Keras tuner not generating check point for every trial

[Geerthy1130]

Hi everyone
I am performing a hyperparameter tuning using a Keras tuner and for most of the trials it doesn’t generate a checkpoint; but for every trial, there is a file called trial(JSON file). Is it possible to create the model out of the JSON file? I tried to extract the hyperparameters successfully but I couldn’t generate the model out of it and I got the following error.

Cell In[8], line 1
    trial_model = tuner_bo.hypermodel.build(hyperparameters)

  File c:\users\geerthy\ejection fraction project\echo_codes\tunermr_ef_median_2l_earlyfusion_with_fc.py:69 in build
    fl_1 = hp.Choice('n_fil_1', values = [2,4, 8, 16, 32, 64])

AttributeError: 'dict' object has no attribute 'Choice` 

Here’s how I call it

best_trials = tuner_bo.oracle.get_best_trials(num_trials=50)
for i in range(0, len(best_trials), 50):
    for j in range(i, min(i + 50, len(best_trials))):
        trial = best_trials[j]
        hyperparameters = trial.hyperparameters.values

        # Train the model with the best hyperparameters
        trial_model = tuner_bo.hypermodel.build(hyperparameters)

Here’s my model class :

class my_tuner_model(HyperModel):
  
    def __init__(self, input_shape):
        self.input_shape = input_shape
    def build(self, hp):
        
        RawInputECG = Input(shape=(12,301,1))
        fl_1 = hp.Choice('n_fil_1', values = [2,4, 8, 16, 32, 64])
        fw_1 = hp.Choice('filWidth_1', values = [2, 5, 10, 25, 50, 100, 125])
        fl_2 = hp.Choice('n_fil_2', values = [2,4, 8, 16, 32, 64])
        fw_2 = hp.Choice('filWidth_2', values = [2, 5, 10, 25, 50, 100, 125])
        convECG1 = Conv2D(filters=fl_1, kernel_size= (12,fw_1), activation=hp.Choice('activation1', ['sigmoid', 'relu', 'tanh']))(RawInputECG)
        flat = Flatten()(convECG1)
        output = Dense(1)(output_FC)  
        model = Model(inputs=[RawInputECG], outputs=output)
        
        optim = hp.Choice('optimizer', values=['adam', 'RMSprop', 'SGD'])
        learn_rate = hp.Choice('learning_rate', values=[1e-1, 1e-2, 1e-3, 1e-4])
                            
        if optim == 'adam':
            optimizers=tf.keras.optimizers.Adam(learning_rate=learn_rate, beta_1=0.9, beta_2=0.999, amsgrad=False)
        elif optim == 'RMSprop':
            optimizers=tf.keras.optimizers.RMSprop(learning_rate=learn_rate, rho=0.9, momentum=0.0, epsilon=1e-07, centered=False)
        elif optim == 'SGD':
            optimizers=tf.keras.optimizers.SGD(learning_rate=learn_rate, momentum=0.0, nesterov=False)
                
        auc = tf.keras.metrics.MeanAbsoluteError()
        model.compile(optimizer=optimizers, loss='mse', metrics=[tf.keras.metrics.RootMeanSquaredError(), auc])
        
        return model 
hypermodel = my_tuner_model(input_shape)

Any suggestions would be helpful
Thank you