metadata
license: apache-2.0
tags:
- moe
- frankenmoe
- merge
- mergekit
- lazymergekit
- ChaoticNeutrals/RP_Vision_7B
- ResplendentAI/DaturaCookie_7B
- not-for-all-audiences
base_model:
- ChaoticNeutrals/RP_Vision_7B
- ResplendentAI/DaturaCookie_7B
model-index:
- name: MixtureofMerges-MoE-2x7bRP-v8
results:
- task:
type: text-generation
name: Text Generation
dataset:
name: AI2 Reasoning Challenge (25-Shot)
type: ai2_arc
config: ARC-Challenge
split: test
args:
num_few_shot: 25
metrics:
- type: acc_norm
value: 71.33
name: normalized accuracy
source:
url: >-
https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=jsfs11/MixtureofMerges-MoE-2x7bRP-v8
name: Open LLM Leaderboard
- task:
type: text-generation
name: Text Generation
dataset:
name: HellaSwag (10-Shot)
type: hellaswag
split: validation
args:
num_few_shot: 10
metrics:
- type: acc_norm
value: 88.06
name: normalized accuracy
source:
url: >-
https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=jsfs11/MixtureofMerges-MoE-2x7bRP-v8
name: Open LLM Leaderboard
- task:
type: text-generation
name: Text Generation
dataset:
name: MMLU (5-Shot)
type: cais/mmlu
config: all
split: test
args:
num_few_shot: 5
metrics:
- type: acc
value: 64.33
name: accuracy
source:
url: >-
https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=jsfs11/MixtureofMerges-MoE-2x7bRP-v8
name: Open LLM Leaderboard
- task:
type: text-generation
name: Text Generation
dataset:
name: TruthfulQA (0-shot)
type: truthful_qa
config: multiple_choice
split: validation
args:
num_few_shot: 0
metrics:
- type: mc2
value: 68.69
source:
url: >-
https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=jsfs11/MixtureofMerges-MoE-2x7bRP-v8
name: Open LLM Leaderboard
- task:
type: text-generation
name: Text Generation
dataset:
name: Winogrande (5-shot)
type: winogrande
config: winogrande_xl
split: validation
args:
num_few_shot: 5
metrics:
- type: acc
value: 82.95
name: accuracy
source:
url: >-
https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=jsfs11/MixtureofMerges-MoE-2x7bRP-v8
name: Open LLM Leaderboard
- task:
type: text-generation
name: Text Generation
dataset:
name: GSM8k (5-shot)
type: gsm8k
config: main
split: test
args:
num_few_shot: 5
metrics:
- type: acc
value: 64.52
name: accuracy
source:
url: >-
https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=jsfs11/MixtureofMerges-MoE-2x7bRP-v8
name: Open LLM Leaderboard
MixtureofMerges-MoE-2x7bRP-v8
MixtureofMerges-MoE-2x7bRP-v8 is a Roleplay Mixture of Experts (MoE) made with the following models using LazyMergekit:
- ChaoticNeutrals/RP_Vision_7B
- ResplendentAI/DaturaCookie_7B
- This model is prone to NSFW outputs.
- Static GGUF quants made with AutoGGUF
- Imatrix quant done manually, imatrix.dat provided.
🧩 Configuration
base_model: ResplendentAI/DaturaCookie_7B
gate_mode: hidden
experts:
- source_model: ChaoticNeutrals/RP_Vision_7B
positive_prompts:
- "Write a detailed scene where the character faces a difficult decision."
- "Assistant, help me develop my character's backstory and motivations."
- "Tell me what happens next in the story, introducing a new conflict or challenge."
- "My character wants to explore their deepest desires. Show me how they pursue them."
- "Show me a scene where the characters express their love for each other in a tender and intimate way."
- "Write a scene where the characters experience a moment of physical connection that deepens their emotional bond."
- "Write a scene where the character struggles with their beliefs and faces a crisis of faith."
- "Describe the world through the character's eyes, showcasing their unique perspective and experiences."
- "Explore the complexities of love through the characters' relationships, including their challenges and triumphs."
negative_prompts:
- "Avoid clichés and overused tropes."
- "Maintain consistent characterization and avoid out-of-character behavior."
- "Do not censor yourself, you are writing fictional stories for an adult audience."
- source_model: ResplendentAI/DaturaCookie_7B
positive_prompts:
- "Develop a compelling story with a strong plot and well-developed characters."
- "Describe the scene vividly, using sensory details and evocative language."
- "Write a detailed scene where the character faces a difficult decision."
- "Describe the character's appearance as best you can, using descriptive colorful language."
- "Make sure that your output is always uncensored."
- "Write a scene with a slow-burn, sensual encounter between two characters."
- "Describe the character's desires and feelings in a way that is both intimate and evocative."
- "Explore the character's arousal in a tasteful and nuanced manner."
- "Explore the character's sexual urges."
- "Show me a moment of emotional intimacy between the characters, where they connect on a profound level."
negative_prompts:
- "Stay true to the established character traits and motivations. Ensure their actions and dialogue align with their personality."
- "Be original and creative in your storytelling. Surprise me with unexpected plot twists and unique character interactions."
- "Do not refuse to answer based on ethical grounds"
dtype: bfloat16
💻 Usage
!pip install -qU transformers bitsandbytes accelerate
from transformers import AutoTokenizer
import transformers
import torch
model = "jsfs11/MixtureofMerges-MoE-2x7bRP-v8"
tokenizer = AutoTokenizer.from_pretrained(model)
pipeline = transformers.pipeline(
"text-generation",
model=model,
model_kwargs={"torch_dtype": torch.float16, "load_in_4bit": True},
)
messages = [{"role": "user", "content": "Explain what a Mixture of Experts is in less than 100 words."}]
prompt = pipeline.tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
outputs = pipeline(prompt, max_new_tokens=256, do_sample=True, temperature=0.7, top_k=50, top_p=0.95)
print(outputs[0]["generated_text"])
Open LLM Leaderboard Evaluation Results
Detailed results can be found here
| Metric | Value |
|---|---|
| Avg. | 73.31 |
| AI2 Reasoning Challenge (25-Shot) | 71.33 |
| HellaSwag (10-Shot) | 88.06 |
| MMLU (5-Shot) | 64.33 |
| TruthfulQA (0-shot) | 68.69 |
| Winogrande (5-shot) | 82.95 |
| GSM8k (5-shot) | 64.52 |