Description GGUF Format model files for this project.
From @bogdan1: Llama-2-7b-base fine-tuned on the Chitanka dataset and a dataset made of scraped news comments dating mostly from 2022/2023. Big Thank you :)
About GGUF
Introduction:
GGUF was introduced by the llama.cpp team on August 21st, 2023, as a replacement for GGML, which is no longer supported. GGUF is a successor file format to GGML, GGMF, and GGJT. It is designed to provide a comprehensive solution for model loading, ensuring unambiguous data representation while offering extensibility to accommodate future enhancements. GGUF eliminates the need for disruptive changes, introduces support for various non-llama models such as falcon, rwkv, and bloom, and simplifies configuration settings by automating prompt format adjustments.
Key Features:
No More Breaking Changes: GGUF is engineered to prevent compatibility issues with older models, ensuring a seamless transition from previous file formats like GGML, GGMF, and GGJT.
Support for Non-Llama Models: GGUF extends its compatibility to a wide range of models beyond llamas, including falcon, rwkv, bloom, and more.
Streamlined Configuration: Say goodbye to complex settings like rope-freq-base, rope-freq-scale, gqa, and rms-norm-eps. GGUF simplifies the configuration process, making it more user-friendly.
Automatic Prompt Format: GGUF introduces the ability to automatically set prompt formats, reducing the need for manual adjustments.
Extensibility: GGUF is designed to accommodate future updates and enhancements, ensuring long-term compatibility and adaptability.
Enhanced Tokenization: GGUF features improved tokenization code, including support for special tokens, which enhances overall performance, especially for models using new special tokens and custom prompt templates.
Supported Clients and Libraries:
GGUF is supported by a variety of clients and libraries, making it accessible and versatile for different use cases:
- Downloads last month
- 92