mayacinka / West-Ramen-7Bx4

🧩 Configuration

base_model: /home/Ubuntu/Desktop/mergekit/models/Mistral-7B-Instruct-v0.2
gate_mode: hidden 
dtype: bfloat16 
experts:
  - source_model: /home/Ubuntu/Desktop/mergekit/models/Mistral-7B-Instruct-v0.2
    positive_prompts:
      - "instructions"
      - "concise"
      - "straightforward"
      - "helpful"
      - "assistant"
    negative_prompts:
      - "vague"
      - "inaccurate"
      - "verbose"
      - "complicated"
      - "speculative"
  - source_model: /home/Ubuntu/Desktop/mergekit/models/NeuralOmniWestBeaglake-7B
    positive_prompts:
      - "storytelling"
      - "role play"
      - "imagine"
      - "artistic"
      - "narrative"
  - source_model: /home/Ubuntu/Desktop/mergekit/models/Kunoichi-DPO-v2-7B
    positive_prompts:
      - "reason"
      - "think step by step"
      - "logic"
      - "knowledge"
    negative_prompts:
      - "artistic"
      - "speculative"
      - "playful"
  - source_model: /home/Ubuntu/Desktop/mergekit/models/Starling-LM-7B-alpha
    positive_prompts:
      - "code"
      - "python"
      - "javascript"
      - "react"
      - "clear"
      - "programming"
    negative_prompts:
      - "error"
      - "art"
      - "role play"

💻 Usage

!pip install -qU transformers bitsandbytes accelerate

from transformers import AutoTokenizer
import transformers
import torch

model = "mayacinka/West-Ramen-7Bx4"

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