sharpbai/llama-13b-hf
sharpbai/llama-13b-hf is a 13 billion parameter auto-regressive language model based on the Transformer architecture, developed by Meta AI's FAIR team. This version is a re-packaged LLaMA-13B model, originally trained between December 2022 and February 2023, and is intended primarily for research in large language models. It excels at common sense reasoning, reading comprehension, and natural language understanding tasks, with a focus on English language performance.
Loading preview...
Model Overview
sharpbai/llama-13b-hf is a 13 billion parameter LLaMA model, originally developed by Meta AI's FAIR team. This specific repository provides a re-packaged version of the yahma/llama-13b-hf model, optimized for convenient parallel downloads with 650MB file chunks. It is an auto-regressive language model built on the Transformer architecture, trained between December 2022 and February 2023.
Key Capabilities & Performance
- Research Focus: Primarily intended for research in large language models, including exploring applications like question answering and natural language understanding, and evaluating model biases and limitations.
- Reasoning Tasks: Demonstrates strong performance on common sense reasoning benchmarks, achieving 78.1% on BoolQ, 80.1% on PIQA, and 79.2% on HellaSwag for the 13B variant.
- Multilingual Data: Trained on a diverse dataset including CCNet, C4, GitHub, Wikipedia, Books, ArXiv, and Stack Exchange, covering 20 languages, though performance is expected to be strongest in English.
Intended Use Cases
- Academic Research: Ideal for researchers studying LLM capabilities, limitations, bias mitigation, and harmful content generation.
- Foundation Model: Serves as a base model for further fine-tuning and risk evaluation in downstream applications.
Important Considerations
- License: Operates under a non-commercial bespoke license, requiring users to have been granted access to the original LLaMA model weights.
- Limitations: As a foundational model, it has not been trained with human feedback and may generate toxic, offensive, or incorrect information. It is not intended for applications critical to human life without extensive risk mitigation.