LLaMA-Factory is a production-grade, unified framework for fine-tuning, evaluating, and serving 100+ large language and multimodal models. It provides a CLI (`llamafactory-cli`), a Gradio web UI (LLaMA Board), and an OpenAI-compatible API server. The architecture is monolithic but highly modular, built on Hugging Face Transformers/TRL/PEFT with optional integration of vLLM, SGLang, DeepSpeed, FSDP, Megatron Bridge, and hardware-specific backends (CUDA, ROCm, NPU, XPU). Training pipelines support full-parameter, LoRA, QLoRA, freeze-tuning, and alignment methods (DPO, PPO, KTO, etc.) with extensive data processing (multimodal, packing, streaming, template-based formatting). The codebase features a plugin system (v1) for kernels, quantization, distributed backends, and batching strategies, though it maintains a parallel legacy (v0) path. Reusable components include dataset processors, chat template hierarchy, multimodal plugin architecture, custom trainer callbacks, and batching strategy plugins. While the framework is actively maintained and well-tested, it suffers from some technical drift due to a dual-codebase structure, hardcoded model registries, version pinning, and fragile monkey-patching in distributed integrations.
This is the research archive for the Hyperspace P2P network, a distributed AGI system. The repository primarily contains documentation, leaderboards, trained model weights, and configuration files for autonomous research agents. It serves as a durable record of experiments run by agents across 5 research domains (machine learning, search, finance, skills, causes). The main outputs include a collaboratively trained language model (qwen2.5-0.5b-hyperspace-v1) trained by 32 anonymous nodes using DiLoCo + SVD compression, and live leaderboards updated via GitHub Actions. While the repo itself is mostly static artifacts, it references a larger codebase (the Hyperspace CLI and P2P stack) that is not included. The repository is hybrid: primarily a research data archive, but also includes a standalone CLI tool (agentboard) for a blockchain-based message board.