WeiNyn/Llama2-7b-hf: A Foundation Model from Meta

This model is the 7 billion parameter pretrained variant of Meta's Llama 2 family, adapted for Hugging Face Transformers. Llama 2 models are generative text models built on an optimized transformer architecture, trained on a new mix of publicly available online data totaling 2 trillion tokens. The 7B model has a context length of 4k tokens.

Key Capabilities & Features

• Foundation Model: Designed for a wide range of natural language generation tasks, serving as a base for further fine-tuning.
• Performance: Shows improved academic benchmark scores over its Llama 1 counterpart, including 45.3 on MMLU and 14.6 on Math for the 7B variant.
• Training Data: Pretrained on 2 trillion tokens with a data cutoff of September 2022.
• Commercial Use: Licensed for both commercial and research applications, primarily in English.

Intended Use Cases

• Natural Language Generation: Adaptable for various text generation tasks.
• Research: Suitable for academic and commercial research into large language models.
• English-centric Applications: Optimized for use with English language inputs and outputs.