منصوبے کے بارے میں

tropical-gemm is a Rust library designed for high-performance tropical (or max-plus/min-plus) matrix multiplication. It is particularly useful for algorithms in graph theory, such as shortest and longest path problems, and for tropical neural networks. Key capabilities include: - Multiple Semirings: Support for MaxPlus, MinPlus, and MaxMul algebraic structures. - Hardware Acceleration: Optimized CPU implementations using SIMD (AVX-512, AVX2, SSE4.1, NEON) and a GPU-accelerated CUDA backend. - Argmax Tracking: Built-in support for tracking argmax values, which is essential for backpropagation in tropical neural networks. - Python Integration: Provides bindings for both NumPy and PyTorch, allowing for seamless integration into Python-based machine learning workflows with automatic gradient computation. The library is written in Rust for performance and safety, and it offers a clean API for both Rust and Python developers.