如何将训练数据集帧转换为5d张量，同时保持帧维度的标签？

import numpy as np

def convert_to_5d_tensor(frames, labels, batch_size, time_steps):
    """
    Convert a list of frames and their corresponding labels into a 5D tensor.
    
    Args:
    frames (list of np.array): List of video frames.
    labels (list of np.array): List of corresponding labels for each frame.
    batch_size (int): Number of videos in a batch.
    time_steps (int): Number of frames per video.
    
    Returns:
    tuple: A tuple containing the 5D tensor of frames and the 2D tensor of labels.
    """
    num_videos = len(frames)
    height, width, channels = frames[0].shape
    
    # Initialize the 5D tensor for frames
    frames_tensor = np.zeros((num_videos, time_steps, height, width, channels))
    labels_tensor = np.zeros((num_videos, time_steps))
    
    for i, (frame_seq, label_seq) in enumerate(zip(frames, labels)):
        if len(frame_seq) != time_steps or len(label_seq) != time_steps:
            raise ValueError("Each video must have exactly {} frames and labels.".format(time_steps))
        
        frames_tensor[i] = frame_seq[:time_steps]
        labels_tensor[i] = label_seq[:time_steps]
    
    return frames_tensor, labels_tensor

# Example usage
frames = [np.random.rand(10, 64, 64, 3) for _ in range(5)]  # 5 videos, each with 10 frames
labels = [np.random.randint(0, 10, 10) for _ in range(5)]  # Corresponding labels

batch_size = 5
time_steps = 10

frames_tensor, labels_tensor = convert_to_5d_tensor(frames, labels, batch_size, time_steps)
print("Frames Tensor Shape:", frames_tensor.shape)
print("Labels Tensor Shape:", labels_tensor.shape)