#Model Optimization

Lightning-Bolts extends PyTorch Lightning with additional components such as callbacks and datasets, facilitating applied research and production. It includes features like the Torch ORT Callback for enhanced training and inference speeds on NVIDIA or AMD GPUs. The SparseMLCallback further introduces sparsity in model fine-tuning using the DeepSparse engine. Encouraging community contributions, Lightning-Bolts evolves to support diverse machine learning needs in different domains without exaggeration.

This repository provides the comprehensive 2024 course materials for Efficient Deep Learning Systems taught at HSE University and Yandex School of Data Analysis. Topics covered include core GPU architecture, CUDA API, experiment management, distributed training, and Python web deployment. Detailed week-by-week content supports learning of both theoretical foundations and practical applications, emphasizing real-world examples and project-based studies.

LLM-VM by Anarchy Labs is an open-source server enhancing LLMs with features like live data augmentation and tool usage. It supports open-source models for faster AGI development, cost reduction, and flexibility. Users can locally run models, join a community, and have data control, featuring inference optimization and library callability. The project is currently in beta, with active development and ongoing improvements, making it an attractive option for those seeking actively maintained tools.

ComfyUI provides a cutting-edge, intuitive interface for crafting sophisticated diffusion workflows using a node-based system that eliminates the need for coding. Supporting a wide array of models including SD1.x, SD2.x, SDXL, and more, it efficiently manages resources with GPU usage as low as 1GB VRAM and supports CPU execution. Features include offline operation, asynchronous queuing, and model loading, making it ideal for versatile and modular diffusion processes. ComfyUI ensures seamless cross-platform compatibility and provides numerous workflow examples for scalable applications.

This repository features FastViT, a rapid hybrid vision transformer utilizing structural reparameterization to boost image classification accuracy. Models have been trained on ImageNet-1K and benchmarked for latency on an iPhone 12 Pro via the ModelBench app. The repository includes setup guides for configuring environments, training, and evaluating models, along with scripts for implementation. It provides a varied collection of pre-trained models tailored for classification tasks, including the option for knowledge distillation. Comprehensive dataset preparation and model export instructions are available, making this a versatile tool for tasks ranging from training to fine-tuning in machine learning.

Olive offers a user-friendly tool for optimizing models with awareness of specific hardware needs. It integrates leading techniques in compression, optimization, and compilation to simplify the engineering process, accommodating multiple vendor-specific toolchains for cloud and edge deployments. With a flexible framework, Olive allows for the easy inclusion of new industry innovations, catering to diverse hardware requirements while adhering to constraints such as accuracy and latency. Recent developments highlight enhancements for AI applications and DirectML performance.