Quickstart Tutorial¶
In this short introduction to the Masterful AutoML Platform, you will train a model start to finish with Masterful.
First, import the necessary libraries.
[1]:
import tensorflow as tf
import tensorflow_datasets as tfds
import sys
sys.path.append('../../../')
import masterful
Use Tensorflow Datasets to load CIFAR-10.
[2]:
TRAINPER=5
(train, test) = tfds.load('cifar10', as_supervised=True, batch_size=32, split=[f'train[:{TRAINPER}%]', 'test[:10%]'],with_info=False)
2021-09-14 18:45:50.621668: I tensorflow/core/platform/cpu_feature_guard.cc:142] This TensorFlow binary is optimized with oneAPI Deep Neural Network Library (oneDNN) to use the following CPU instructions in performance-critical operations: AVX2 FMA
To enable them in other operations, rebuild TensorFlow with the appropriate compiler flags.
Masterful works best with dense labels, so apply one-hot to your sparse labels.
[3]:
def one_hot(image, label):
return image, tf.one_hot(label, 10)
train = train.map(one_hot)
test = test.map(one_hot)
Build a simple model to classify CIFAR-10.
[4]:
def get_model():
model = tf.keras.models.Sequential()
model.add(tf.keras.layers.experimental.preprocessing.Rescaling(1./255, input_shape=(32, 32, 3)))
model.add(tf.keras.layers.Conv2D(16, (3, 3), activation='relu',))
model.add(tf.keras.layers.MaxPooling2D((2, 2)))
model.add(tf.keras.layers.Conv2D(32, (3, 3), activation='relu',))
model.add(tf.keras.layers.MaxPooling2D((2, 2)))
model.add(tf.keras.layers.Conv2D(64, (3, 3), activation='relu',))
model.add(tf.keras.layers.MaxPooling2D((2, 2)))
model.add(tf.keras.layers.GlobalAveragePooling2D())
model.add(tf.keras.layers.Dense(10, activation='softmax'))
return model
model = get_model()
Train the model to completion.
[5]:
model.compile(optimizer='adam',
loss=tf.keras.losses.CategoricalCrossentropy(from_logits=False),
metrics=['accuracy'])
early_stopping = tf.keras.callbacks.EarlyStopping(patience=10, verbose=1)
reduce_lr = tf.keras.callbacks.ReduceLROnPlateau(patience=4, verbose=1)
model.fit(train, validation_data=test, callbacks=[early_stopping, reduce_lr], epochs=tf.int32.max, verbose=2)
2021-09-14 18:45:51.460624: I tensorflow/compiler/mlir/mlir_graph_optimization_pass.cc:185] None of the MLIR Optimization Passes are enabled (registered 2)
Epoch 1/2147483647
79/79 - 3s - loss: 2.2747 - accuracy: 0.1304 - val_loss: 2.2188 - val_accuracy: 0.1630
Epoch 2/2147483647
79/79 - 1s - loss: 2.0878 - accuracy: 0.2176 - val_loss: 1.9740 - val_accuracy: 0.2360
Epoch 3/2147483647
79/79 - 1s - loss: 1.9572 - accuracy: 0.2708 - val_loss: 1.8702 - val_accuracy: 0.2750
Epoch 4/2147483647
79/79 - 1s - loss: 1.8666 - accuracy: 0.3068 - val_loss: 1.8110 - val_accuracy: 0.2860
Epoch 5/2147483647
79/79 - 1s - loss: 1.7961 - accuracy: 0.3400 - val_loss: 1.7728 - val_accuracy: 0.3110
Epoch 6/2147483647
79/79 - 1s - loss: 1.7362 - accuracy: 0.3596 - val_loss: 1.7350 - val_accuracy: 0.3370
Epoch 7/2147483647
79/79 - 1s - loss: 1.6865 - accuracy: 0.3808 - val_loss: 1.7020 - val_accuracy: 0.3620
Epoch 8/2147483647
79/79 - 1s - loss: 1.6432 - accuracy: 0.4080 - val_loss: 1.6791 - val_accuracy: 0.3730
Epoch 9/2147483647
79/79 - 1s - loss: 1.6032 - accuracy: 0.4284 - val_loss: 1.6526 - val_accuracy: 0.3800
Epoch 10/2147483647
79/79 - 1s - loss: 1.5683 - accuracy: 0.4464 - val_loss: 1.6310 - val_accuracy: 0.3920
Epoch 11/2147483647
79/79 - 1s - loss: 1.5354 - accuracy: 0.4568 - val_loss: 1.6151 - val_accuracy: 0.4010
Epoch 12/2147483647
79/79 - 1s - loss: 1.5049 - accuracy: 0.4632 - val_loss: 1.5887 - val_accuracy: 0.4210
Epoch 13/2147483647
79/79 - 1s - loss: 1.4763 - accuracy: 0.4724 - val_loss: 1.5797 - val_accuracy: 0.4200
Epoch 14/2147483647
79/79 - 1s - loss: 1.4485 - accuracy: 0.4764 - val_loss: 1.5675 - val_accuracy: 0.4170
Epoch 15/2147483647
79/79 - 1s - loss: 1.4224 - accuracy: 0.4892 - val_loss: 1.5543 - val_accuracy: 0.4230
Epoch 16/2147483647
79/79 - 1s - loss: 1.3973 - accuracy: 0.4984 - val_loss: 1.5448 - val_accuracy: 0.4300
Epoch 17/2147483647
79/79 - 1s - loss: 1.3735 - accuracy: 0.5064 - val_loss: 1.5414 - val_accuracy: 0.4320
Epoch 18/2147483647
79/79 - 1s - loss: 1.3503 - accuracy: 0.5148 - val_loss: 1.5421 - val_accuracy: 0.4400
Epoch 19/2147483647
79/79 - 1s - loss: 1.3279 - accuracy: 0.5268 - val_loss: 1.5367 - val_accuracy: 0.4460
Epoch 20/2147483647
79/79 - 1s - loss: 1.3054 - accuracy: 0.5400 - val_loss: 1.5309 - val_accuracy: 0.4430
Epoch 21/2147483647
79/79 - 1s - loss: 1.2828 - accuracy: 0.5464 - val_loss: 1.5276 - val_accuracy: 0.4400
Epoch 22/2147483647
79/79 - 1s - loss: 1.2613 - accuracy: 0.5516 - val_loss: 1.5298 - val_accuracy: 0.4380
Epoch 23/2147483647
79/79 - 1s - loss: 1.2421 - accuracy: 0.5620 - val_loss: 1.5295 - val_accuracy: 0.4440
Epoch 24/2147483647
79/79 - 1s - loss: 1.2234 - accuracy: 0.5700 - val_loss: 1.5303 - val_accuracy: 0.4400
Epoch 25/2147483647
79/79 - 1s - loss: 1.2044 - accuracy: 0.5832 - val_loss: 1.5313 - val_accuracy: 0.4480
Epoch 00025: ReduceLROnPlateau reducing learning rate to 0.00010000000474974513.
Epoch 26/2147483647
79/79 - 1s - loss: 1.1333 - accuracy: 0.6120 - val_loss: 1.4211 - val_accuracy: 0.4940
Epoch 27/2147483647
79/79 - 1s - loss: 1.1132 - accuracy: 0.6256 - val_loss: 1.4189 - val_accuracy: 0.4910
Epoch 28/2147483647
79/79 - 1s - loss: 1.1075 - accuracy: 0.6244 - val_loss: 1.4190 - val_accuracy: 0.4910
Epoch 29/2147483647
79/79 - 1s - loss: 1.1029 - accuracy: 0.6252 - val_loss: 1.4195 - val_accuracy: 0.4920
Epoch 30/2147483647
79/79 - 1s - loss: 1.0990 - accuracy: 0.6256 - val_loss: 1.4197 - val_accuracy: 0.4910
Epoch 31/2147483647
79/79 - 1s - loss: 1.0954 - accuracy: 0.6264 - val_loss: 1.4198 - val_accuracy: 0.4920
Epoch 00031: ReduceLROnPlateau reducing learning rate to 1.0000000474974514e-05.
Epoch 32/2147483647
79/79 - 1s - loss: 1.0900 - accuracy: 0.6296 - val_loss: 1.4106 - val_accuracy: 0.4980
Epoch 33/2147483647
79/79 - 1s - loss: 1.0827 - accuracy: 0.6308 - val_loss: 1.4100 - val_accuracy: 0.5020
Epoch 34/2147483647
79/79 - 1s - loss: 1.0817 - accuracy: 0.6340 - val_loss: 1.4100 - val_accuracy: 0.5040
Epoch 35/2147483647
79/79 - 1s - loss: 1.0811 - accuracy: 0.6344 - val_loss: 1.4101 - val_accuracy: 0.5040
Epoch 36/2147483647
79/79 - 1s - loss: 1.0807 - accuracy: 0.6352 - val_loss: 1.4101 - val_accuracy: 0.5040
Epoch 37/2147483647
79/79 - 1s - loss: 1.0803 - accuracy: 0.6352 - val_loss: 1.4101 - val_accuracy: 0.5030
Epoch 00037: ReduceLROnPlateau reducing learning rate to 1.0000000656873453e-06.
Epoch 38/2147483647
79/79 - 1s - loss: 1.0785 - accuracy: 0.6340 - val_loss: 1.4100 - val_accuracy: 0.5030
Epoch 39/2147483647
79/79 - 1s - loss: 1.0784 - accuracy: 0.6344 - val_loss: 1.4099 - val_accuracy: 0.5030
Epoch 40/2147483647
79/79 - 1s - loss: 1.0783 - accuracy: 0.6344 - val_loss: 1.4099 - val_accuracy: 0.5030
Epoch 41/2147483647
79/79 - 1s - loss: 1.0782 - accuracy: 0.6344 - val_loss: 1.4098 - val_accuracy: 0.5030
Epoch 42/2147483647
79/79 - 1s - loss: 1.0782 - accuracy: 0.6356 - val_loss: 1.4098 - val_accuracy: 0.5020
Epoch 43/2147483647
79/79 - 1s - loss: 1.0781 - accuracy: 0.6348 - val_loss: 1.4098 - val_accuracy: 0.5020
Epoch 44/2147483647
79/79 - 1s - loss: 1.0781 - accuracy: 0.6348 - val_loss: 1.4098 - val_accuracy: 0.5020
Epoch 45/2147483647
79/79 - 1s - loss: 1.0780 - accuracy: 0.6356 - val_loss: 1.4098 - val_accuracy: 0.5020
Epoch 00045: ReduceLROnPlateau reducing learning rate to 1.0000001111620805e-07.
Epoch 46/2147483647
79/79 - 1s - loss: 1.0778 - accuracy: 0.6348 - val_loss: 1.4098 - val_accuracy: 0.5020
Epoch 47/2147483647
79/79 - 1s - loss: 1.0778 - accuracy: 0.6348 - val_loss: 1.4098 - val_accuracy: 0.5020
Epoch 48/2147483647
79/79 - 1s - loss: 1.0778 - accuracy: 0.6348 - val_loss: 1.4098 - val_accuracy: 0.5020
Epoch 49/2147483647
79/79 - 1s - loss: 1.0778 - accuracy: 0.6348 - val_loss: 1.4098 - val_accuracy: 0.5020
Epoch 00049: ReduceLROnPlateau reducing learning rate to 1.000000082740371e-08.
Epoch 50/2147483647
79/79 - 1s - loss: 1.0778 - accuracy: 0.6348 - val_loss: 1.4098 - val_accuracy: 0.5020
Epoch 51/2147483647
79/79 - 1s - loss: 1.0778 - accuracy: 0.6348 - val_loss: 1.4098 - val_accuracy: 0.5020
Epoch 52/2147483647
79/79 - 1s - loss: 1.0778 - accuracy: 0.6348 - val_loss: 1.4098 - val_accuracy: 0.5020
Epoch 53/2147483647
79/79 - 1s - loss: 1.0778 - accuracy: 0.6348 - val_loss: 1.4098 - val_accuracy: 0.5020
Epoch 00053: ReduceLROnPlateau reducing learning rate to 1.000000082740371e-09.
Epoch 00053: early stopping
[5]:
<keras.callbacks.History at 0x7fcad4a9ae80>
Evaluate the model to see how well it did without Masterful. You should get a model around 51% accurate. In practice, you would need to run multiple experiments and run a t-test to get a statistically significant result.
[6]:
model.evaluate(test)
32/32 [==============================] - 0s 3ms/step - loss: 1.4098 - accuracy: 0.5020
[6]:
[1.4098398685455322, 0.5019999742507935]
Now train with Masterful. First, specify some design choices about your model by using masterful.spec.create_model_and_data_specs. Next, call masterful.autofit, which runs Masterful’s metalearning algorithm and then applies it.
[7]:
del model
model = get_model()
model_spec, data_spec = masterful.spec.create_model_and_data_specs(
model=model,
dataset=train.unbatch(),
task=masterful.spec.Task.CLASSIFICATION,
image_range=masterful.spec.ImageRange.ZERO_255,
num_classes=10,
sparse=False,
from_logits=False)
masterful.autofit(model, model_spec, train.unbatch(), data_spec, validation_data=test.unbatch())
WARNING:tensorflow:From ../../../masterful/utils/find_batchsize.py:40: is_gpu_available (from tensorflow.python.framework.test_util) is deprecated and will be removed in a future version.
Instructions for updating:
Use `tf.config.list_physical_devices('GPU')` instead.
WARNING:tensorflow:From ../../../masterful/utils/find_batchsize.py:40: is_gpu_available (from tensorflow.python.framework.test_util) is deprecated and will be removed in a future version.
Instructions for updating:
Use `tf.config.list_physical_devices('GPU')` instead.
../../../masterful/utils/find_batchsize.py:47: UserWarning: WARNING: No GPU available. Using fixed batch size of 32
warnings.warn(
500/500 [==============================] - 6s 10ms/step - loss: 2.2633
The logs are being saved to /Users/yaosh/.masterful/UTC_2021-09-14__22-46-49.945599/log-analyze-UTC_2021-09-14__22-46-49.945599.txt
The logs are being saved to /Users/yaosh/.masterful/UTC_2021-09-14__22-46-49.945599/log-analyze-UTC_2021-09-14__22-46-49.945599.txt
Masterful backend.find_augmentation_policy() STARTING...
Baseline analysis STARTING... Args: {'x': <BatchDataset shapes: ((None, 32, 32, 3), (None, 10)), types: (tf.uint8, tf.float32)>, 'epochs': 2147483647, 'validation_data': <BatchDataset shapes: ((None, 32, 32, 3), (None, 10)), types: (tf.uint8, tf.float32)>, 'callbacks': [<keras.callbacks.EarlyStopping object at 0x7fcab64c75e0>, <keras.callbacks.ReduceLROnPlateau object at 0x7fcab64c7520>]}
Masterful backend.analyze() DONE.
A file with name: interim-bridge-analyze.pstpb was found while override was set to False.
500/500 [==============================] - 7s 11ms/step - loss: 0.9418
The logs are being saved to /Users/yaosh/.masterful/UTC_2021-09-14__23-25-06.931697/log-fit-UTC_2021-09-14__23-25-06.931697.txt
The logs are being saved to /Users/yaosh/.masterful/UTC_2021-09-14__23-25-06.931697/log-fit-UTC_2021-09-14__23-25-06.931697.txt
Masterful backend.fit() STARTING...
Loaded config from /Users/yaosh/.masterful/policies/interim-bridge-fit
Masterful backend.fit with policy /Users/yaosh/.masterful/policies/interim-bridge-fit phase 1 of 2 STARTING...
Epoch 1/2147483647
79/79 [==============================] - 19s 51ms/step - loss: 1.5700 - val_loss: 1.4144
Epoch 2/2147483647
79/79 [==============================] - 4s 48ms/step - loss: 1.5917 - val_loss: 1.3346
Epoch 3/2147483647
79/79 [==============================] - 4s 48ms/step - loss: 1.5551 - val_loss: 1.3347
Epoch 4/2147483647
79/79 [==============================] - 4s 48ms/step - loss: 1.5999 - val_loss: 1.3054
Epoch 5/2147483647
79/79 [==============================] - 4s 48ms/step - loss: 1.5796 - val_loss: 1.2822
Epoch 6/2147483647
79/79 [==============================] - 4s 48ms/step - loss: 1.5731 - val_loss: 1.3102
Epoch 7/2147483647
79/79 [==============================] - 4s 48ms/step - loss: 1.5587 - val_loss: 1.3041
Epoch 8/2147483647
79/79 [==============================] - 4s 49ms/step - loss: 1.5436 - val_loss: 1.3193
Epoch 9/2147483647
79/79 [==============================] - 4s 49ms/step - loss: 1.5609 - val_loss: 1.3037
Epoch 10/2147483647
79/79 [==============================] - 4s 49ms/step - loss: 1.5560 - val_loss: 1.2685
Epoch 11/2147483647
79/79 [==============================] - 4s 49ms/step - loss: 1.5782 - val_loss: 1.3367
Epoch 12/2147483647
79/79 [==============================] - 4s 49ms/step - loss: 1.5487 - val_loss: 1.2593
Epoch 13/2147483647
79/79 [==============================] - 4s 49ms/step - loss: 1.5633 - val_loss: 1.3003
Epoch 14/2147483647
79/79 [==============================] - 4s 49ms/step - loss: 1.5642 - val_loss: 1.3707
Epoch 15/2147483647
79/79 [==============================] - 4s 49ms/step - loss: 1.5522 - val_loss: 1.3814
Epoch 16/2147483647
79/79 [==============================] - 4s 52ms/step - loss: 1.5609 - val_loss: 1.3255
Epoch 17/2147483647
79/79 [==============================] - 4s 48ms/step - loss: 1.5218 - val_loss: 1.2996
Epoch 18/2147483647
79/79 [==============================] - 4s 50ms/step - loss: 1.5129 - val_loss: 1.2794
Epoch 19/2147483647
79/79 [==============================] - 4s 48ms/step - loss: 1.4661 - val_loss: 1.2664
Epoch 20/2147483647
79/79 [==============================] - 4s 48ms/step - loss: 1.4742 - val_loss: 1.2593
Epoch 21/2147483647
79/79 [==============================] - 4s 48ms/step - loss: 1.4712 - val_loss: 1.2543
Epoch 22/2147483647
79/79 [==============================] - 4s 48ms/step - loss: 1.4860 - val_loss: 1.2446
Epoch 23/2147483647
79/79 [==============================] - 4s 48ms/step - loss: 1.4615 - val_loss: 1.2713
Epoch 24/2147483647
79/79 [==============================] - 4s 48ms/step - loss: 1.4886 - val_loss: 1.2341
Epoch 25/2147483647
79/79 [==============================] - 4s 48ms/step - loss: 1.4912 - val_loss: 1.2808
Epoch 26/2147483647
79/79 [==============================] - 4s 48ms/step - loss: 1.4933 - val_loss: 1.2643
Epoch 27/2147483647
79/79 [==============================] - 4s 48ms/step - loss: 1.4550 - val_loss: 1.2435
Epoch 28/2147483647
79/79 [==============================] - 4s 48ms/step - loss: 1.4859 - val_loss: 1.2653
Epoch 29/2147483647
79/79 [==============================] - 4s 48ms/step - loss: 1.4823 - val_loss: 1.2459
Epoch 30/2147483647
79/79 [==============================] - 4s 48ms/step - loss: 1.4740 - val_loss: 1.2510
Epoch 31/2147483647
79/79 [==============================] - 4s 48ms/step - loss: 1.4524 - val_loss: 1.2561
Epoch 32/2147483647
79/79 [==============================] - 4s 49ms/step - loss: 1.4516 - val_loss: 1.2597
Epoch 33/2147483647
79/79 [==============================] - 4s 48ms/step - loss: 1.4719 - val_loss: 1.2360
Epoch 34/2147483647
79/79 [==============================] - 4s 48ms/step - loss: 1.4542 - val_loss: 1.2570
Epoch 35/2147483647
79/79 [==============================] - 4s 48ms/step - loss: 1.4308 - val_loss: 1.2507
Epoch 36/2147483647
79/79 [==============================] - 4s 48ms/step - loss: 1.4596 - val_loss: 1.2436
Epoch 37/2147483647
79/79 [==============================] - 4s 48ms/step - loss: 1.4558 - val_loss: 1.2471
Epoch 38/2147483647
79/79 [==============================] - 4s 49ms/step - loss: 1.4581 - val_loss: 1.2481
Epoch 39/2147483647
79/79 [==============================] - 4s 48ms/step - loss: 1.4448 - val_loss: 1.2445
Epoch 40/2147483647
79/79 [==============================] - 4s 48ms/step - loss: 1.4449 - val_loss: 1.2450
Epoch 41/2147483647
79/79 [==============================] - 4s 48ms/step - loss: 1.4487 - val_loss: 1.2459
Epoch 42/2147483647
79/79 [==============================] - 4s 48ms/step - loss: 1.4421 - val_loss: 1.2477
Epoch 43/2147483647
79/79 [==============================] - 4s 48ms/step - loss: 1.4520 - val_loss: 1.2484
Epoch 44/2147483647
79/79 [==============================] - 4s 48ms/step - loss: 1.4518 - val_loss: 1.2481
Masterful Restore LR: Restoring lr to 0.0012499999720603228.
Fitting model phase 1 of 2 DONE.
Masterful backend.fit with policy /Users/yaosh/.masterful/policies/interim-bridge-fit phase 2 of 2 STARTING...
Epoch 45/2147483691
79/79 [==============================] - 4s 51ms/step - loss: 2.4773 - val_loss: 1.8307
Epoch 46/2147483691
79/79 [==============================] - 4s 52ms/step - loss: 2.2624 - val_loss: 1.8694
Epoch 47/2147483691
79/79 [==============================] - 4s 51ms/step - loss: 2.2504 - val_loss: 1.8442
Epoch 48/2147483691
79/79 [==============================] - 4s 51ms/step - loss: 2.2481 - val_loss: 1.8308
Epoch 49/2147483691
79/79 [==============================] - 4s 52ms/step - loss: 2.2466 - val_loss: 1.8446
Epoch 50/2147483691
79/79 [==============================] - 4s 51ms/step - loss: 2.2454 - val_loss: 1.8291
Epoch 51/2147483691
79/79 [==============================] - 4s 51ms/step - loss: 2.2443 - val_loss: 1.8542
Epoch 52/2147483691
79/79 [==============================] - 4s 50ms/step - loss: 2.2436 - val_loss: 1.8459
Epoch 53/2147483691
79/79 [==============================] - 4s 51ms/step - loss: 2.2446 - val_loss: 1.8340
Epoch 54/2147483691
79/79 [==============================] - 4s 51ms/step - loss: 2.2428 - val_loss: 1.8411
Epoch 55/2147483691
79/79 [==============================] - 4s 51ms/step - loss: 2.2436 - val_loss: 1.8519
Epoch 56/2147483691
79/79 [==============================] - 4s 51ms/step - loss: 2.2414 - val_loss: 1.8213
Epoch 57/2147483691
79/79 [==============================] - 4s 50ms/step - loss: 2.2405 - val_loss: 1.7987
Epoch 58/2147483691
79/79 [==============================] - 4s 51ms/step - loss: 2.2405 - val_loss: 1.8074
Epoch 59/2147483691
79/79 [==============================] - 4s 51ms/step - loss: 2.2400 - val_loss: 1.8353
Epoch 60/2147483691
79/79 [==============================] - 4s 51ms/step - loss: 2.2405 - val_loss: 1.8140
Epoch 61/2147483691
79/79 [==============================] - 4s 51ms/step - loss: 2.2405 - val_loss: 1.8092
Epoch 62/2147483691
79/79 [==============================] - 4s 51ms/step - loss: 2.2409 - val_loss: 1.8203
Epoch 63/2147483691
79/79 [==============================] - 4s 51ms/step - loss: 2.2387 - val_loss: 1.8120
Epoch 64/2147483691
79/79 [==============================] - 4s 51ms/step - loss: 2.2399 - val_loss: 1.8119
Epoch 65/2147483691
79/79 [==============================] - 4s 51ms/step - loss: 2.2398 - val_loss: 1.8108
Epoch 66/2147483691
79/79 [==============================] - 4s 52ms/step - loss: 2.2398 - val_loss: 1.8127
Epoch 67/2147483691
79/79 [==============================] - 4s 50ms/step - loss: 2.2401 - val_loss: 1.8033
Epoch 68/2147483691
79/79 [==============================] - 4s 51ms/step - loss: 2.2391 - val_loss: 1.8117
Epoch 69/2147483691
79/79 [==============================] - 4s 51ms/step - loss: 2.2395 - val_loss: 1.8116
Epoch 70/2147483691
79/79 [==============================] - 4s 51ms/step - loss: 2.2392 - val_loss: 1.8134
Epoch 71/2147483691
79/79 [==============================] - 4s 51ms/step - loss: 2.2389 - val_loss: 1.8125
Epoch 72/2147483691
79/79 [==============================] - 4s 50ms/step - loss: 2.2400 - val_loss: 1.8129
Epoch 73/2147483691
79/79 [==============================] - 4s 52ms/step - loss: 2.2394 - val_loss: 1.8126
Epoch 74/2147483691
79/79 [==============================] - 4s 51ms/step - loss: 2.2391 - val_loss: 1.8115
Epoch 75/2147483691
79/79 [==============================] - 4s 50ms/step - loss: 2.2393 - val_loss: 1.8116
Epoch 76/2147483691
79/79 [==============================] - 4s 51ms/step - loss: 2.2398 - val_loss: 1.8111
Epoch 77/2147483691
79/79 [==============================] - 4s 51ms/step - loss: 2.2395 - val_loss: 1.8106
Masterful Restore LR: Restoring lr to 0.0012499999720603228.
Masterful backend.fit phase 2 of 2 DONE.
Masterful backend.fit() DONE.
32/32 [==============================] - 0s 3ms/step - loss: 1.2481 - categorical_accuracy: 0.5580
[7]:
Evaluate the model trained with Masterful. Your model should now be about 55% accurate.
[8]:
model.evaluate(test)
32/32 [==============================] - 0s 3ms/step - loss: 1.2481 - categorical_accuracy: 0.5580
[8]:
[1.2481117248535156, 0.5580000281333923]