unsloth/Llama-3.1-8B

The unsloth/Llama-3.1-8B model is an 8 billion parameter instruction-tuned generative language model from the Llama 3.1 family, developed by Meta. It features an optimized transformer architecture with Grouped-Query Attention and a 128K context length, making it suitable for multilingual dialogue and general natural language generation tasks. This version is specifically optimized by Unsloth for faster fine-tuning with reduced memory usage, offering 2.4x faster training and 58% less memory consumption.