Hamzah-Asadullah/GenericRPV1-7B-GGUF

This model is available on Ollama: ollama run Hamzah-Asadullah/GenericRPV1-7B

TL;DR: GRP (GenericRP) is a merge of four models, three of which designed for further enhancing the model's "academic" and day-to-day abilities, one designed to be less censored and extremely good in roleplaying. All models used are licensed under Apache2.0.

Since that's what I do in my free-time, I don't get paid for it. Training stable models requires compute, you can help me buy more here ٩( ๑╹ ꇴ╹)۶

GRP is a 7B parameter model designed as a general model, which includes strong instruction-following, strong performance in STEM & programming, but also roleplaying which most models lack.
This is a merge. A simple linear merge method was used, and following models were included in that merge, each with a weight of 0.25:

• jeffmeloy/Qwen2.5-7B-nerd-uncensored-v1.0: The uncensored model with strong roleplaying abilities, which therefore lacked normal instruction following and had large gaps in Mathematics and other STEM-related topics
• Spestly/Athena-1-7B: A model designed for super instruction-following and exceptional knowledge-dense parameters
• prithivMLmods/Viper-Coder-HybridMini-v1.3: A model designed for enhanced coding abilities
• rombodawg/Rombos-LLM-V2.5-Qwen-7b: A simple model which outperforms its base-model: Qwen2.5-7B-Instruct

All of those models use the ChatML format, but experimentation is key ;) Feel free to report your findings here.

XeTute Technologies

XeTute Technologies is an unofficial Pakistani organisation created by Hamzah Asadullah.

Links

HuggingFace GitHub

Buy me a Coffee Apex Webpage

Pakistan

Pakistan is a country in South-Asia known for its rich culture despite the British, its stunning landscape, and PAF (Pakistan Armed Forces), its military. Long live the Islamic Republic of Pakistan.