Llama3 8B Slerp Med Chinese

🚀 llama3-8B-slerp-med-chinese

llama3-8B-slerp-med-chinese is a merged model that combines the strengths of multiple models, offering enhanced performance in relevant tasks.

🚀 Quick Start

llama3-8B-slerp-med-chinese is a merge of the following models using LazyMergekit:

• winninghealth/WiNGPT2-Llama-3-8B-Base
• johnsnowlabs/JSL-MedLlama-3-8B-v1.0

✨ Features

• Multi - model Merge: Combines winninghealth/WiNGPT2-Llama-3-8B-Base and johnsnowlabs/JSL-MedLlama-3-8B-v1.0 using the slerp merge method.
• Language Support: Supports multiple languages including Chinese, English, and French.

📦 Installation

The installation steps are included in the usage example. You can install the necessary libraries with the following command:

!pip install -qU transformers accelerate

💻 Usage Examples

Basic Usage

!pip install -qU transformers accelerate

from transformers import AutoTokenizer
import transformers
import torch

model = "shanchen/llama3-8B-slerp-med-chinese"
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"])

📚 Documentation

🧩 Configuration

slices:
  - sources:
      - model: winninghealth/WiNGPT2-Llama-3-8B-Base
        layer_range: [0,32]
      - model: johnsnowlabs/JSL-MedLlama-3-8B-v1.0
        layer_range: [0,32]
merge_method: slerp
base_model: winninghealth/WiNGPT2-Llama-3-8B-Base
parameters:
  t:
    - filter: self_attn
      value: [0, 0.5, 0.5, 0.5, 1]
    - filter: mlp
      value: [1, 0.5, 0.5, 0.5, 0]
    - value: 0.5
dtype: bfloat16

📄 License

The model is licensed under the llama3 license.