Fine-tune LLMs from one YAML config - train 8B models on 4GB laptop GPUs with layer streaming
A flexible framework for cutting-edge LLM inference and fine-tuning optimizations with CPU-GPU heterogeneous computing
Post-training recipes and SDK for customizing language models with distributed fine-tuning
Fine-tune 600+ LLMs and 300+ MLLMs with PEFT, full-parameter training, and advanced RL algorithms.
Real-time SOTA object detection & segmentation built on DINOv2 transformers
Hands-on programming tutorials for learning large language models from scratch
