'eval_batch_size': 200,
'epochs': 20,
'evaluate_test': True,
'eval_flexible': False,
'save_dir': 'temp-sms-spam/model-zoo/example-3-2-{date}-{random_id}',
'save_accuracy_limit': 0.97,
}
data = rnn_model(params)
solver = hype.TensorflowSolver(data=data,
hyper_params=params,
**solver_params)
return solver
hyper_params_spec = hype.spec.new(
max_sequence_length=hype.spec.choice(range(20, 50)),
min_frequency=hype.spec.choice([1, 3, 5, 10]),
embedding_size=hype.spec.choice([32, 64, 128]),
rnn_cell=hype.spec.choice(['basic_rnn', 'lstm', 'gru']),
rnn_hidden_size=hype.spec.choice([16, 32, 64]),
dropout_keep_prob=hype.spec.uniform(0.5, 1.0),
learning_rate=10**hype.spec.uniform(-4, -3),
)
strategy_params = {
'io_load_dir': 'temp-sms-spam/example-3-2',
'io_save_dir': 'temp-sms-spam/example-3-2',
}
tuner = hype.HyperTuner(hyper_params_spec, solver_generator, **strategy_params)
tuner.tune()
layer = tf.add(tf.matmul(tf.sigmoid(layer), weights[2]), biases[2])
output = tf.add(tf.matmul(tf.sigmoid(layer), weights['output']),
biases['output'])
cost = tf.reduce_mean(tf.squared_difference(output, y), name='loss')
optimizer = tf.train.GradientDescentOptimizer(
learning_rate=params.learning_rate).minimize(cost, name='minimize')
tf.reduce_mean(tf.cast(tf.abs(output - y) < 0.5, tf.float32),
name='accuracy')
def solver_generator(params):
solver_params = {
'batch_size': 167,
'epochs': 50,
'evaluate_test': True,
'eval_flexible': True,
}
dnn_model(params)
solver = hype.TensorflowSolver(data=data,
hyper_params=params,
**solver_params)
return solver
hyper_params_spec = hype.spec.new(learning_rate=10**hype.spec.uniform(-1,
-3), )
tuner = hype.HyperTuner(hyper_params_spec, solver_generator)
tuner.tune()
