Trouble loading a saved model in tfjs-node, "Op type not registered 'DisableCopyOnRead'"

[StratosGiak]

System Information

• OS Platform and Distribution: Windows 11 and Ubuntu 22.04
• TensorFlow.js installed from: npm
• tfjs-node version: 4.18.0

I have a SavedModel made by exporting this model using Python with Tensorflow 2.16.1. The model works fine when loaded back into Python, however when I try to load it in Node using tfjs.node.loadSavedModel() I get the following errors

E tensorflow/core/grappler/optimizers/meta_optimizer.cc:903] tfg_optimizer{} failed: NOT_FOUND: Op type not registered 'DisableCopyOnRead' in binary running on LAPTOP. Make sure the Op and Kernel are registered in the binary running in this process. Note that if you are loading a saved graph which used ops from tf.contrib, accessing (e.g.) `tf.contrib.resampler` should be done before importing the graph, as contrib ops are lazily registered when the module is first accessed.
        While importing FunctionDef: __inference__traced_save_11547
        when importing GraphDef to MLIR module in GrapplerHook
E tensorflow/core/framework/node_def_util.cc:630] NodeDef mentions attribute debug_name which is not in the op definition: Op<name=VarHandleOp; signature= -> resource:resource; attr=container:string,default=""; attr=shared_name:string,default=""; attr=dtype:type; attr=shape:shape; attr=allowed_devices:list(string),default=[]; is_stateful=true> This may be expected if your graph generating binary is newer  than this binary. Unknown attributes will be ignored. NodeDef: {{node Variable}}
I tensorflow/cc/saved_model/loader.cc:212] Running initialization op on SavedModel bundle at path: ./models/opennsfw2
E tensorflow/core/grappler/optimizers/meta_optimizer.cc:903] tfg_optimizer{} failed: NOT_FOUND: Op type not registered 'DisableCopyOnRead' in binary running on LAPTOP. Make sure the Op and Kernel are registered in the binary running in this process. Note that if you are loading a saved graph which used ops from tf.contrib, accessing (e.g.) `tf.contrib.resampler` should be done before importing the graph, as contrib ops are lazily registered when the module is first accessed.
        While importing FunctionDef: __inference__traced_save_11547
        when importing GraphDef to MLIR module in GrapplerHook
I tensorflow/cc/saved_model/loader.cc:301] SavedModel load for tags { serve }; Status: success: OK. Took 1504673 microseconds.
C:\dev\Projects\nsfw_test\node_modules\@tensorflow\tfjs-node\dist\nodejs_kernel_backend.js:435
        return this.binding.loadSavedModel(path, tags);
                            ^

Error: Failed to load SavedModel: Converting GraphDef to Graph has failed with an error: 'Op type not registered 'DisableCopyOnRead' in binary running on LAPTOP. Make sure the Op and Kernel are registered in the binary running in this process. Note that if you are loading a saved graph which used ops from tf.contrib, accessing (e.g.) `tf.contrib.resampler` should be done before importing the graph, as contrib ops are lazily registered when the module is first accessed.' The binary trying to i
    at NodeJSKernelBackend.loadSavedModelMetaGraph (C:\dev\Projects\nsfw_test\node_modules\@tensorflow\tfjs-node\dist\nodejs_kernel_backend.js:435:29)
    at Object.<anonymous> (C:\dev\Projects\nsfw_test\node_modules\@tensorflow\tfjs-node\dist\saved_model.js:448:45)
    at step (C:\dev\Projects\nsfw_test\node_modules\@tensorflow\tfjs-node\dist\saved_model.js:49:23)
    at Object.next (C:\dev\Projects\nsfw_test\node_modules\@tensorflow\tfjs-node\dist\saved_model.js:30:53)
    at fulfilled (C:\dev\Projects\nsfw_test\node_modules\@tensorflow\tfjs-node\dist\saved_model.js:21:58)

I'm not sure what the problem is, or if it strictly has to do with tfjs to be honest. Is the DisableCopyOnRead op not supported by tfjs-node? Is there a workaround, perhaps by exporting the model some other way? I also tried exporting it in Python with Tensorflow version 2.14.0, the earliest one allowed by the package containing the model, but that had no effect.