ComfyUI Node: VNCCS Dataset Generator

VNCCS Dataset Generator:

The DatasetGenerator node is designed to facilitate the creation of datasets specifically for Lora training, which is a technique used in machine learning to fine-tune models. This node is particularly useful for AI artists who are working with character data from games, as it allows for the organization and preparation of character-specific datasets. By leveraging this node, you can efficiently generate datasets that are structured and formatted to meet the requirements of Lora training, ensuring that your models can be trained effectively with the character data you provide. The node automates the process of organizing character information, creating necessary directories, and handling file processing, which significantly reduces the manual effort required in dataset preparation.

VNCCS Dataset Generator Input Parameters:

character

The character parameter specifies the name of the character for which the dataset is being generated. This parameter is crucial as it determines the specific character data that will be processed and included in the dataset. The character name is used to locate the corresponding character configuration and data files within the base path. There are no explicit minimum, maximum, or default values provided for this parameter, but it should match the name of a character directory within the base path.

game_name

The game_name parameter is used to associate the character with a specific game. This helps in organizing the dataset by appending the game name to the character's name, creating a unique identifier for the dataset. This parameter is important for distinguishing datasets when multiple games are involved. Similar to the character parameter, there are no explicit constraints on the values for game_name, but it should be a string that accurately represents the game associated with the character.

additional_caption

The additional_caption parameter allows you to add extra descriptive text to the dataset. This can be useful for providing additional context or information about the dataset, which might be beneficial during the training process. While this parameter is optional, including a meaningful caption can enhance the dataset's metadata. There are no specific constraints on this parameter, and it defaults to an empty string if not provided.

VNCCS Dataset Generator Output Parameters:

lora_dir

The lora_dir output parameter provides the path to the directory where the generated dataset is stored. This directory contains all the processed files and is organized according to the character and game name specified in the input parameters. The lora_dir is essential for accessing the dataset after it has been created, allowing you to use it for Lora training or other purposes. The output is a string representing the directory path, and it is crucial for locating the dataset files.

VNCCS Dataset Generator Usage Tips:

• Ensure that the character name provided matches the directory structure within your base path to avoid configuration loading issues.
• Use descriptive game names and additional captions to maintain clear and organized datasets, especially when working with multiple characters or games.

VNCCS Dataset Generator Common Errors and Solutions:

Failed to load character configuration <character>

• Explanation: This error occurs when the node is unable to find or load the configuration file for the specified character.
• Solution: Verify that the character name is correct and that the corresponding configuration file exists in the expected directory. Ensure that the file is accessible and properly formatted.

Dataset created in folder <lora_dir>, processed 0 files

• Explanation: This message indicates that the dataset directory was created, but no files were processed, possibly due to missing or incompatible data files.
• Solution: Check the character's data directory to ensure that the necessary files are present and compatible with the dataset generation process. Make sure the files are not corrupted or in an unsupported format.