ComfyUI Node: Load Qwen3.5 ViT (Anima)

Load Qwen3.5 ViT (Anima):

The LoadQwen35ViT node is designed to load the Qwen 3.5 Vision Transformer (ViT) model, which is a crucial component for image-conditioned generation within the Anima 2B framework. This node facilitates the integration of visual features extracted from reference images into the text encoder's token sequence, thereby enabling advanced functionalities such as style transfer, character consistency, and image-guided generation. By leveraging the Vision Transformer, this node enhances the creative possibilities for AI artists, allowing for more nuanced and visually coherent outputs. The primary goal of this node is to ensure that the ViT model is correctly loaded and ready for use, providing a seamless experience for users who wish to incorporate image-based conditioning into their generative processes.

Load Qwen3.5 ViT (Anima) Input Parameters:

vit_name

The vit_name parameter specifies the name of the Vision Transformer weights file to be loaded. This parameter is crucial as it determines which set of pre-trained weights will be used by the ViT model, directly impacting the model's performance and the quality of the generated outputs. The available options for this parameter are typically files with the .safetensors extension, such as qwen35_vit.safetensors. Users should ensure that the specified file exists in the appropriate directory, as the node will attempt to locate and load the weights from there. This parameter does not have a default value, and users must provide a valid file name to proceed.

Load Qwen3.5 ViT (Anima) Output Parameters:

vit

The vit output parameter represents the loaded Vision Transformer model, ready for use in image-conditioned generation tasks. This output is crucial as it encapsulates the model's ability to process and integrate visual features into the generative process, thereby enhancing the creative capabilities of the Anima 2B framework. The vit model is returned as a tuple, ensuring that it can be easily utilized in subsequent nodes or processes within the workflow. This output is essential for users who wish to leverage the full potential of image-guided generation, providing a robust foundation for creating visually coherent and stylistically consistent outputs.

Load Qwen3.5 ViT (Anima) Usage Tips:

• Ensure that the vit_name parameter is correctly set to a valid weights file, such as qwen35_vit.safetensors, to avoid loading errors and ensure optimal model performance.
• Regularly update the weights file to the latest version to benefit from improvements and enhancements in the Vision Transformer model, which can lead to better image-conditioned generation results.

Load Qwen3.5 ViT (Anima) Common Errors and Solutions:

ViT weights not found: <vit_name>

• Explanation: This error occurs when the specified weights file cannot be found in the expected directory. It indicates that the file name provided in the vit_name parameter does not match any existing files.
• Solution: Verify that the vit_name parameter is set to the correct file name and that the file exists in the designated directory. If the file is missing, run extract_vit.py to download and extract the necessary weights from Qwen/Qwen3.5-4B.

Missing keys: <missing_keys>

• Explanation: This warning indicates that some expected keys are missing from the loaded weights file, which may affect the model's performance.
• Solution: Ensure that the weights file is complete and not corrupted. Consider re-downloading or regenerating the weights file using extract_vit.py to ensure all necessary keys are present.

Unexpected keys: <unexpected_keys>

• Explanation: This warning suggests that there are additional keys in the weights file that are not expected by the model, which might lead to inconsistencies.
• Solution: Check if the weights file is compatible with the current version of the model. If necessary, update the model or the weights file to ensure compatibility and remove any unexpected keys.