This repository is the official companion code for the book "Build a Reasoning Model (From Scratch)" by Sebastian Raschka (Manning, 2025). It is primarily a data_research codebase focused on building, evaluating, and training reasoning LLMs from scratch using the Qwen3 0.6B model as a base. The code covers inference-time scaling (temperature/top-p sampling, self-consistency), reinforcement learning from verifiable rewards (GRPO/RLVR), self-refinement, and knowledge distillation from larger teachers (DeepSeek-R1, Qwen3 235B). It includes a full custom Qwen3 implementation with KV cache, RoPE, and grouped-query attention, plus training scripts, notebooks, benchmark evaluations (MATH-500, MMLU), and Hugging Face export adapters. While educational in nature, it demonstrates production-quality engineering for research experiments. The repository is hybrid: it provides a reusable Python package for text generation and training, but its dominant character is data_research due to its notebook-driven, experimental, and benchmark-focused workflow.
AudioX-Turbo is a unified framework for efficient anything-to-audio generation (text, video, audio conditions → audio) using Distribution Matching Distillation (DMD) adapted to flow matching and adversarial training. It supports six core tasks: Text-to-Audio, Text-to-Music, Video-to-Audio, Video-to-Music, Text+Video-to-Audio, and Text+Video-to-Music. The system includes training scripts, an interactive Gradio demo, and a modular architecture with interchangeable components (attention, bottlenecks, etc.) via a factory pattern. Key innovations enable high-quality 4-step audio generation, reducing computational cost by up to 25× compared to multi-step diffusion models.
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.
This repository is a research codebase for implementing and experimenting with dense retrieval techniques. It provides implementations of multiple advanced retrieval models including DPR, CITADEL, COIL, ColBERT, SPLADE, and their variants, along with data augmentation strategies like DRAGON and DRAMA. The codebase enables researchers to train, evaluate, and compare different retrieval approaches on standard benchmarks like Natural Questions, MS MARCO, and conversational datasets. It features a modular architecture with separate components for data processing, model definitions, training tasks, and evaluation, configured through Hydra for easy experimentation.
SDPO (Self-Distilled Policy Optimization) is a framework for training large language models using reinforcement learning algorithms, built upon the verl (Volcano Engine Reinforcement Learning) foundation. The system provides a modular architecture for implementing PPO, GRPO, SDPO, and other RL algorithms with distributed training capabilities via Ray. It integrates with inference engines like vLLM and SGLang, supports multiple model architectures (Llama, Qwen with Megatron optimization), and includes reward modeling for domains like mathematics and coding. The framework features extensible components for reward managers, tools, and agent interactions, with experimental capabilities for fully asynchronous training and multi-turn reasoning. While actively maintained, the system's complexity and extensive dependency ecosystem present moderate technical drift risks.
This repository implements a meta-skill (ex-skill) that generates digital personas of former romantic partners by analyzing chat histories from WeChat and iMessage. It is designed to be consumed by AI agent runtimes like OpenClaw, Claude Code, or DeepSeek Harness via the AgentSkills standard. The system guides users through conversational intake of relationship details (astrology, MBTI, attachment style), automatically decrypts and extracts WeChat/iMessage chat databases, analyzes communication patterns using LLM prompt templates, and writes a structured persona skill that can be interactively chatted with. The codebase consists of a SKILL.md orchestrator, six detailed LLM prompts (intake, analysis, persona builder, merger, correction handler), four Python tools (database decryption, message parsing, skill file management, version management), and an example persona. It is a specialized tool for cyber-immortality and digital persona creation, blending reverse engineering with psychological profiling.