README.md

---
base_model:
- Dampfinchen/Llama-3.1-8B-Ultra-Instruct
tags:
- merge
- mergekit
- Undi95/Meta-Llama-3.1-8B-Claude
- Dampfinchen/Llama-3.1-8B-Ultra-Instruct
license: llama3.1
language:
- en
- de
---

# llama3.1-8b-spaetzle-v59

llama3.1-8b-spaetzle-v59 is a dare ties merge of the models 
* [Undi95/Meta-Llama-3.1-8B-Claude](https://huggingface.co/Undi95/Meta-Llama-3.1-8B-Claude)
* [Dampfinchen/Llama-3.1-8B-Ultra-Instruct](https://huggingface.co/Dampfinchen/Llama-3.1-8B-Ultra-Instruct)

The GGUF is simply built with b3472 llama.cpp.

EQ-Bench v2_de: 67.38 (171/171) (which is not bad...)

## 🧩 Configuration

```yaml
models:
  - model: Dampfinchen/Llama-3.1-8B-Ultra-Instruct
    # no parameters necessary for base model
  - model: Undi95/Meta-Llama-3.1-8B-Claude
    parameters:
      density: 0.65
      weight: 0.4
merge_method: dare_ties
base_model: Dampfinchen/Llama-3.1-8B-Ultra-Instruct
parameters:
  int8_mask: true
dtype: bfloat16
random_seed: 0
tokenizer_source: base
```

## 💻 Usage

```python
!pip install -qU transformers accelerate

from transformers import AutoTokenizer
import transformers
import torch

model = "cstr/llama3.1-8b-spaetzle-v59"
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"])
```