LLM Compressor v0.10: Faster compression with distributed GPTQ

The LLM Compressor v0.10 introduces enhanced compression techniques for large language models, featuring distributed quantization for faster performance across multiple GPUs, improved accuracy with numerical optimizations, and efficient offloading mechanisms for large models. Key highlights include multi-GPU support, better memory management, and new quantization formats aimed at enhancing the usability of model compression. This release not only accelerates training times but also broadens access to model compression on less powerful hardware, making it a significant advancement for developers working with large LLMs.