sixtyoneeighty/FNCARL9000

sixtyoneeighty-4x7b-v2

sixtyoneeighty-4x7b-v2 is a Mixure of Experts (MoE) made with the following models using LazyMergekit:

• Open-Orca/Mistral-7B-OpenOrca
• NeuralNovel/Mistral-7B-Instruct-v0.2-Neural-Story
• S-miguel/The-Trinity-Coder-7B
• chihoonlee10/T3Q-Mistral-Orca-Math-DPO

🧩 Configuration

base_model: cognitivecomputations/dolphin-2.8-mistral-7b-v02
gate_mode: hidden
dtype: bfloat16
experts_per_token: 2
experts:
  - source_model: Open-Orca/Mistral-7B-OpenOrca
    positive_prompts:
      - "What are some fun activities to do in Seattle?"
      - "What are some fun historical facts about New York City?"
    negative_prompts:
      - "Write a Python script to scrape data from a website."
      - "Explain the key differences between Bayesian and frequentist statistics."

  - source_model: NeuralNovel/Mistral-7B-Instruct-v0.2-Neural-Story
    positive_prompts:
      - "Write me a fictional story about dragons and wizards?"
      - "From now on take on the role of Dwayne Johnson"
    negative_prompts:
      - "When is the next solar eclipse."
      - "What year did World War II end?"

  - source_model: S-miguel/The-Trinity-Coder-7B
    positive_prompts:
      - "Can you review my JavaScript code and suggest ways to optimize it for better performance?"
      - "I'm getting an 'undefined variable' error in my Python script. Here's the code: [code snippet]"
    negative_prompts:
      - "What are some effective strategies for managing stress and anxiety?"
      - "Compare and contrast the themes in 'The Great Gatsby' and 'The Catcher in the Rye'."

  - source_model: chihoonlee10/T3Q-Mistral-Orca-Math-DPO
    positive_prompts:
      - "What's a square root of 1337?"
      - "Find the midpoint of the line segment with the given end points (-5,7) and (-2,1)"
    negative_prompts:
      - "What are some effective strategies for managing stress and anxiety?"
      - "Compare and contrast the themes in 'The Great Gatsby' and 'The Catcher in the Rye'."

💻 Usage

!pip install -qU transformers bitsandbytes accelerate

from transformers import AutoTokenizer
import transformers
import torch

model = "jambroz/sixtyoneeighty-7b-MOE"

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"])