Explore the Hands-On Large Language Models repo, a Jupyter notebook-based practical guide from fundamentals to fine-tuning, designed for hands-on LLM learning on free Colab GPUs.
Alibaba’s Logics-Parsing-v2 converts complex document images into structured HTML, handling formulas, tables, flowcharts, music sheets, and pseudocode with a single model.
Genie Envisioner offers a two-stage training pipeline using video diffusion for robotic manipulation, separating world model adaptation from action policy learning. Here’s how it works and how to get started.
NousResearch’s finetuning-subnet enables continuous, incentivized fine-tuning of LLMs using synthetic data from a separate subnet, pioneering cross-subnet communication in Bittensor.
MAGI implements a multi-round debate protocol among three LLMs to match stronger models’ accuracy via iterative critique and voting. It offers fault tolerance, adaptive escalation, and persona presets.
Magika replaces magic-byte heuristics with a tiny deep learning model for file type detection, achieving ~99% accuracy across 200+ types with 5ms CPU inference.
A curated directory cataloging over 200 production-ready open-source AI projects across the machine learning stack, from training frameworks to self-hosted UIs.
MicroGPT-C uses a deterministic C scaffold to coordinate tiny GPT-2 models, achieving 90%+ accuracy on logic games with 8x memory compression and infinite sequence lengths.
MultiWorld offers a unified framework for multi-agent multi-view video world modeling using a frozen VGGT backbone for implicit 3D understanding. It supports scalable multi-agent control and autoregressive inference.
Explore the EthicalML awesome-production-machine-learning repo, a curated catalog of 200+ open source MLOps tools covering the full production ML lifecycle. Essential for ML engineers building production stacks.
Omni-Diffusion models text, image, and speech tokens jointly via masked discrete diffusion, enabling any-to-any multimodal generation with a single unified model.
onnxmltools is a Python library for converting machine learning models from various frameworks into the ONNX format, enabling interoperability across runtimes and platforms.
OpenGame from CUHK MMLab generates full web games from natural language prompts using a dual-skill LLM architecture that maintains cross-file consistency and integration fixes.
Orion bypasses CoreML to access Apple’s Neural Engine directly via private frameworks, enabling on-device inference and fine-tuning of small LLMs with 8.5x reduced training overhead.
paper2code transforms arxiv papers into Python code with ambiguity auditing and inline citations, prioritizing traceability over completeness in ML implementations.
SimScale provides a sim-real co-training pipeline for autonomous driving planners, combining synthetic simulation data with real-world data to improve robustness and generalization across multiple planner types.
tribev2 offers pretrained models to predict brain responses to videos using cortical mesh modeling. Supports video, text, and audio inputs with easy inference setup.
A curated repo breaking down large language model internals with numeric attention math, tokenization, and transformer architecture, targeting engineers who want to understand LLMs under the hood.
vllm-mlx is a Python inference server for Apple Silicon that supports OpenAI and Anthropic APIs, featuring SSD-tiered KV cache for long-context agents and continuous batching for performance.
The Awesome-LLM-Ensemble repo catalogs research on combining multiple LLMs with a clear three-phase taxonomy: before, during, and after inference ensemble methods.
