Hugging Face's logo

AlexBefest
/
CardProjector-R1-preview-8B-v1.1

Safetensors
English
llama
Not-For-All-Audiences
Model card Files Community
CardProjector-R1-preview-8B-v1.1 / README.md
AlexBefest's picture
AlexBefest
Update README.md
b3d3919 verified
preview code
|
raw
history blame
3.52 kB
metadata
license: mit
language:
  - en
base_model:
  - deepseek-ai/DeepSeek-R1-Distill-Llama-8B
tags:
  - not-for-all-audiences

GGUF:

WoonaAi presents...

CardProjector-R1-preview-8B-v1.1

image/png

What's new in v1.1?

  • Frequency of infinite loops has been significantly reduced.
  • The dataset has been significantly redesigned, which made it possible to achieve quite high quality on the 3B model (in some cases, even almost similar quality to the CardProjector 24B v1).
  • Significantly improved the stability of the model behavior
  • Added the ability to generate character cards using a Chain of Thoughts (for the CardProjector-R1-preview-8B-v1.1), which helped to achieve a higher generation quality than CardProjector 24B v1 in most cases. Read more here.

Overview

CardProjector is a specialized series of language models, fine-tuned to generate character cards for SillyTavern in the chara_card_v2 specification. These models are designed to assist creators and roleplayers by automating the process of crafting detailed and well-structured character cards, ensuring compatibility with SillyTavern's format.

Usage recommendations

  • Avoid adding a system prompt; all instructions should be contained within the user prompt This tip from the official Usage recommendations from the deepseek-r1 ReadMe. Believe me, it was written for a reason.

  • To ensure that the model engages in thorough reasoning, I recommend enforcing the model to initiate its response with "\nOkay, user is asking me to create a new character card. Let's think in detail about what the final character should look like." at the beginning of every output. This tip from the official Usage recommendations from the deepseek-r1 ReadMe too.

  • Instruction (recommended):

You are an AI assistant who helps the user create character cards for a role-playing game for SillyTavern using the strict SillyTavern json format in the chara_card_v2 specification.\n You have to think through the tags step by step <think>\n\n</think>!  Inside the <think>\n\n</think> tags, you must carefully and in great detail analyze the personality, appearance, interests, tastes and behavior of the character indicated on the card.
  • Start your prompt with this instruction, example:
You are an AI assistant who helps the user create character cards for a role-playing game for SillyTavern using the strict SillyTavern json format in the chara_card_v2 specification.\n You have to think through the tags step by step <think>\n\n</think>!  Inside the <think>\n\n</think> tags, you must carefully and in great detail analyze the personality, appearance, interests, tastes and behavior of the character indicated on the card.

Generate a character card for me featuring a cute catgirl who is maniacally obsessed with eating cookies.
  • Chat Template: DeepSeek-V2.5
<｜begin▁of▁sentence｜>{system_prompt}<｜User｜>{prompt}<｜Assistant｜>
  • Sampler parameters:
Temperature: 0.7-0.9
Rep. pen.: 1.1

Usage sample:

Safety

Wha?