Llava Llama 3 8b V1 1 Transformers

🚀 多模态大模型 llava-llama-3-8b-v1_1-hf

llava-llama-3-8b-v1_1-hf 是一款图像文本多模态大模型，基于 XTuner 工具包，使用 ShareGPT4V-PT 和 InternVL-SFT 数据集，对 meta-llama/Meta-Llama-3-8B-Instruct 和 CLIP-ViT-Large-patch14-336 进行微调得到。

🚀 快速开始

通过 pipeline 进行对话

from transformers import pipeline
from PIL import Image    
import requests

model_id = "xtuner/llava-llama-3-8b-v1_1-transformers"
pipe = pipeline("image-to-text", model=model_id, device=0)
url = "http://images.cocodataset.org/val2017/000000039769.jpg"

image = Image.open(requests.get(url, stream=True).raw)
prompt = ("<|start_header_id|>user<|end_header_id|>\n\n<image>\nWhat are these?<|eot_id|>"
          "<|start_header_id|>assistant<|end_header_id|>\n\n")
outputs = pipe(image, prompt=prompt, generate_kwargs={"max_new_tokens": 200})
print(outputs)
>>> [{'generated_text': 'user\n\n\nWhat are these?assistant\n\nThese are two cats, one brown and one gray, lying on a pink blanket. sleep. brown and gray cat sleeping on a pink blanket.'}]

通过纯 transformers 进行对话

import requests
from PIL import Image

import torch
from transformers import AutoProcessor, LlavaForConditionalGeneration

model_id = "xtuner/llava-llama-3-8b-v1_1-transformers"

prompt = ("<|start_header_id|>user<|end_header_id|>\n\n<image>\nWhat are these?<|eot_id|>"
          "<|start_header_id|>assistant<|end_header_id|>\n\n")
image_file = "http://images.cocodataset.org/val2017/000000039769.jpg"

model = LlavaForConditionalGeneration.from_pretrained(
    model_id, 
    torch_dtype=torch.float16, 
    low_cpu_mem_usage=True, 
).to(0)

processor = AutoProcessor.from_pretrained(model_id)

raw_image = Image.open(requests.get(image_file, stream=True).raw)
inputs = processor(prompt, raw_image, return_tensors='pt').to(0, torch.float16)

output = model.generate(**inputs, max_new_tokens=200, do_sample=False)
print(processor.decode(output[0][2:], skip_special_tokens=True))
>>> These are two cats, one brown and one gray, lying on a pink blanket. sleep. brown and gray cat sleeping on a pink blanket.

复现实验

请参考 文档。

✨ 主要特性

• 多模态融合：结合图像和文本信息，实现更丰富的交互。
• 多种格式支持：支持 HuggingFace LLaVA 格式、XTuner LLaVA 格式和 GGUF 格式。

📚 详细文档

模型信息

llava-llama-3-8b-v1_1-hf 是由 XTuner 基于 meta-llama/Meta-Llama-3-8B-Instruct 和 CLIP-ViT-Large-patch14-336，使用 ShareGPT4V-PT 和 InternVL-SFT 数据集微调得到的 LLaVA 模型。

注意：此模型为 HuggingFace LLaVA 格式。

资源链接

• GitHub: xtuner
• 官方 LLaVA 格式模型: xtuner/llava-llama-3-8b-v1_1-hf
• XTuner LLaVA 格式模型: xtuner/llava-llama-3-8b-v1_1
• GGUF 格式模型: xtuner/llava-llama-3-8b-v1_1-gguf

模型细节

模型效果

📄 许可证

@misc{2023xtuner,
    title={XTuner: A Toolkit for Efficiently Fine-tuning LLM},
    author={XTuner Contributors},
    howpublished = {\url{https://github.com/InternLM/xtuner}},
    year={2023}
}