"Model failed to serialize in JSON" in Food Vision Big model fitting - Milestone Project 1

[MercurialOutsider]

I'm getting an error when I try to fit the feature extraction model for Food Vision Big. This is the code:

Fit the feature extraction model with callbacks

history_101_food_classes_feature_extract = model.fit(train_data,
                                                     epochs=3,
                                                     steps_per_epoch=len(train_data),
                                                     validation_data=test_data,
                                                     validation_steps=int(0.15 * len(test_data)),
                                                     callbacks=[create_tensorboard_callback(dir_name="training_logs",
                                                                                            experiment_name="efficientnetb0_101_classes_all_data_feature_extract"),
                                                                model_checkpoint])

[santiagorg2401]

Hey,
Are you saving only the weights on that model_checkpoint callback? An issue on Tf2.10.0> is causing troubles when serializing to save the entire model.

[MercurialOutsider]

That was definitely the issue! In the link you sent, the poster used an older version of TensorFlow, I just set save_weights_only to False in the model_checkpoint callback like you mentioned and it worked. Thank you for the help!

[santiagorg2401]

Anytime :D