Skip to content

Release Notes

v1.0 - Mono Quant Initial Release (2026-02-03)

Features

Core Quantization

  • ✅ INT8 quantization with per-channel scaling
  • ✅ INT4 quantization with group-wise scaling (group_size=128)
  • ✅ FP16 quantization for memory reduction
  • ✅ Dynamic quantization (no calibration data required)
  • ✅ Static quantization with calibration

Calibration

  • ✅ MinMaxObserver (default, fast)
  • ✅ MovingAverageMinMaxObserver (robust, EMA smoothing)
  • ✅ HistogramObserver (outlier-aware, KL divergence)
  • ✅ Calibration data from tensors or DataLoader

User Interface

  • ✅ Unified quantize() Python API
  • ✅ QuantizationResult with .save() and .validate() methods
  • ✅ CLI with git-style subcommands (monoquant)
  • ✅ Progress bars with CI/TTY auto-detection

Serialization

  • ✅ PyTorch format (.pt/.pth) support
  • ✅ Safetensors format support
  • ✅ Metadata preservation (bits, scheme, scales, zero-points)
  • ✅ Model dequantization back to FP32

Validation

  • ✅ SQNR (signal-to-quantization-noise ratio) computation
  • ✅ Model size comparison
  • ✅ Load testing (round-trip validation)
  • ✅ Accuracy warnings for aggressive quantization

Advanced Features

  • ✅ Model-agnostic design (any PyTorch model)
  • ✅ Layer skipping for INT4 (protects sensitive layers)
  • ✅ Symmetric and asymmetric quantization schemes
  • ✅ Custom exception hierarchy with actionable suggestions
  • ✅ Zero-point clamping to prevent runtime errors

Statistics

  • Requirements delivered: 30/30 (100%)
  • Integration points: 8/8 verified
  • E2E flows: 8/8 working
  • Lines of code: 5,228 Python
  • Files: 26 source files
  • Technical debt: None identified

Dependencies

Required: - Python >= 3.11 - torch >= 2.0 - numpy >= 1.24

Included: - safetensors >= 0.4 (Safetensors format) - click >= 8.1 (CLI) - tqdm >= 4.66 (progress bars)

Optional: - mkdocs >= 1.6.0 (documentation) - mkdocs-material >= 9.7.0 (documentation theme) - mkdocstrings[python] >= 1.0.0 (API documentation)

Documentation

  • Installation guide
  • Quick start tutorial
  • Basic usage examples
  • CLI reference
  • API documentation
  • User guide
  • Examples (dynamic INT8, static INT4, custom observer, CI/CD)

Known Limitations

  • CLI does not support loading calibration data from files (use Python API)
  • INT4 quantization requires calibration data (no dynamic INT4)
  • No quantization-aware training (QAT) - build-phase only
  • No ONNX/TFLite export (use dedicated conversion tools)

Future Enhancements (v2)

  • Genetic optimization for quantization parameters
  • Experiment tracking and logging
  • Mixed precision (different bits per layer)
  • LLM.int8() style outlier detection
  • Automatic layer sensitivity analysis

Older Versions

No older versions. v1.0 is the initial release.