Mimeo is a Python CLI application that transforms a named expert's public body of work (essays, talks, interviews, podcasts, papers, etc.) into a structured, source-verified Agent Skill file (SKILL.md) and/or AGENTS.md, designed for AI coding agents to adopt the expert's reasoning style. The pipeline orchestrates identity resolution, source discovery via the Parallel API, content fetching (web, YouTube transcripts, optional audio transcription), LLM-based distillation and clustering (via OpenRouter), quote verification against source text, structured skill authoring, critique/refinement loops, and optional avatar generation. The output is a curated set of markdown files (principles, frameworks, mental models, anti-patterns, heuristics, quotes) organized per expert in the `output/` directory. The codebase is well-architected with async stages, a fingerprint-based caching layer, SSRF-safe HTTP helpers, prompt injection mitigations, and extensive offline tests using fake LLM/Parallel clients. While the primary purpose is the application tool, the generated output forms a reusable knowledge base of expert reasoning patterns, giving the repo a hybrid character.
DistillKit is a production-ready toolkit for knowledge distillation of large language models, supporting both online and offline workflows. It features advanced logit compression techniques (polynomial approximation + quantization + bit-packing) that enable efficient offline distillation at scale. The toolkit provides composable loss functions (KL, JSD, TVD, ranking losses, hidden state alignment) and integrates with Hugging Face Transformers, TRL, and Accelerate. It powers Arcee AI's distilled model releases and includes CLI tools for easy experimentation.
PMetal is a comprehensive machine learning framework specifically optimized for Apple Silicon hardware. It provides a complete suite for LLM training, fine-tuning, inference, and deployment through multiple interfaces (CLI, TUI, GUI, Python SDK, and REST API). The framework leverages Apple-specific technologies like Metal GPU, ANE, and UltraFusion for performance optimization while supporting standard ML workflows including LoRA/QLoRA fine-tuning, knowledge distillation, model merging, and quantization to formats like GGUF. The architecture follows a modular workspace design with over 20 specialized crates handling different aspects of the ML lifecycle, unified through a job specification system and configuration framework.
OpenCrew is a multi-agent operating system designed to extend OpenClaw into a manageable AI team. It enables users to create specialized agents (Chief of Staff, CTO, Builder, CIO, Knowledge Officer, Operations, Research) that collaborate through messaging platforms like Slack, Discord, and Feishu. The system implements structured collaboration through role-based channels, thread-based task isolation, and agent-to-agent communication protocols. Key features include autonomy levels (L0-L3), task classification (QAPS), two-mode agent collaboration (delegation and discussion), and three-layer knowledge distillation. The application is actively maintained with recent updates focusing on cross-agent collaboration capabilities.
This repository is a curated collection of Jupyter notebooks and Python scripts that demonstrate fine-tuning, inference, quantization, and optimization techniques for state-of-the-art vision and multimodal models from the Hugging Face ecosystem. It covers a wide range of tasks including visual question answering, object detection, image classification, video understanding, multimodal retrieval, and preference alignment (DPO). The code is primarily educational, designed to be run interactively in Google Colab, with each notebook or script serving as a self-contained tutorial for a specific model or technique. Notable patterns include LoRA/QLoRA fine-tuning, knowledge distillation via a custom ImageDistilTrainer, any-to-any embedding for retrieval, and video frame downsampling for multimodal input. The repository is not a production library or application; it is a learning resource for engineers and researchers exploring vision-language models.
This repository implements the EViT (Expediting Vision Transformers via Token Reorganizations) method for accelerating Vision Transformers by dynamically reducing the number of tokens processed in intermediate layers through token discarding or fusion. It provides training, evaluation, and visualization scripts for ImageNet classification, leveraging the timm library for core utilities. The codebase includes dynamic token rate scheduling via cosine annealing, throughput and MAC measurement tools, and knowledge distillation support. While the core EViT model is specialized, the repository contains reusable components such as helper functions for scheduling and efficiency metrics (helpers.py) and a knowledge distillation loss wrapper (losses.py). These components are self-contained and demonstrate non-obvious solutions to common problems in efficient deep learning. However, the codebase suffers from technical drift due to outdated dependencies (e.g., torch 1.9.0, timm 0.4.12) and appears inactive based on the commit history.