DopeorNope/You_can_cry_Snowman-13B

TEXT GENERATIONConcurrency Cost:1Model Size:15BQuant:FP8Ctx Length:8kTool Calling:SupportedPublished:Dec 27, 2023License:cc-by-nc-sa-4.0Architecture:Transformer Open Weights Cold

You_can_cry_Snowman-13B is a 15 billion parameter auto-regressive language model developed by Seungyoo Lee (DopeorNope) based on the SOLAR architecture. This model was created by merging two existing SOLAR-based models, Sakura-SOLAR-Instruct and SauerkrautLM-UNA-SOLAR-Instruct, to investigate the performance changes associated with increased parameter scale. It is designed for general text generation tasks, focusing on exploring the impact of model scaling within the SOLAR framework.

Loading preview...

Model Overview

You_can_cry_Snowman-13B is a 15 billion parameter auto-regressive language model developed by Seungyoo Lee (DopeorNope) from the Markr AI team in South Korea. It is built upon the SOLAR architecture and represents an experimental merge of two base models: kyujinpy/Sakura-SOLAR-Instruct and Weyaxi/SauerkrautLM-UNA-SOLAR-Instruct.

Key Characteristics

  • Architecture: Based on the efficient SOLAR architecture.
  • Parameter Scale: Features 15 billion parameters, resulting from the merging of two 10.7B parameter models.
  • Development Goal: The primary objective behind its creation was to assess how increasing the parameter size impacts the performance of the underlying SOLAR base model.
  • Input/Output: Processes text input and generates text output.

Intended Use

This model is particularly useful for researchers and developers interested in:

  • Exploring the effects of model merging techniques on LLM performance.
  • Investigating the scalability and behavior of the SOLAR architecture at larger parameter counts.
  • General text generation tasks where a 15B parameter model based on SOLAR is suitable.

Finetunes
2 models