Aeala/GPT4-x-Alpasta-13b

TEXT GENERATIONConcurrency Cost:1Model Size:13BQuant:FP8Ctx Length:4kPublished:May 6, 2023Architecture:Transformer0.0K Cold

Aeala/GPT4-x-Alpasta-13b is a 13 billion parameter language model created by Aeala, merging ChanSung's AlpacaGPT4-LoRA-13B-elina with dvruette's llama-13b sft do2 finetune. This model combines elements from instruction-tuned Alpaca-GPT4 and a fine-tuned LLaMA base. It is designed for general language generation tasks, leveraging its merged architecture for diverse capabilities.

Loading preview...

Overview

Aeala/GPT4-x-Alpasta-13b is a 13 billion parameter language model developed by Aeala. It is a merge of two distinct models: ChanSung's AlpacaGPT4-LoRA-13B-elina and dvruette's llama-13b sft do2 finetune. This combination aims to leverage the instruction-following capabilities of the Alpaca-GPT4-LoRA variant with the foundational strengths of a fine-tuned LLaMA 13B model.

Key Capabilities

  • Instruction Following: Inherits capabilities from the AlpacaGPT4-LoRA component, suggesting proficiency in understanding and executing user instructions.
  • General Text Generation: Benefits from the LLaMA 13B base, providing a strong foundation for various language generation tasks.
  • Merged Architecture: Represents an experimental blend of different fine-tuning approaches, potentially offering a unique balance of performance characteristics.

Good For

  • Exploratory Use Cases: Suitable for developers interested in experimenting with merged model architectures.
  • General Purpose Language Tasks: Can be applied to a range of applications requiring text generation or instruction adherence.
  • Research and Development: Provides a base for further fine-tuning or analysis of combined model strengths.