ComfyUI Node: interrupt on NaN

interrupt on NaN:

The interrupt on NaN node is designed to ensure the stability and reliability of your AI model by monitoring the output for any NaN (Not a Number) or infinite values during processing. This node is particularly useful in preventing the propagation of invalid values that can disrupt the model's performance or lead to unexpected results. By either replacing these problematic values with zeros or interrupting the current processing, the node helps maintain the integrity of the model's output. This functionality is crucial for AI artists who rely on consistent and accurate results from their models, as it helps avoid potential issues that could arise from numerical instabilities.

interrupt on NaN Input Parameters:

model

This parameter represents the AI model that you are working with. It is the primary input to the node and is essential for the node to function correctly. The model is monitored for any NaN or infinite values during its processing.

replace_values

This is a boolean parameter that determines the action to be taken when NaN or infinite values are detected in the model's output. If set to True, the node will replace these values with zeros, ensuring that the processing continues smoothly without interruption. If set to False, the node will interrupt the current processing to prevent the propagation of invalid values. The default value for this parameter is True.

interrupt on NaN Output Parameters:

model

The output of this node is the same AI model that was input, but with the added functionality of monitoring and handling NaN or infinite values. Depending on the replace_values parameter, the model will either have these values replaced with zeros or the processing will be interrupted to maintain the integrity of the output.

interrupt on NaN Usage Tips:

• To ensure continuous processing without interruptions, set the replace_values parameter to True. This will replace any NaN or infinite values with zeros, allowing the model to continue running smoothly.
• If you prefer to halt processing when invalid values are detected, set the replace_values parameter to False. This can be useful for debugging and identifying the source of numerical instabilities in your model.

interrupt on NaN Common Errors and Solutions:

"Processing interrupted due to NaN or infinite values"

• Explanation: This error occurs when the node detects NaN or infinite values in the model's output and the replace_values parameter is set to False.
• Solution: To avoid this interruption, set the replace_values parameter to True, which will replace the invalid values with zeros and allow the processing to continue.

"Model output contains NaN or infinite values"

• Explanation: This warning indicates that the model's output has NaN or infinite values, but they have been replaced with zeros because the replace_values parameter is set to True.
• Solution: While this allows processing to continue, it is advisable to investigate the source of these invalid values in your model to prevent them from occurring in the first place.