README.md

metadata

license: apache-2.0
tags:
  - moe
  - frankenmoe
  - merge
  - mergekit
  - lazymergekit
  - mlabonne/AlphaMonarch-7B
  - FPHam/Karen_TheEditor_V2_STRICT_Mistral_7B
  - SanjiWatsuki/Kunoichi-DPO-v2-7B
  - OmnicromsBrain/NeuralStar-7b-Lazy
base_model:
  - mlabonne/AlphaMonarch-7B
  - FPHam/Karen_TheEditor_V2_STRICT_Mistral_7B
  - SanjiWatsuki/Kunoichi-DPO-v2-7B
  - OmnicromsBrain/NeuralStar-7b-Lazy
model-index:
  - name: NeuralStar_AlphaWriter_4x7b
    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: 70.22
            name: normalized accuracy
        source:
          url: >-
            https://huggingface.co/spaces/open-llm-leaderboard/open_llm_leaderboard?query=OmnicromsBrain/NeuralStar_AlphaWriter_4x7b
          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: 88.31
            name: normalized accuracy
        source:
          url: >-
            https://huggingface.co/spaces/open-llm-leaderboard/open_llm_leaderboard?query=OmnicromsBrain/NeuralStar_AlphaWriter_4x7b
          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: 64.6
            name: accuracy
        source:
          url: >-
            https://huggingface.co/spaces/open-llm-leaderboard/open_llm_leaderboard?query=OmnicromsBrain/NeuralStar_AlphaWriter_4x7b
          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: 71.7
        source:
          url: >-
            https://huggingface.co/spaces/open-llm-leaderboard/open_llm_leaderboard?query=OmnicromsBrain/NeuralStar_AlphaWriter_4x7b
          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
            name: accuracy
        source:
          url: >-
            https://huggingface.co/spaces/open-llm-leaderboard/open_llm_leaderboard?query=OmnicromsBrain/NeuralStar_AlphaWriter_4x7b
          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: 63
            name: accuracy
        source:
          url: >-
            https://huggingface.co/spaces/open-llm-leaderboard/open_llm_leaderboard?query=OmnicromsBrain/NeuralStar_AlphaWriter_4x7b
          name: Open LLM Leaderboard

NeuralStar_AlphaWriter_4x7b

I was blown away by the writing results I was getting from mlabonne/Beyonder-4x7B-v3 while writing in NovelCrafter.

Inspired by his LLM Course and fueled by his LazyMergeKit. I couldnt help but wonder what a writing model would be like if all 4 “experts” excelled in creative writing.

I present NeuralStar-AlphaWriter-4x7b:

NeuralStar_AlphaWriter_4x7b is a Mixture of Experts (MoE) made with the following models using LazyMergekit:

• mlabonne/AlphaMonarch-7B
• FPHam/Karen_TheEditor_V2_STRICT_Mistral_7B
• SanjiWatsuki/Kunoichi-DPO-v2-7B
• OmnicromsBrain/NeuralStar-7b-Lazy

⚡ Quantized Models

Special thanks to MRadermacher for the Static and iMatrx quantized models

.GGUF https://huggingface.co/mradermacher/NeuralStar_AlphaWriter_4x7b-GGUF

iMatrix https://huggingface.co/mradermacher/NeuralStar_AlphaWriter_4x7b-i1-GGUF

Q4_K_M and Q5_K_M .gguf Here created with mlabonne/Autogguf

🧩 Configuration

base_model: mlabonne/AlphaMonarch-7B
experts:  
  - source_model: mlabonne/AlphaMonarch-7B
    positive_prompts: 
    - "chat"
    - "assistant"
    - "tell me"
    - "explain"
    - "I want"
  - source_model: FPHam/Karen_TheEditor_V2_STRICT_Mistral_7B
    positive_prompts:
    - "edit"
    - "rewrite"
    - "evaluate"
    - "spelling"
    - "grammer"
  - source_model: SanjiWatsuki/Kunoichi-DPO-v2-7B
    positive_prompts:
    - "storywriting"
    - "write"
    - "scene"
    - "prose"
    - "character"
  - source_model: OmnicromsBrain/NeuralStar-7b-Lazy
    positive_prompts:
    - "codex"
    - "plot"
    - "outline"
    - "scenebeat"
    - "count"

💻 Usage

!pip install -qU transformers bitsandbytes accelerate

from transformers import AutoTokenizer
import transformers
import torch

model = "OmnicromsBrain/NeuralStar_AlphaWriter_4x7b"

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)	70.22
HellaSwag (10-Shot)	88.31
MMLU (5-Shot)	64.60
TruthfulQA (0-shot)	71.70
Winogrande (5-shot)	82.00
GSM8k (5-shot)	63.00