8-3 Meta AI: Llama3 モデル

import transformers
import torch
from huggingface_hub import login

# Log in to Hugging Face
login(token="hf_XXXXXXXXXXXXXXXX") # Replace with your own Hugging Face token

# Model name
model_id = "meta-llama/Meta-Llama-3-8B-Instruct"

# Create a pipeline for text generation
pipeline = transformers.pipeline(
    "text-generation",  # Specify the task type as text generation
    model=model_id,  # Specify the model ID
    model_kwargs={
        "torch_dtype": torch.bfloat16, # Set the data type of the model (bfloat16)
        "cache_dir": "./model" # Set the path to store the model (default is ~/.cache/huggingface)
    },
    device_map="auto",  # Automatically select device (CPU or GPU)
)

# Set up the conversation messages, with roles and content
messages = [
    {"role": "user", "content": "What is a cat? Describe in one sentence."},  # User asks a question
]

# Define end-of-sequence tokens to identify where generation should stop
terminators = [
    pipeline.tokenizer.eos_token_id,
    pipeline.tokenizer.convert_tokens_to_ids("<|eot_id|>")
]

# Use the pipeline to generate text
outputs = pipeline(
    messages,  # Input messages
    max_new_tokens=256,  # Maximum number of new tokens to generate
    eos_token_id=terminators,  # List of end-of-sequence token IDs
    do_sample=True,  # Enable random sampling for generation
    temperature=0.6,  # Controls randomness of generation
    top_p=0.9,  # Only consider tokens with cumulative probability up to 0.9
)

# Print the result
print(f'result: {outputs[0]["generated_text"][-1]["content"]}')