, N:M sparsity
Papers: Wanda ICLR 2024 (arXiv 2306.11695), SparseGPT (arXiv 2301.00774)
Installation
# Wanda implementation
git clone https://github.com/locuslab/wanda
cd wanda
pip install -r requirements.txt
# Optional: SparseGPT
git clone https://github.com/IST-DASLab/sparsegpt
cd sparsegpt
pip install -e .
# Dependencies
pip install torch transformers accelerate
Quick Start
Wanda Pruning (One-Shot, No Retraining)
Source: ICLR 2024 (arXiv 2306.11695)
import torch
from transformers import AutoModelForCausalLM, AutoTokenizer
# Load model
model = AutoModelForCausalLM.from_pretrained(
"meta-llama/Llama-2-7b-hf",
torch_dtype=torch.float16,
device_map="cuda"
)
tokenizer = AutoTokenizer.from_pretrained("meta-llama/Llama-2-7b-hf")
# Calibration data (small dataset for activation statistics)
calib_data = [
"The quick brown fox jumps over the lazy dog.",
"Machine learning is transforming the world.",
"Artificial intelligence powers modern applications.",
]
# Wanda pruning function
def wanda_prune(model, calib_data, sparsity=0.5):
"""
Wanda: Prune by weight magnitude × input activation.
Args:
sparsity: Fraction of weights to prune (0.5 = 50%)
"""
# 1. Collect activation statistics
activations = {}
def hook_fn(name):
def hook(module, input, output):
# Store input activation norms
activations[name] = input[0].detach().abs().mean(dim=0)
return hook
# Register hooks for all linear layers
hooks = []
for name, module in model.named_modules():
if isinstance(module, torch.nn.Linear):
hooks.append(module.register_forward_hook(hook_fn(name)))
# Run calibration data
model.eval()
with torch.no_grad():
for text in calib_data:
inputs = tokenizer(text, return_tensors="pt").to(model.device)
model(**inputs)
# Remove hooks
for hook in hooks:
hook.remove()
# 2. Prune weights based on |weight| × activation
for name, module in model.named_modules():
if isinstance(module, torch.nn.Linear) and name in activations:
W = module.weight.data
act = activations[name]
# Compute importance: |weight| × activation
importance = W.abs() * act.unsqueeze(0)
# Flatten and find threshold
threshold = torch.quantile(importance.flatten(), sparsity)
# Create mask
mask = importance >= threshold
# Apply mask (prune)
W *= mask.float()
return model
# Apply Wanda pruning (50% sparsity, one-shot, no retraining)
pruned_model = wanda_prune(model, calib_data, sparsity=0.5)
# Save
pruned_model.save_pretrained("./llama-2-7b-wanda-50")
SparseGPT (Second-Order Pruning)
Source: arXiv 2301.00774
from sparsegpt import SparseGPT
# Load model
model = AutoModelForCausalLM.from_pretrained("meta-llama/Llama-2-7b-hf")
# Initialize SparseGPT
pruner = SparseGPT(model)
# Calibration data
calib_data = load_calibration_data() # ~128 samples
# Prune (one-shot, layer-wise reconstruction)
pruned_model = pruner.prune(
calib_data=calib_data,
sparsity=0.5, # 50% sparsity
prunen=0, # Unstructured (0) or N:M structured
prunem=0,
percdamp=0.01, # Damping for Hessian inverse
)
# Results: Near-lossless pruning at 50% sparsity
N:M Structured Pruning (Hardware Accelerator)
def nm_prune(weight, n=2, m=4):
"""
N:M pruning: Keep N weights per M consecutive weights.
Example: 2:4 = keep 2 out of every 4 weights.
Compatible with NVIDIA sparse tensor cores (2:4, 4:8).
"""
# Reshape weight into groups of M
shape = weight.shape
weight_flat = weight.flatten()
# Pad to multiple of M
pad_size = (m - weight_flat.numel() % m) % m
weight_padded = F.pad(weight_flat, (0, pad_size))
# Reshape into (num_groups, m)
weight_grouped = weight_padded.reshape(-1, m)
# Find top-N in each group
_, indices = torch.topk(weight_grouped.abs(), n, dim=-1)
# Create mask
mask = torch.zeros_like(weight_grouped)
mask.scatter_(1, indices, 1.0)
# Apply mask
weight_pruned = weight_grouped * mask
# Reshape back
weight_pruned = weight_pruned.flatten()[:weight_flat.numel()]
return weight_pruned.reshape(shape)
# Apply 2:4 sparsity (NVIDIA hardware)
for name, module in model.named_modules():
if isinstance(module, torch.nn.Linear):
module.weight.data = nm_prune(module.weight.data, n=2, m=4)
# 50% sparsity, 2× speedup on A100 with sparse tensor cores
Core Concepts
1. Pruning Criteria
Magnitude Pruning (baseline):
# Prune weights with smallest absolute values
importance = weight.abs()
threshold = torch.quantile(importance, sparsity)
mask = importance >= threshold
Wanda (weights × activations):
# Importance = |weight| × input_activation
importance = weight.abs() * activation
# Better than magnitude alone (considers usage)
SparseGPT (second-order):
# Uses Hessian (second derivative) for importance
# More accurate but computationally expensive
importance = weight^2 / diag(Hessian)
2. Structured vs Unstructured
Unstructured (fine-grained):
- Prune individual weights
- Higher quality (better accuracy)
- No hardware speedup (irregular sparsity)
Structured (coarse-grained):
- Prune entire neurons, heads, or layers
- Lower quality (more accuracy loss)
- Hardware speedup (regular sparsity)
Semi-structured (N:M):
- Best of both worlds
- 50% sparsity (2:4) → 2× speedup on NVIDIA GPUs
- Minimal accuracy loss
3. Sparsity Patterns
# Unstructured (random)
# [1, 0, 1, 0, 1, 1, 0, 0]
# Pros: Flexible, high quality
# Cons: No speedup
# Structured (block)
# [1, 1, 0, 0, 1, 1, 0, 0]
# Pros: Hardware friendly
# Cons: More accuracy loss
# N:M (semi-structured)
# [1, 0, 1, 0] [1, 1, 0, 0] (2:4 pattern)
# Pros: Hardware speedup + good quality
# Cons: Requires specific hardware (NVIDIA)
Pruning Strategies
Strategy 1: Gradual Magnitude Pruning
def gradual_prune(model, initial_sparsity=0.0, final_sparsity=0.5, num_steps=100):
"""Gradually increase sparsity during training."""
for step in range(num_steps):
# Current sparsity
current_sparsity = initial_sparsity + (final_sparsity - initial_sparsity) * (step / num_steps)
# Prune at current sparsity
for module in model.modules():
if isinstance(module, torch.nn.Linear):
weight = module.weight.data
threshold = torch.quantile(weight.abs().flatten(), current_sparsity)
mask = weight.abs() >= threshold
weight *= mask.float()
# Train one step
train_step(model)
return model
Strategy 2: Layer-wise Pruning
def layer_wise_prune(model, sparsity_per_layer):
"""Different sparsity for different layers."""
# Early layers: Less pruning (more important)
# Late layers: More pruning (less critical)
sparsity_schedule = {
"layer.0": 0.3, # 30% sparsity
"layer.1": 0.4,
"layer.2": 0.5,
"layer.3": 0.6, # 60% sparsity
}
for name, module in model.named_modules():
if isinstance(module, torch.nn.Linear):
# Find layer index
for layer_name, sparsity in sparsity_schedule.items():
if layer_name in name:
# Prune at layer-specific sparsity
prune_layer(module, sparsity)
break
return model
Strategy 3: Iterative Pruning + Fine-tuning
def iterative_prune_finetune(model, target_sparsity=0.5, iterations=5):
"""Prune gradually with fine-tuning between iterations."""
current_sparsity = 0.0
sparsity_increment = target_sparsity / iterations
for i in range(iterations):
# Increase sparsity
current_sparsity += sparsity_increment
# Prune
prune_model(model, sparsity=current_sparsity)
# Fine-tune (recover accuracy)
fine_tune(model, epochs=2, lr=1e-5)
return model
# Results: Better accuracy than one-shot at high sparsity
Production Deployment
Complete Pruning Pipeline
from transformers import Trainer, TrainingArguments
def production_pruning_pipeline(
model_name="meta-llama/Llama-2-7b-hf",
target_sparsity=0.5,
method="wanda", # or "sparsegpt"
):
# 1. Load model
model = AutoModelForCausalLM.from_pretrained(model_name, torch_dtype=torch.float16)
tokenizer = AutoTokenizer.from_pretrained(model_name)
# 2. Load calibration data
calib_dataset = load_dataset("wikitext", "wikitext-2-raw-v1", split="train[:1000]")
# 3. Apply pruning
if method == "wanda":
pruned_model = wanda_prune(model, calib_dataset, sparsity=target_sparsity)
elif method == "sparsegpt":
pruner = SparseGPT(model)
pruned_model = pruner.prune(calib_dataset, sparsity=target_sparsity)
# 4. (Optional) Fine-tune to recover accuracy
training_args = TrainingArguments(
output_dir="./pruned-model",
num_train_epochs=1,
per_device_train_batch_size=4,
learning_rate=1e-5,
bf16=True,
)
trainer = Trainer(
model=pruned_model,
args=training_args,
train_dataset=finetune_dataset,
)
trainer.train()
# 5. Save
pruned_model.save_pretrained("./pruned-llama-7b-50")
tokenizer.save_pretrained("./pruned-llama-7b-50")
return pruned_model
# Usage
pruned_model = production_pruning_pipeline(
model_name="meta-llama/Llama-2-7b-hf",
target_sparsity=0.5,
method="wanda"
)
Evaluation
from lm_eval import evaluator
# Evaluate pruned vs original model
original_results = evaluator.simple_evaluate(
model="hf",
model_args="pretrained=meta-llama/Llama-2-7b-hf",
tasks=["arc_easy", "hellaswag", "winogrande"],
)
pruned_results = evaluator.simple_evaluate(
model="hf",
model_args="pretrained=./pruned-llama-7b-50",
tasks=["arc_easy", "hellaswag", "winogrande"],
)
# Compare
print(f"Original: {original_results['results']['arc_easy']['acc']:.3f}")
print(f"Pruned: {pruned_results['results']['arc_easy']['acc']:.3f}")
print(f"Degradation: {(original_results - pruned_results):.3f}")
# Typical results at 50% sparsity:
# - Wanda: <1% accuracy loss
# - SparseGPT: <0.5% accuracy loss
# - Magnitude: 2-3% accuracy loss
Best Practices
1. Sparsity Selection
# Conservative (safe)
sparsity = 0.3 # 30%, <0.5% loss
# Balanced (recommended)
sparsity = 0.5 # 50%, ~1% loss
# Aggressive (risky)
sparsity = 0.7 # 70%, 2-5% loss
# Extreme (model-dependent)
sparsity = 0.9 # 90%, significant degradation
2. Method Selection
# One-shot, no retraining → Wanda or SparseGPT
if no_retraining_budget:
use_method = "wanda" # Faster
# Best quality → SparseGPT
if need_best_quality:
use_method = "sparsegpt" # More accurate
# Hardware speedup → N:M structured
if need_speedup:
use_method = "nm_prune" # 2:4 or 4:8
3. Avoid Common Pitfalls
# ❌ Bad: Pruning without calibration data
prune_random(model) # No activation statistics
# ✅ Good: Use calibration data
prune_wanda(model, calib_data)
# ❌ Bad: Too high sparsity in one shot
prune(model, sparsity=0.9) # Massive accuracy loss
# ✅ Good: Gradual or iterative
iterative_prune(model, target=0.9, steps=10)
Performance Comparison
Pruning methods at 50% sparsity (LLaMA-7B):
| Method | Accuracy Loss | Speed | Memory | Retraining Needed |
|---|---|---|---|---|
| Magnitude | -2.5% | 1.0× | -50% | No |
| Wanda | -0.8% | 1.0× | -50% | No |
| SparseGPT | -0.4% | 1.0× | -50% | No |
| N:M (2:4) | -1.0% | 2.0× | -50% | No |
| Structured | -3.0% | 2.0× | -50% | No |
Source: Wanda paper (ICLR 2024), SparseGPT paper
Resources
- Wanda Paper (ICLR 2024): https://arxiv.org/abs/2306.11695
- Wanda GitHub: https://github.com/locuslab/wanda
- SparseGPT Paper: https://arxiv.org/abs/2301.00774
- SparseGPT GitHub: https://github.com/IST-DASLab/sparsegpt
- NVIDIA Sparse Tensor Cores: https://developer.nvidia.com/blog/accelerating-inference-with-sparsity-using-ampere-and-tensorrt/
Source
git clone https://github.com/Orchestra-Research/AI-Research-SKILLs/blob/main/19-emerging-techniques/model-pruning/SKILL.mdView on GitHub Overview
Model pruning reduces large language models’ size and speeds up inference by removing redundant weights. It covers unstructured and structured pruning, N:M sparsity, magnitude pruning, and one-shot methods like Wanda and SparseGPT, enabling compression with minimal accuracy loss and faster runs on hardware accelerators. This is especially valuable for deploying models on mobile and edge devices or for efficient serving with reduced memory footprint.
How This Skill Works
Pruning assigns importance to weights using magnitude, activation statistics, or second-order criteria, then removes low-importance weights. Wanda pruning uses weight magnitudes times input activations, while SparseGPT relies on second-order information. One-shot pruning removes weights in a single pass, often achieving 40–60% sparsity with minimal accuracy loss, and can be tailored to hardware through N:M or structured sparsity.
When to Use It
- Reduce model size by 40-60% with minimal accuracy loss (often <1%).
- Accelerate inference with hardware-friendly sparsity (2-4× speedups).
- Deploy on constrained hardware like mobile or edge devices.
- Compress without retraining using one-shot pruning methods.
- Enable efficient serving with a reduced memory footprint.
Quick Start
- Step 1: Install Wanda and SparseGPT and set up environment (torch, transformers, accelerate).
- Step 2: Load your pre-trained model and tokenizer; prepare a small calibration dataset for activation statistics.
- Step 3: Run Wanda pruning (or SparseGPT) to target your desired sparsity (e.g., 0.5) and save the pruned model.
Best Practices
- Start with moderate sparsity (40-60%) and verify accuracy and latency trade-offs.
- Choose pruning type (unstructured vs structured; N:M) based on target hardware.
- Use activation-aware (Wanda) or second-order (SparseGPT) criteria for pruning decisions.
- Benchmark inference speed on your actual accelerators after pruning.
- Consider pairing pruning with quantization or distillation for additional gains.
Example Use Cases
- Wanda pruning applied to Llama-2-7B achieving ~50% sparsity with minimal accuracy loss.
- SparseGPT used to compress large LLMs without retraining for faster inference.
- 2–4× speedups on GPUs with hardware-friendly sparsity (N:M) pruning.
- Pruned models deployed on mobile/edge devices due to reduced memory footprint.
- Pruned models used for efficient production serving with lower resource usage.
Frequently Asked Questions
Related Skills
long-context
Orchestra-Research/AI-Research-SKILLs
Extend context windows of transformer models using RoPE, YaRN, ALiBi, and position interpolation techniques. Use when processing long documents (32k-128k+ tokens), extending pre-trained models beyond original context limits, or implementing efficient positional encodings. Covers rotary embeddings, attention biases, interpolation methods, and extrapolation strategies for LLMs.
model-merging
Orchestra-Research/AI-Research-SKILLs
Merge multiple fine-tuned models using mergekit to combine capabilities without retraining. Use when creating specialized models by blending domain-specific expertise (math + coding + chat), improving performance beyond single models, or experimenting rapidly with model variants. Covers SLERP, TIES-Merging, DARE, Task Arithmetic, linear merging, and production deployment strategies.
gguf-quantization
Orchestra-Research/AI-Research-SKILLs
GGUF format and llama.cpp quantization for efficient CPU/GPU inference. Use when deploying models on consumer hardware, Apple Silicon, or when needing flexible quantization from 2-8 bit without GPU requirements.
hqq-quantization
Orchestra-Research/AI-Research-SKILLs
Half-Quadratic Quantization for LLMs without calibration data. Use when quantizing models to 4/3/2-bit precision without needing calibration datasets, for fast quantization workflows, or when deploying with vLLM or HuggingFace Transformers.
knowledge-distillation
Orchestra-Research/AI-Research-SKILLs
Compress large language models using knowledge distillation from teacher to student models. Use when deploying smaller models with retained performance, transferring GPT-4 capabilities to open-source models, or reducing inference costs. Covers temperature scaling, soft targets, reverse KLD, logit distillation, and MiniLLM training strategies.
moe-training
Orchestra-Research/AI-Research-SKILLs
Train Mixture of Experts (MoE) models using DeepSpeed or HuggingFace. Use when training large-scale models with limited compute (5× cost reduction vs dense models), implementing sparse architectures like Mixtral 8x7B or DeepSeek-V3, or scaling model capacity without proportional compute increase. Covers MoE architectures, routing mechanisms, load balancing, expert parallelism, and inference optimization.