MNN Releases

MNN 2.2.0 is a significant release focused on performance optimization and framework extensibility. Key additions include ARMv8.6-A instruction support (smmla/bfmmla) achieving near-theoretical performance gains, a new NNAPI backend for Android NPU/APU/DSP acceleration, and CUDA optimizations. The release also expands training capabilities with new derivative operators and adds support for A16/M2 CPU families.

Oct 31, 2022

MNN 2.1.0 is a significant release focusing on framework generality, performance optimization, and model compression. Key improvements include CUDA performance enhancements via cutlass-based matrix multiplication and Winograd algorithms, Eager mode support in MNN-Express, and Winograd Int8 optimization for quantized convolutions. The release also includes Metal shader compilation improvements and various bug fixes.

Aug 31, 2022

MNN 2.0.0 is a major release featuring significant framework enhancements including expanded operator support (Onnx: 117→158, Torchscripts: 34→163), new MNN-CV and MNN-Numpy modules with 227 combined functions, and substantial CPU/GPU performance improvements through AVX512 and TensorCore optimizations.

Jun 29, 2022

MNN 1.2.0 introduces significant framework enhancements including TorchScript model support for PyTorch migration, ARM BF16 backend for performance gains on mid/low-end devices, and CoreML backend for Apple Neural Engine acceleration. The release also delivers major performance improvements across ARM, X86, and OpenCL backends, along with model compression enhancements through sparse operators and improved offline quantization.

Jun 18, 2021

MNN 1.1.0 is a major release featuring geometric computing - a framework refactoring that separates shape/geometric computation from hardware backends, enabling ~20% GPU performance gains. The release adds TensorRT and CUDA backends, introduces weight-only quantization (4x model size reduction at 8-bit), and adds support for Transformer-based ASR models with Control Flow, Dynamic Shape, and Zero Shape features. Performance optimizations across ARM (10-20% improvement), OpenCL (20-100% improvement), and x86 (on par with OpenVINO) backends are also included.

Nov 5, 2020

MNN 1.0.0 marks a major milestone, transforming MNN from an inference-only engine into a full deep learning framework with training capabilities. This release introduces the Express API for dynamic graph construction, training, and quantization-aware training (QAT), along with significant performance improvements across Arm V8.2, Arm64, and x86 platforms, and a beta Python API.

May 8, 2020