Llama Language Models: Latest Updates and Availability
Model Sizes and Training Details
Google AI has recently announced the release of four sizes of Llama language models, with 7B, 13B, 34B, and 70B parameters. All models are trained with a global batch-size of 4M tokens, and the pretraining data consists of WEB Token counts only.
Availability on HuggingFace
All three model sizes (7B, 13B, and 70B) are now available for download on HuggingFace. The custom model "TheBlokeLlama-2-70B-chat-GPTQ" is also available for download under the "Download custom model or LoRA" section.
Open Source and Commercial Use
The Llama language models are open source and free to use for research and commercial purposes. This includes the model weights and starting code for pre-trained and fine-tuned models.
Komentar