The use of large language models is varied, with options such as API endpoints from companies like OpenAI or downloaded artifacts from HuggingFace. A unique C/C++ LLM inference engine, llama.cpp, is designed for Apple Silicon GPUs. A step-by-step guide to creating a model starts with linear regression in PyTorch, utilizing features like training data, loss functions like RMSE, and optimization techniques like gradient descent. The process involves feeding data into the model, calculating predictions, and adjusting weights to optimize the model’s output. The end result is a trained model object with updated parameters and reduced loss, ready to make predictions based on new input values.
https://vickiboykis.com/2024/02/28/gguf-the-long-way-around/