ComfyUI Node: NNT Load Model

NNT Load Model:

The NntLoadModel node is designed to facilitate the loading of various types of neural network models, making it an essential tool for AI artists who work with different model formats. This node supports loading models in several formats, including State Dict, TorchScript, PyTorch Model, ONNX, TorchScript Mobile, and Quantized models. By providing a unified interface for loading these diverse model types, NntLoadModel simplifies the process of integrating pre-trained models into your workflow. It also offers the option to load associated optimizers, which can be particularly beneficial for those looking to continue training or fine-tuning models. The node's ability to handle multiple formats ensures flexibility and adaptability, allowing you to work with a wide range of models without needing to worry about the underlying technical details.

NNT Load Model Input Parameters:

filename

The filename parameter specifies the name of the file containing the model you wish to load. This parameter is crucial as it directs the node to the correct file within the specified directory. There are no specific minimum or maximum values for this parameter, but it must be a valid string representing a file name.

directory

The directory parameter indicates the location where the model file is stored. This is important for the node to locate and access the model file. Like filename, this parameter should be a valid directory path string.

load_format

The load_format parameter determines the format in which the model is saved. It supports options such as "State Dict," "TorchScript," "PyTorch Model," "ONNX," "TorchScript Mobile," and "Quantized." This parameter is essential as it guides the node on how to correctly interpret and load the model file. There are no default values, and you must specify one of the supported formats.

load_optimizer

The load_optimizer parameter is a boolean option that specifies whether to load the optimizer state along with the model. If set to "True," the node will attempt to load the optimizer state if it is available in the model file. This can be useful for resuming training from a saved state. The default value is typically "False."

NNT Load Model Output Parameters:

model

The model output parameter represents the loaded neural network model. This output is crucial as it provides you with the model object that can be used for inference, further training, or analysis. The model's structure and weights are restored based on the specified load format.

status_msg

The status_msg output parameter provides a message indicating the success or failure of the model loading process. It offers insights into what was loaded and from where, helping you verify that the correct model and optimizer (if applicable) have been loaded successfully.

NNT Load Model Usage Tips:

• Ensure that the filename and directory parameters are correctly specified to avoid file not found errors.
• Choose the appropriate load_format based on the model file you are working with to ensure successful loading.
• If you plan to continue training a model, set load_optimizer to "True" to load the optimizer state, if available.

NNT Load Model Common Errors and Solutions:

Invalid state dict file

• Explanation: This error occurs when the file specified does not contain a valid state dictionary or is missing the model_state_dict key.
• Solution: Verify that the file is a valid state dictionary and contains the necessary keys. Ensure the file path and format are correct.

Model and optimizer loaded from <file_path>

• Explanation: This message indicates that both the model and optimizer were successfully loaded from the specified file path.
• Solution: No action is needed as this is a confirmation of successful loading.

Model loaded from <file_path>

• Explanation: This message indicates that the model was successfully loaded, but the optimizer was not loaded, possibly because load_optimizer was set to "False" or the optimizer state was not available.
• Solution: If you intended to load the optimizer, check the load_optimizer setting and ensure the optimizer state is present in the file.