Automodel provides out-of-the-box Hugging Face integration for training and fine-tuning LLMs and VLMs at scale. It implements PyTorch DTensor SPMD parallelism with custom kernels and optimized training recipes. The library supports a wide range of foundation models including Gemma, Llama, Qwen, DeepSeek, Mistral, and GLM families, offering day-zero compatibility with Hugging Face model hubs for distributed pretraining and fine-tuning workflows.