---
tags:
- merge
- mergekit
- cognitivecomputations/dolphin-2.8-mistral-7b-v02
- NousResearch/Hermes-2-Pro-Mistral-7B
base_model:
- cognitivecomputations/dolphin-2.8-mistral-7b-v02
license: apache-2.0
---

![](https://raw.githubusercontent.com/saucam/models/main/Nereus.png)

# 🌊 Nereus-7B

Nereus-7B excels at conversations, coding, and tasks that require structured output in JSON.
It is a merge of the following models using [mergekit](https://github.com/arcee-ai/mergekit):
* [cognitivecomputations/dolphin-2.8-mistral-7b-v02](https://huggingface.co/cognitivecomputations/dolphin-2.8-mistral-7b-v02)
* [NousResearch/Hermes-2-Pro-Mistral-7B](https://huggingface.co/NousResearch/Hermes-2-Pro-Mistral-7B)

## 🧩 Configuration

```yamlmodels:
  - model: cognitivecomputations/dolphin-2.8-mistral-7b-v02
    parameters:
      density: 0.5
      weight: 0.4
    # No parameters necessary for base model
  - model: NousResearch/Hermes-2-Pro-Mistral-7B
    parameters:
      density: 0.5
      weight: 0.6
merge_method: dare_ties
base_model: cognitivecomputations/dolphin-2.8-mistral-7b-v02
parameters:
  int8_mask: true
dtype: bfloat16
```

## Eval Results

|Benchmark|                       Model                        |agieval|gpt4all|bigbench|truthfulqa|Average|
|---------|----------------------------------------------------|------:|------:|-------:|---------:|------:|
|nous     |[Nereus-7B](https://huggingface.co/saucam/Nereus-7B)|   42.8|  72.21|   39.17|     54.32|  52.12|


 |Benchmark|                       Model                        |winogrande| arc |gsm8k|mmlu|truthfulqa|hellaswag|Average|
|---------|----------------------------------------------------|---------:|----:|----:|---:|---------:|--------:|------:|
|openllm  |[Nereus-7B](https://huggingface.co/saucam/Nereus-7B)|     76.95|62.54|46.25|59.6|     54.32|    83.23|  63.82|


For detailed results [see here](https://github.com/saucam/model_evals/blob/main/saucam/Nereus-7B/README.md)

## 💻 Usage

```python
!pip install -qU transformers accelerate

from transformers import AutoTokenizer
import transformers
import torch

model = "saucam/Nereus-7B"
messages = [{"role": "user", "content": "What is a large language model?"}]

tokenizer = AutoTokenizer.from_pretrained(model)
prompt = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
pipeline = transformers.pipeline(
    "text-generation",
    model=model,
    torch_dtype=torch.float16,
    device_map="auto",
)

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"])
```

Sample responses
```
What is a large language model?<|im_end|>
<|im_start|>assistant
A large language model is a type of artificial intelligence (AI) system that has been trained on vast amounts of text data to understand, generate, and predict patterns in human language. It is designed to process and analyze natural language input, making it capable of tasks such as text generation, translation, language translation, and text classification. These models are typically based on deep learning techniques, particularly neural networks, and are trained on large datasets, often consisting of billions of words. Some well-known large language models include GPT-3 by OpenAI, BERT by Google, and T5 by Google. These models can be fine-tuned for specific tasks or domains to improve their performance. They have revolutionized the field of natural language processing and have numerous applications in areas such as chatbots, search engines, and automated writing assistance.
```