Awesome GitHub RepositoriesLoss Function Customization

This project is a modular research toolkit designed for developing, training, and evaluating deep learning models for object detection, segmentation, and video instance tracking. It provides a flexible training engine that manages complex neural network execution, including distributed training, custom lifecycle hooks, and weight optimization. The framework is built around a hierarchical configuration system that allows users to define architectures, data pipelines, and training hyperparameters through composable, inheritable files. The project distinguishes itself through its highly modular

MMSegmentation is an open-source semantic segmentation toolbox built on PyTorch that provides a modular, configurable framework for building, training, evaluating, and deploying segmentation models. At its core, it offers a config-driven pipeline that assembles training, evaluation, and inference workflows by parsing hierarchical configuration files, with a modular component registry that enables plug-and-play composition of neural network modules, optimizers, datasets, and metrics. The framework supports the full model lifecycle through a unified runner interface that controls training, testi

Vowpal Wabbit is an open-source machine learning system designed for online learning, where models update incrementally from streaming data without requiring full retraining. It provides a reduction-based learning framework that composes complex tasks from simpler algorithms, and includes a feature hashing trick that maps unbounded feature names into a fixed-size vector space to keep memory usage constant regardless of dataset size. The system supports distributed training across a cluster using an allreduce protocol for synchronized updates, and offers an active learning query strategy that s

sd-scripts is a suite of utilities designed for fine-tuning generative models, preprocessing datasets, and converting model weights. It provides a collection of scripts for executing Stable Diffusion training through methods such as DreamBooth, textual inversion, and full fine-tuning, alongside a framework for creating and managing Low-Rank Adaptation weights. The project features specialized capabilities for model weight conversion between different architectures and precision formats. It includes tools for merging adaptation weights into base models, extracting weights from trained models,

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

MMDetection3D is an open-source toolbox for 3D perception, providing a unified framework for detecting and segmenting objects in three-dimensional environments. It supports a range of core tasks including monocular 3D object detection from single camera images, LiDAR-based 3D object detection from raw point clouds, and multi-modal fusion that combines camera images with LiDAR data. The toolbox also covers point cloud semantic segmentation, assigning class labels to every point in a scan for scene understanding. The project distinguishes itself through a config-driven pipeline that orchestrate

MMF is a modular framework for building, training, and evaluating vision-and-language models. It provides a configuration-driven experiment system where model, dataset, and training parameters are defined through composable YAML files, alongside a curated model zoo of pretrained checkpoints for state-of-the-art multimodal architectures. The framework includes a multimodal dataset loader that downloads, processes, and batches vision-and-language data, and a vision-language model trainer supporting distributed training, mixed precision, and checkpoint-based resumption. The framework distinguish

SLIME is a distributed reinforcement learning framework for large language model post-training that bridges Megatron training with SGLang inference servers. It orchestrates scalable RL loops across GPU clusters, decoupling training and inference into independent processes that communicate over HTTP and NCCL for independent scaling and fault tolerance. The system supports multi-agent reinforcement learning workflows with parallel agent instances, customizable rollout strategies, and personalized agent serving that improves models from prior conversations without disrupting API serving. The fra

AdalFlow este un framework de agenți AI autonomi și o bibliotecă de aplicații LLM concepută pentru construirea de fluxuri de lucru modulare. Servește ca interfață agnostic-model și orchestrator de pipeline-uri RAG, permițând utilizatorilor să dezvolte agenți ReAct care utilizează raționamentul iterativ și execuția de instrumente externe pentru a rezolva sarcini complexe. Proiectul se distinge printr-un sistem de optimizare a prompt-urilor care utilizează gradient descent textual pentru a rafina automat template-urile de prompt și exemplele few-shot. Acesta tratează feedback-ul modelului ca pe un semnal diferențiabil, permițând o formă de backpropagation pentru LLM-uri pentru a îmbunătăți iterativ calitatea output-ului pe baza metricilor de evaluare. Framework-ul acoperă o suprafață largă de capabilități, inclusiv retrieval-augmented generation cu căutare semantică vectorială și reranking, urmărirea execuției bazată pe span pentru observabilitate și parsare structurată bazată pe schemă. Oferă un strat de comunicare unificat pentru numeroși furnizori de modele proprietare și open-source și suportă conversia funcțiilor Python în interfețe standardizate de instrumente. Sistemul este implementat în Python și se integrează cu MLflow pentru urmărirea și analiza fluxurilor de lucru.

TextGrad is a differentiable text optimization library and framework designed for simulated language model backpropagation. It functions as a textual gradient engine that treats language model feedback as gradients to iteratively refine prompts and unstructured text variables. The system utilizes a computation graph to trace errors from a defined loss function back to input text, allowing it to determine specific improvements. It differentiates itself by implementing natural-language backpropagation and gradient aggregation, which merges multiple pieces of textual critique into consolidated i