Red-Daffodil-7B

Red-Daffodil-7B

Roleplaying focused model based on Mistral-7b-v0.1. Aims to produce natural human-like text, free from slop.

Uses a model based on aselection of LoRA trained on almost entirely human text, which is then merged with opus-v0-7b and AdventurousWinds-Mk2-7b. The resulting model is merged together with Noromaid-7b-v0.2 back into the Mistral base model with LM Cocktail to maintain generalisation and intelligence.

Prompt template:

### Input:
User: {prompt}

### Response:
Character: 

Alpaca prompt template should work fine too.

opus-v0-7b: https://huggingface.co/dreamgen/opus-v0-7b

AdventurousWinds-Mk2-7b https://huggingface.co/PocketDoc/Dans-AdventurousWinds-Mk2-7b

Noromaid: https://huggingface.co/NeverSleep/Noromaid-7b-v0.2/

LM Cocktail: https://github.com/FlagOpen/FlagEmbedding/tree/master/LM_Cocktail