---
license: cc-by-nc-4.0
tags:
- merge
- mergekit
- lazymergekit
base_model:
- liminerity/M7-7b
- rwitz/experiment26-truthy-iter-0
---

# Zebrafish-7B

Zebrafish-7B is my first model using the new merge method called [Model Stock](https://arxiv.org/abs/2403.19522).

Zebrafish-7B is a merge of the following models using [LazyMergekit](https://colab.research.google.com/drive/1obulZ1ROXHjYLn6PPZJwRR6GzgQogxxb?usp=sharing):
* [liminerity/M7-7b](https://huggingface.co/liminerity/M7-7b)
* [rwitz/experiment26-truthy-iter-0](https://huggingface.co/rwitz/experiment26-truthy-iter-0)

Special thanks to Charles Goddard for the quick implementation!

## 🏆 Evaluation

### Nous

| Model | Average | AGIEval | GPT4All | TruthfulQA | Bigbench |
|---|---:|---:|---:|---:|---:|
| [mlabonne/AlphaMonarch-7B](https://huggingface.co/mlabonne/AlphaMonarch-7B) [📄](https://gist.github.com/mlabonne/1d33c86824b3a11d2308e36db1ba41c1) | 62.74 | 45.37 | 77.01 | 78.39 | 50.2 |
| [**mlabonne/Zebrafish-7B**](https://huggingface.co/mlabonne/Zebrafish-7B) [📄](https://gist.github.com/mlabonne/719d5e106eefbcffb951b65616dcbec4) | **62.41** | **44.92** | **77.18** | **78.25** | **49.28** |
| [mlabonne/Beyonder-4x7B-v3](https://huggingface.co/mlabonne/Beyonder-4x7B-v3) [📄](https://gist.github.com/mlabonne/3740020807e559f7057c32e85ce42d92) | 61.91 | 45.85 | 76.67 | 74.98 | 50.12 |
| [mlabonne/NeuralBeagle14-7B](https://huggingface.co/mlabonne/NeuralBeagle14-7B) [📄](https://gist.github.com/mlabonne/ad0c665bbe581c8420136c3b52b3c15c) | 60.25 | 46.06 | 76.77 | 70.32 | 47.86 |
| [mistralai/Mistral-7B-Instruct-v0.2](https://huggingface.co/mistralai/Mistral-7B-Instruct-v0.2) [📄](https://gist.github.com/mlabonne/05d358e17dffdf9eee7c2322380c9da6) | 54.81 | 38.5 | 71.64 | 66.82 | 42.29 |

## 🧩 Configuration

```yaml
models:
  - model: mistralai/Mistral-7B-v0.1
  - model: liminerity/M7-7b
  - model: rwitz/experiment26-truthy-iter-0
merge_method: model_stock
base_model: mistralai/Mistral-7B-v0.1
dtype: bfloat16
```

## 💻 Usage

```python
!pip install -qU transformers accelerate

from transformers import AutoTokenizer
import transformers
import torch

model = "mlabonne/Zebrafish-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"])
```