Les fonctionnalités principales de ag14774/diffdist sont : Outils de développement, More to explore, PyTorch Utilities.
Les alternatives open-source à ag14774/diffdist incluent : mariogeiger/hessian — hessian in pytorch. lyken17/pytorch-opcounter — pytorch-OpCounter is a profiling utility for PyTorch neural networks designed to quantify model efficiency by… kevinmusgrave/pytorch-metric-learning — PyTorch Metric Learning is an open-source library for training neural networks to produce similarity-preserving… awentzonline/pytorch-cns. avelino/awesome-go — This project serves as a comprehensive language ecosystem index, functioning as a centralized, community-curated… adrientorris/awesome-blazor.
pytorch-OpCounter is a profiling utility for PyTorch neural networks designed to quantify model efficiency by calculating floating point operations and multiply-accumulate counts. It functions as a complexity analyzer to measure the computational cost and theoretical workload of different model architectures. The tool allows for the definition of custom operation counting rules to support third-party modules not covered by default. It uses forward hooks to intercept module calls and recursive traversal of the module tree to aggregate operations across child sub-modules. The project provides
PyTorch Metric Learning is an open-source library for training neural networks to produce similarity-preserving embedding spaces. It provides a modular framework where interchangeable loss functions, mining strategies, and evaluation tools can be composed to learn representations that map similar items to nearby points and dissimilar items to distant points in the embedding space. The library distinguishes itself through a highly configurable architecture that separates concerns across several interchangeable components. Users can assemble custom loss functions from pluggable distance metrics
