---
license: apache-2.0
tags:
- moe
- merge
- mergekit
- vicgalle/CarbonBeagle-11B
- Sao10K/Fimbulvetr-10.7B-v1
- bn22/Nous-Hermes-2-SOLAR-10.7B-MISALIGNED
- Yhyu13/LMCocktail-10.7B-v1
---

# Umbra-v2-MoE-4x10.7

Umbra-v2-MoE-4x10.7 is a Mixure of Experts (MoE) made with the following models using [LazyMergekit](https://colab.research.google.com/drive/1obulZ1ROXHjYLn6PPZJwRR6GzgQogxxb?usp=sharing):
* [vicgalle/CarbonBeagle-11B](https://huggingface.co./vicgalle/CarbonBeagle-11B)
* [Sao10K/Fimbulvetr-10.7B-v1](https://huggingface.co./Sao10K/Fimbulvetr-10.7B-v1)
* [bn22/Nous-Hermes-2-SOLAR-10.7B-MISALIGNED](https://huggingface.co./bn22/Nous-Hermes-2-SOLAR-10.7B-MISALIGNED)
* [Yhyu13/LMCocktail-10.7B-v1](https://huggingface.co./Yhyu13/LMCocktail-10.7B-v1)

## 🧩 Configuration

```yamlbase_model: vicgalle/CarbonBeagle-11B
gate_mode: hidden
dtype: bfloat16

experts:
  - source_model: vicgalle/CarbonBeagle-11B
    positive_prompts:
    - "versatile"
    - "adaptive"
    - "comprehensive"
    - "integrated"
    - "balanced"
    - "all-rounder"
    - "flexible"
    - "wide-ranging"
    - "multi-disciplinary"
    - "holistic"
    - "innovative"
    - "eclectic"
    - "resourceful"
    - "dynamic"
    - "robust"

    negative_prompts:
    - "narrow"
    - "specialized"
    - "limited"
    - "focused"

  - source_model: Sao10K/Fimbulvetr-10.7B-v1
    positive_prompts:
    - "creative"
    - "storytelling"
    - "expressive"
    - "imaginative"
    - "engaging"
    - "verbose"
    - "narrative"
    - "descriptive"
    - "elaborate"
    - "fictional"
    - "artistic"
    - "vivid"
    - "colorful"
    - "fantastical"
    - "lyrical"

    negative_prompts:
    - "sorry"
    - "I cannot"
    - "factual"
    - "concise"
    - "straightforward"
    - "objective"
    - "dry"

  - source_model: bn22/Nous-Hermes-2-SOLAR-10.7B-MISALIGNED
    positive_prompts:
    - "intelligent"
    - "analytical"
    - "accurate"
    - "knowledgeable"
    - "logical"
    - "data-driven"
    - "scientific"
    - "rational"
    - "precise"
    - "methodical"
    - "empirical"
    - "systematic"
    - "efficient"
    - "scholarly"
    - "statistical"
    - "calculate"
    - "compute"
    - "solve"
    - "work"
    - "python"
    - "javascript"
    - "programming"
    - "algorithm"
    - "tell me"
    - "assistant"

    negative_prompts:
    - "creative"
    - "imaginative"
    - "abstract"
    - "emotional"
    - "artistic"
    - "speculative"

  - source_model: Yhyu13/LMCocktail-10.7B-v1
    positive_prompts:
    - "instructive"
    - "verbose"
    - "descriptive"
    - "clear"
    - "detailed"
    - "informative"
    - "explanatory"
    - "elucidative"
    - "articulate"
    - "comprehensive"
    - "educational"
    - "thorough"
    - "specific"
    - "clarifying"
    - "structured"

    negative_prompts:
    - "concise"
    - "vague"```

## 💻 Usage

```python
!pip install -qU transformers bitsandbytes accelerate

from transformers import AutoTokenizer
import transformers
import torch

model = "Steelskull/Umbra-v2-MoE-4x10.7"

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"])
```