A lightweight PyTorch library for efficient model quantization and memory optimization. Perfect for running large language models on consumer hardware.

• 🎯 8-bit & 4-bit quantization primitives
• 💾 Memory-efficient optimizers
• 🚀 LLM.int8() inference support
• 🔄 QLoRA-style fine-tuning
• 🖥️ Cross-platform hardware support

import torch
from Quanta.functional.quantization import quantize_8bit, dequantize_8bit

# Quantize your model
q_tensor, scale, zero_point = quantize_8bit(model_weights)

🚧 Early Development - Currently implementing core quantization features.

MIT License

Inspired by bitsandbytes