TheBloke/guanaco-13B-HF
TheBloke/guanaco-13B-HF is a 13 billion parameter instruction-tuned language model, created by Tim Dettmers and converted to fp16 Hugging Face format by TheBloke. This model is a merged LoRA of the original Guanaco 13B, designed for general-purpose conversational AI and instruction following. It offers a balance of performance and resource efficiency for various natural language processing tasks.
Loading preview...
Overview
The TheBloke/guanaco-13B-HF model is a 13 billion parameter instruction-tuned language model, originally developed by Tim Dettmers. This specific repository provides the model in a merged LoRA, fp16 Hugging Face format, prepared by TheBloke for broader compatibility and ease of use.
Key Characteristics
- Base Model: Derived from Tim Dettmers' Guanaco 13B.
- Format: Provided as fp16 Hugging Face model files, suitable for direct integration into Hugging Face ecosystems.
- Conversion: The model is a result of merging the original LoRA (Low-Rank Adaptation) weights into the base model, then saving it in the standard fp16 Hugging Face format.
- Instruction-Tuned: Optimized for following instructions and engaging in conversational interactions.
Usage and Availability
This model is part of a series of conversions by TheBloke, which also includes:
- 4-bit GPTQ models for GPU inference.
- 4-bit, 5-bit, and 8-bit GGML models for CPU and GPU inference.
This guanaco-13B-HF version is the merged, unquantized fp16 model, offering a balance between precision and performance for developers seeking a standard Hugging Face implementation.