库名称：transformers 许可证：mit 任务标签：视频文本转文本

LLaVA-Video-7B-Qwen2-TPO

LLaVA-Video-7B-Qwen2-TPO，由论文《Temporal Preference Optimization for Long-form Video Understanding》提出，基于LLaVA-Video-7B-Qwen2进行了时间偏好优化。该模型在一系列基准测试中确立了最先进的性能，相比LLaVA-Video-7B平均性能提升了1.5%。特别值得一提的是，它在Video-MME基准测试中成为领先的7B参数模型。

项目页面：https://ruili33.github.io/tpo_website/ 代码：https://github.com/ruili33/TPO

评估结果

快速开始

使用以下代码快速上手该模型。更多信息请参考我们的GitHub代码库。

# pip install git+https://github.com/LLaVA-VL/LLaVA-NeXT.git
from llava.model.builder import load_pretrained_model
from llava.mm_utils import get_model_name_from_path, process_images, tokenizer_image_token
from llava.constants import IMAGE_TOKEN_INDEX, DEFAULT_IMAGE_TOKEN, DEFAULT_IM_START_TOKEN, DEFAULT_IM_END_TOKEN, IGNORE_INDEX
from llava.conversation import conv_templates, SeparatorStyle
from PIL import Image
import requests
import copy
import torch
import sys
import warnings
from decord import VideoReader, cpu
import numpy as np
warnings.filterwarnings("ignore")
def load_video(self, video_path, max_frames_num,fps=1,force_sample=False):
    if max_frames_num == 0:
        return np.zeros((1, 336, 336, 3))
    vr = VideoReader(video_path, ctx=cpu(0),num_threads=1)
    total_frame_num = len(vr)
    video_time = total_frame_num / vr.get_avg_fps()
    fps = round(vr.get_avg_fps()/fps)
    frame_idx = [i for i in range(0, len(vr), fps)]
    frame_time = [i/fps for i in frame_idx]
    if len(frame_idx) > max_frames_num or force_sample:
        sample_fps = max_frames_num
        uniform_sampled_frames = np.linspace(0, total_frame_num - 1, sample_fps, dtype=int)
        frame_idx = uniform_sampled_frames.tolist()
        frame_time = [i/vr.get_avg_fps() for i in frame_idx]
    frame_time = ",".join([f"{i:.2f}s" for i in frame_time])
    spare_frames = vr.get_batch(frame_idx).asnumpy()
    # import pdb;pdb.set_trace()
    return spare_frames,frame_time,video_time
pretrained = "ruili0/LLaVA-Video-7B-TPO"
model_name = "llava_qwen"
device = "cuda"
device_map = "auto"
tokenizer, model, image_processor, max_length = load_pretrained_model(pretrained, None, model_name, torch_dtype="bfloat16", device_map=device_map)  # 可添加其他llava_model_args参数
model.eval()
video_path = "local_demo/assets/dc_demo.mp4"
max_frames_num = "64"
video,frame_time,video_time = load_video(video_path, max_frames_num, 1, force_sample=True)
video = image_processor.preprocess(video, return_tensors="pt")["pixel_values"].cuda().bfloat16()
video = [video]
conv_template = "qwen_1_5"  # 确保为不同模型使用正确的对话模板
time_instruciton = f"视频时长为{video_time:.2f}秒，从中均匀采样了{len(video[0])}帧。这些帧位于{frame_time}时间点。请回答以下与该视频相关的问题。"
question = DEFAULT_IMAGE_TOKEN + f"{time_instruciton}\n请详细描述这个视频内容。"
conv = copy.deepcopy(conv_templates[conv_template])
conv.append_message(conv.roles[0], question)
conv.append_message(conv.roles[1], None)
prompt_question = conv.get_prompt()
input_ids = tokenizer_image_token(prompt_question, tokenizer, IMAGE_TOKEN_INDEX, return_tensors="pt").unsqueeze(0).to(device)
cont = model.generate(
    input_ids,
    images=video,
    modalities= ["video"],
    do_sample=False,
    temperature=0,
    max_new_tokens=4096,
)
text_outputs = tokenizer.batch_decode(cont, skip_special_tokens=True)[0].strip()
print(text_outputs)

许可证

本项目使用的部分数据集和检查点受其原始许可证约束。用户必须遵守这些原始许可证的所有条款和条件，包括但不限于数据集使用的OpenAI使用条款和基础语言模型(Qwen2许可证)的特定许可。除原始许可证规定外，本项目不施加任何额外限制。此外，提醒用户确保其使用行为符合所有适用法律法规。

引用

Bibtex格式:

@article{li2025temporal,
      title={Temporal Preference Optimization for Long-Form Video Understanding},
      author={Li, Rui and Wang, Xiaohan and Zhang, Yuhui and Wang, Zeyu and Yeung-Levy, Serena},
      journal={arXiv preprint arXiv:2501.13919},
      year={2025}
    }

参考文献:

[1]. Liu, Z., Zhu, L., Shi, B., Zhang, Z., Lou, Y., Yang, S., ... & Lu, Y. (2024). NVILA: Efficient Frontier Visual Language Models. arXiv preprint arXiv:2412.04468.

[2]. Zhang, Y., Wu, J., Li, W., Li, B., Ma, Z., Liu, Z., & Li, C. (2024). Video instruction tuning with synthetic data. arXiv preprint arXiv:2410.02713.