README.md

---
language:
- en
license: apache-2.0
library_name: transformers
tags:
- moe
- merge
- abideen/NexoNimbus-7B
- mlabonne/NeuralMarcoro14-7B
model-index:
- name: NexoNimbus-MoE-2x7B
  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: 66.81
      name: normalized accuracy
    source:
      url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=abideen/NexoNimbus-MoE-2x7B
      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: 85.66
      name: normalized accuracy
    source:
      url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=abideen/NexoNimbus-MoE-2x7B
      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: 64.51
      name: accuracy
    source:
      url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=abideen/NexoNimbus-MoE-2x7B
      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: 53.06
    source:
      url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=abideen/NexoNimbus-MoE-2x7B
      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: 81.53
      name: accuracy
    source:
      url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=abideen/NexoNimbus-MoE-2x7B
      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: 53.53
      name: accuracy
    source:
      url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=abideen/NexoNimbus-MoE-2x7B
      name: Open LLM Leaderboard
---

# NexoNimbus-MoE-2x7B

![image/png](https://cdn-uploads.huggingface.co/production/uploads/64e380b2e12618b261fa6ba0/_bzC6xkVIHW0tSigBxUI3.png)

NexoNimbus-MoE-2x7B is a Mixure of Experts (MoE) made with the following models:
* [abideen/NexoNimbus-7B](https://huggingface.co/abideen/NexoNimbus-7B)
* [mlabonne/NeuralMarcoro14-7B](https://huggingface.co/mlabonne/NeuralMarcoro14-7B)

🏆 Evaluation
NexoNimbus-MoE-2x7B is the 10th best-performing 13B LLM on the Open LLM Leaderboard:


![image/png](https://cdn-uploads.huggingface.co/production/uploads/64e380b2e12618b261fa6ba0/z8E728H5fJqVtKNeGuwjX.png)


|    Task     |Version| Metric |Value|   |Stderr|
|-------------|------:|--------|----:|---|-----:|
|arc_challenge|      0|acc     |62.28|±  |  1.41|
|             |       |acc_norm|66.80|±  |  1.37|
|hellaswag    |      0|acc     |66.83|±  |  0.46|
|             |       |acc_norm|85.66|±  |  0.34|
|gsm8k        |      0|acc     |53.52|±  |  1.37|
|winogrande   |      0|acc     |81.53|±  |  1.09|
|mmlu         |      0|acc     |64.51|±  |  1.00|

Average: 67.51%

### TruthfulQA
|    Task     |Version|Metric|Value|   |Stderr|
|-------------|------:|------|----:|---|-----:|
|truthfulqa_mc|      1|mc1   |35.98|±  |  1.68|
|             |       |mc2   |53.05|±  |  1.53|


## 🧩 Configuration

```yaml
base_model: teknium/OpenHermes-2.5-Mistral-7B
gate_mode: hidden
dtype: bfloat16
experts:
  - source_model: abideen/NexoNimbus-7B
    positive_prompts:
    - "Mathematics"
    - "Physics"
    - "Chemistry"
    - "Biology"
    - "Medicine"
    - "Engineering"
    - "Computer Science"

    negative_prompts:
    - "History"
    - "Philosophy"
    - "Linguistics"
    - "Literature"
    - "Art and Art History"
    - "Music Theory and Composition"
    - "Performing Arts (Theater, Dance)"

  - source_model: mlabonne/NeuralMarcoro14-7B 
    positive_prompts:
    - "Earth Sciences (Geology, Meteorology, Oceanography)"
    - "Environmental Science"
    - "Astronomy and Space Science"
    - "Psychology"
    - "Sociology"
    - "Anthropology"
    - "Political Science"
    - "Economics"
    negative_prompts:
    - "Education"
    - "Law"
    - "Theology and Religious Studies"
    - "Communication Studies"
    - "Business and Management"
    - "Agricultural Sciences"
    - "Nutrition and Food Science"
    - "Sports Science"
```

## 💻 Usage

Here's a [Colab notebook](https://colab.research.google.com/drive/1B1Q7vO95cDkEJbKIPhOWr6exB9-Q_lr-?usp=sharing) to run NexoNimbus-MoE-2x7B in 4-bit precision on a free T4 GPU.

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

from transformers import AutoTokenizer
import transformers
import torch

model = "abideen/NexoNimbus-MoE-2x7B"

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 is data science."}]
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"])
```

"Data science is an interdisciplinary field that combines mathematics, statistics, computer science, and domain expertise in order to extract meaningful insights and knowledge from structured and unstructured data. It involves the process of collecting, cleaning, transforming, analyzing, and visualizing data in order to identify patterns, trends, and relationships that can inform decision-making and drive business strategies. Data scientists use various tools and techniques, such as machine learning, deep learning, and natural language processing, to develop predictive models, optimize processes, and automate decision-making. The field of data science is rapidly evolving as more and more data is generated and the demand for data-driven insights continues to grow."
# [Open LLM Leaderboard Evaluation Results](https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard)
Detailed results can be found [here](https://huggingface.co/datasets/open-llm-leaderboard/details_abideen__NexoNimbus-MoE-2x7B)

|             Metric              |Value|
|---------------------------------|----:|
|Avg.                             |67.51|
|AI2 Reasoning Challenge (25-Shot)|66.81|
|HellaSwag (10-Shot)              |85.66|
|MMLU (5-Shot)                    |64.51|
|TruthfulQA (0-shot)              |53.06|
|Winogrande (5-shot)              |81.53|
|GSM8k (5-shot)                   |53.53|