README.md

---
tags:
- merge
- mergekit
- lazymergekit
- aaditya/Llama3-OpenBioLLM-8B
base_model:
- aaditya/Llama3-OpenBioLLM-8B
license: llama3
language:
- en
---

# Llama-3-Galen-8B-32k-v1

<img src="https://cdn-uploads.huggingface.co/production/uploads/60c8619d95d852a24572b025/R73wGdZE3GWeF9QZPvruG.jpeg" width="600" />

Llama-3-Galen-8B-32k-v1 is a RoPE scaled, DARE TIES merge of the following models using [LazyMergekit](https://colab.research.google.com/drive/1obulZ1ROXHjYLn6PPZJwRR6GzgQogxxb?usp=sharing):
* [aaditya/Llama3-OpenBioLLM-8B](https://huggingface.co/aaditya/Llama3-OpenBioLLM-8B)
* [johnsnowlabs/JSL-MedLlama-3-8B-v2.0](https://huggingface.co/johnsnowlabs/JSL-MedLlama-3-8B-v2.0)

> **This model is capable of handling a context size of 32K right out of the box, enabled with Dynamic RoPE scaling.**

## 🧩 Configuration

```yaml
models:
  - model: johnsnowlabs/JSL-MedLlama-3-8B-v2.0
    # No parameters necessary for base model
  - model: aaditya/Llama3-OpenBioLLM-8B
    parameters:
      density: 0.53
      weight: 0.5
merge_method: dare_ties
base_model: johnsnowlabs/JSL-MedLlama-3-8B-v2.0
parameters:
  int8_mask: true
dtype: bfloat16
```

## 💻 Usage

```python
!pip install -qU transformers accelerate

from transformers import AutoTokenizer
import transformers
import torch

model = "abhinand/Llama-3-Galen-8B-32k-v1"
messages = [{"role": "user", "content": "How long does it take to recover from COVID-19?"}]

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"])
```