OpenLLaMA: Open-Source Reproduction of Meta AI's LLaMA Large Language Model. Run in Google Colab.

In this video, we'll explore OpenLLaMA, an open-source reproduction of Meta AI's LLaMA large language model. We'll load the 7B model in a Google Colab notebook and run a couple of prompts with the HuggingFace Transformers weights. We'll also discuss a strategy for sampling that improves the generated responses.

The creators of OpenLLaMA have released a public preview of the 7B model, trained with 200 billion tokens, and provide PyTorch and Jax weights of pre-trained OpenLLaMA models. They also compare …

colab discuss google huggingface language language model large language model llama meta meta ai notebook prompts sampling strategy transformers video