Perform crop on the bounding boxes given the offsets. Args: boxes (ndarray or None): bounding boxes to perform crop. The dimension

Perform crop on the bounding boxes given the offsets. Args: boxes (ndarray or None): bounding boxes to perform crop. The dimension

iterates through data and handles wrap around

iterates through data and handles wrap around

Disable the redundant torch default initialization to accelerate model creation.

Divides an image into patches of a specified size. Args: image (PIL.Image.Image): The input image. patch_size (int): The size

To generate videos based on the embeddings generated_ids: the index of the generated tokens embedding: the embeddings for sy

To generate videos based on the embeddings generated_ids: the index of the generated tokens embedding: the embeddings for sy

To generate the images based on the embeddings generated_ids: the index of the generated tokens embedding: the embeddings fo

To generate the images based on the embeddings generated_ids: the index of the generated tokens embedding: the embeddings fo

This function is used to generate the tokens and output embeddings that employed to generate images/videos/audios inputs: dict

To generate videos based on the embeddings generated_ids: the index of the generated tokens embedding: the embeddings for sy

To generate videos based on the embeddings generated_ids: the index of the generated tokens embedding: the embeddings for sy

Sinusoid position encoding table

Sinusoid position encoding table

sample and process the alpaca dataset in to the following format: [ { "image_name": "00000000000", "output_mo

Args: args: dataset_name_list: List[str] repeats: List[int], the training epochs for each dataset

sample and process the llava instruction dataset into the following format: [ { "image_name": "00000000000.jpg",