base_model: bunnycore/LLama-3.1-Hyper-Stock
experts_per_token: 2
gate_mode: hidden
dtype: bfloat16
experts:
  - source_model: SvalTek/L3-ColdBrew-Astrid
    positive_prompts:
      - "ColdBrew --"
      - "Write a story about a lonely robot exploring an abandoned space station."
      - "Describe the feeling of standing at the edge of a massive waterfall."
      - "Roleplay as a tavern keeper sharing tales with an adventurer."
      - "Imagine a bustling marketplace in a desert city and narrate the sights and sounds."
      - "[Genres: Science Fiction]
[Tags: humor, old school, sci fi]"
      - "[Mode: Interactive Storyteller]"
      - "[Mode: DM]"
    negative_prompts:
      - "Think step by step"

  - source_model: FPHam/L3-8B-Everything-COT
    positive_prompts:
      - "[OOC:"
      - "What are <thinking> and <reflection> blocks used for?"
      - "Reflect on why humans lie to each other."
      - "What are the most important aspects of good technical documentation?"
      - "Analyzeand explain the logic behind a recursive algorithm for sorting data."
      - "Think step by step with a logical reasoning and intellectual sense before you provide any response."
    negative_prompts:
      - "Alex is on a spaceship, standing before a mysterious control panel."

💻 Usage

python

!pip install -qU transformers bitsandbytes accelerate

from transformers import AutoTokenizer
import transformers
import torch

model = "SvalTek/ColdBrew-Oxford"

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"])