hsiaoer

class Model(tf.keras.Model):
    ...
    def __call__(self, inputs, training):
        # Input layer
        y = tf.reshape(inputs, self._input_shape)
        y = self.conv1(y)
        y = self.max_pool2d(y)
        y = self.conv2(y)
        y = self.max_pool2d(y)
        y = tf.layers.flatten(y)

hsiaoer

def model_fn(features, labels, mode, params):
    ...
    logits = model(image, training=False)

    predictions = {
      'classes': tf.argmax(logits, axis=1),
      'probabilities': tf.nn.softmax(logits, name='softmax_tensor'),
    }
    if mode == tf.estimator.ModeKeys.PREDICT:
      return tf.estimator.EstimatorSpec(

hsiaoer

python official/mnist/mnist.py --export_dir /tmp/mnist_saved_model --model-dir /tmp/mnist_graph_def_with_ckpts

hsiaoer

>> ls /tmp/mnist_graph_def_with_ckpts
checkpoint
model.ckpt-48000
model.ckpt-35626
model.ckpt-39410
model.ckpt-43218
model.ckpt-47043
model.ckpt-48000
graph.pbtxt

hsiaoer

# From anywhere though I suggest you make it outside of the git repos
mkdir training_summaries

# Runs tensorboard in the background at http://localhost:6006
tensorboard --logdir training_summaries & 

# Using my modified import_pb_to_tensorboard.py in the tensorflow repo (feel free to edit to your liking)
import_pb_to_tensorboard.py --model_dir /tmp/mnist_graph_def_with_ckpts/graph.pbtxt --log_dir training_summaries/mnist --graph_type=PbTxt

hsiaoer

X = df_final.drop([‘fraud_reported_Y’,’policy_csl’,’policy_bind_date’,’incident_date’],axis=1).values
y = df_final[‘fraud_reported_Y’].values

hsiaoer

feats = [‘policy_state’,’insured_sex’,’insured_education_level’,’insured_occupation’,’insured_hobbies’,’insured_relationship’,’collision_type’,’incident_severity’,’authorities_contacted’,’incident_state’,’incident_city’,’incident_location’,’property_damage’,’police_report_available’,’auto_make’,’auto_model’,’fraud_reported’,’incident_type’]
df_final = pd.get_dummies(df,columns=feats,drop_first=True)

hsiaoer

X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.3)

hsiaoer

classifier.compile(optimizer= ‘adam’,loss = ‘binary_crossentropy’,metrics = [‘accuracy’])

hsiaoer

def make_classifier():
    classifier = Sequential()
    classiifier.add(Dense(3, kernel_initializer = ‘uniform’, activation = ‘relu’, input_dim=5))
    classiifier.add(Dense(3, kernel_initializer = ‘uniform’, activation = ‘relu’))
    classifier.add(Dense(1, kernel_initializer = ‘uniform’, activation = ‘sigmoid’))
    classifier.compile(optimizer= ‘adam’,loss = ‘binary_crossentropy’,metrics = [‘accuracy’])
    return classifier