bug: Cannot save model from model.save(PATH)

[slavazhulanov]

I trained the model with this tutorial. At the end, the WEIGHTS are saved, but I need to save the MODEL. When I save the MODEL I get the error:

ValueError: Model <keras_cv.models.object_detection.retina_net.retina_net.RetinaNet object at 0x00000287F451B6D0> cannot be saved either because the input shape is not available or because the forward pass of the model is not defined.To define a forward pass, please override Model.call(). To specify an input shape, either call build(input_shape)directly, or call the model on actual data usingModel(), Model.fit(), or Model.predict(). If you have a custom training step, please make sure to invoke the forward pass in train step through Model.call, i.e. model(inputs), as opposed to model.call().

Next, I plan to convert the MODEL to ONNX format and use it in OpenCV.

[jbischof]

The error message is correct! Your model is not built yet.

To specify an input shape, either call build(input_shape) directly, or call the model on actual data using Model(), Model.fit(), or Model.predict().

[slavazhulanov]

@jbischof Could you take a look and tell me what I'm doing wrong?

[slavazhulanov]

The behavior of the save() method changes after a call to predict()

[slavazhulanov]

WARNING:absl:Found untraced functions such as retina_net_label_encoder_layer_call_fn, retina_net_label_encoder_layer_call_and_return_conditional_losses, anchor_generator_layer_call_fn, anchor_generator_layer_call_and_return_conditional_losses, multi_class_non_max_suppression_layer_call_fn while saving (showing 5 of 110). These functions will not be directly callable after loading.

[lb-97]

Hi @slavazhulanov, I tried to run the colab tutorial and I was able to save my model. Could you post your colab here, so that I can take a full look at you code?

[slavazhulanov]

Hello @lb-97 I have attached a link here. The tutorial uses .save_weights() and my error is related to .save()

[ID6109]

In callbacks, try to change save_weights_only=True to False. It should work then after rerunning the entire code. @slavazhulanov

[slavazhulanov]

@ID6109 it didn't change anything. You can try it yourself on the link to Google Colab

[ID6109]

@slavazhulanov Try to save the model as model.save(CHECKPOINT_PATH, include_optimizer = None, save_traces = None)

[ianstenbit]

Closing as stale -- it seems like KerasCV is WAI in this case

[Aillian]

having the same issue when i model.save as .h5/.keras i get: ValueError: Unable to create dataset (name already exists)

when i model.save as .tf , it saves the model, but when loading it i get:

ValueError: `RetinaNet` does not accept a `loss` to `compile()`. Instead, please pass `box_loss` and `classification_loss`. `loss` will be ignored during training.

what is the problem? what is the correct way to save and load the model?

[jbischof]

@Aillian can you send along a colab gist replicating your issue?