neuralmagic/sparseml

Health v2 · maintenance only

Flags: archived

Adoption not part of the score

2145 stars · 156 forks observed · 2026-08-28

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-30, confidence not recorded

SparseML is a Python library for applying sparsification recipes (pruning, quantization, sparsity) to neural networks with a few lines of code, producing faster and smaller models across PyTorch, TensorFlow/Keras, and ONNX. Development ceased and the project was deprecated on June 2, 2025, following Neural Magic's acquisition by Red Hat.

Use cases

• prune a pytorch model for faster inference
• quantize a neural network to reduce its size
• apply sparsification recipes to deep learning models
• optimize an onnx model for deployment
• compress a computer vision model for edge devices
• speed up nlp model inference with sparsity

When to choose

• you need to prune or quantize PyTorch, TensorFlow, or ONNX models and are comfortable using an unmaintained library
• you want recipe-driven model sparsification with minimal code changes

When to avoid

• you need actively maintained tooling - the project was deprecated in June 2025 and receives no updates or support
• you are starting a new model optimization project - consider maintained alternatives or Neural Magic's vLLM-based offerings

Facets

library · maturity abandoned

machine-learning deep-learning llm-training deep-learning machine-learning computer-vision python cross-platform model-optimization pruning quantization sparsification pytorch tensorflow onnx deprecated neural-magic natural-language-processing

Member repositories

For agents

markdown · JSON · MCP: product_card(name="neuralmagic/sparseml")