ryan0712/llama-3-8b-slow-DUS-max-layer1-method2

llama-3-8b-slow-DUS-layer1-method2

llama-3-8b-slow-DUS-layer1-method2 is a merge of the following models using LazyMergekit:

• NousResearch/Meta-Llama-3-8B
• NousResearch/Meta-Llama-3-8B
• NousResearch/Meta-Llama-3-8B
• NousResearch/Meta-Llama-3-8B
• NousResearch/Meta-Llama-3-8B
• NousResearch/Meta-Llama-3-8B
• NousResearch/Meta-Llama-3-8B
• NousResearch/Meta-Llama-3-8B
• NousResearch/Meta-Llama-3-8B
• NousResearch/Meta-Llama-3-8B
• NousResearch/Meta-Llama-3-8B
• NousResearch/Meta-Llama-3-8B
• NousResearch/Meta-Llama-3-8B
• NousResearch/Meta-Llama-3-8B
• NousResearch/Meta-Llama-3-8B
• NousResearch/Meta-Llama-3-8B

🧩 Configuration

slices:

      - sources:
          - model: NousResearch/Meta-Llama-3-8B
            layer_range: [0, 1]
    
      - sources:
          - model: NousResearch/Meta-Llama-3-8B
            layer_range: [1, 2]
    
      - sources:
          - model: NousResearch/Meta-Llama-3-8B
            layer_range: [2, 3]
    
      - sources:
          - model: NousResearch/Meta-Llama-3-8B
            layer_range: [3, 4]
    
      - sources:
          - model: NousResearch/Meta-Llama-3-8B
            layer_range: [4, 5]
    
      - sources:
          - model: NousResearch/Meta-Llama-3-8B
            layer_range: [5, 6]
    
      - sources:
          - model: NousResearch/Meta-Llama-3-8B
            layer_range: [6, 7]
    
      - sources:
          - model: NousResearch/Meta-Llama-3-8B
            layer_range: [7, 8]
    
      - sources:
          - model: NousResearch/Meta-Llama-3-8B
            layer_range: [8, 9]
    
      - sources:
          - model: NousResearch/Meta-Llama-3-8B
            layer_range: [9, 10]
    
      - sources:
          - model: NousResearch/Meta-Llama-3-8B
            layer_range: [10, 11]
    
      - sources:
          - model: NousResearch/Meta-Llama-3-8B
            layer_range: [11, 12]
    
      - sources:
          - model: NousResearch/Meta-Llama-3-8B
            layer_range: [12, 13]
    
      - sources:
          - model: NousResearch/Meta-Llama-3-8B
            layer_range: [13, 14]
    
      - sources:
          - model: NousResearch/Meta-Llama-3-8B
            layer_range: [14, 15]
    
      - sources:
          - model: NousResearch/Meta-Llama-3-8B
            layer_range: [30, 31]
    
merge_method: passthrough
dtype: bfloat16

💻 Usage

!pip install -qU transformers accelerate

from transformers import AutoTokenizer
import transformers
import torch

model = "ryan0712/llama-3-8b-slow-DUS-layer1-method2"
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"])