Keras + Albumentations: Cats vs Dogs Classification 🔗

Complete training pipeline with automatic device detection (CUDA/MPS/CPU).

# %pip install -q tensorflow keras tensorflow-datasets
# %pip install -q albumentationsx
# %pip install -q opencv-python-headless matplotlib

import platform
import numpy as np
import matplotlib.pyplot as plt
from typing import Tuple

# Display plots inline in the notebook
%matplotlib inline

import tensorflow as tf
import tensorflow_datasets as tfds
import keras
from keras import layers, models, optimizers, metrics, callbacks
import albumentations as A

print(f"TensorFlow: &#123;tf.__version__&#125;")
print(f"Keras: &#123;keras.__version__&#125;")
print(f"Albumentations: &#123;A.__version__&#125;")

TensorFlow: 2.19.1 Keras: 3.11.3 Albumentations: 2.0.11

def detect_device() -> Tuple[str, int]:
    cuda_gpus = tf.config.list_physical_devices("GPU")
    if cuda_gpus:
        print(f"✓ CUDA GPU detected: &#123;len(cuda_gpus)&#125; device(s)")
        for gpu in cuda_gpus:
            try:
                tf.config.experimental.set_memory_growth(gpu, True)
            except:
                pass
        return "CUDA_GPU", 64
    if platform.system() == "Darwin" and platform.machine() == "arm64":
        print("✓ Apple Silicon MPS detected")
        return "MPS", 32
    print("✓ CPU detected")
    return "CPU", 16

DEVICE, BATCH_SIZE = detect_device()
print(f"→ Using &#123;DEVICE&#125; with batch size &#123;BATCH_SIZE&#125;")

✓ Apple Silicon MPS detected → Using MPS with batch size 32

class Config:
    input_shape = (128, 128, 3)
    num_classes = 1
    batch_size = BATCH_SIZE
    epochs = 10
    learning_rate = 1e-3
    train_split = "train[:70%]"
    val_split = "train[70%:90%]"
    test_split = "train[90%:]"
    model_path = f"cats_dogs_&#123;DEVICE.lower()&#125;.keras"

config = Config()
print(f"Batch size: &#123;config.batch_size&#125;, Epochs: &#123;config.epochs&#125;")

Batch size: 32, Epochs: 10

Data Augmentation - Using Exact Parameters from Working Script 🔗

def create_augmentation_pipeline(is_training: bool = True) -> A.Compose:
    if is_training:
        return A.Compose([
            A.RandomResizedCrop(size=(128, 128), scale=(0.8, 1.0)),
            A.HorizontalFlip(p=0.5),
            A.RandomBrightnessContrast(brightness_limit=0.2, contrast_limit=0.2, p=0.5),
            A.HueSaturationValue(hue_shift_limit=20, sat_shift_limit=30, val_shift_limit=20, p=0.3),
            A.OneOf([
                A.GaussNoise(std_range=(0.1, 0.2), p=1),
                A.GaussianBlur(blur_limit=(3, 5), p=1),
            ], p=0.2),
            A.CoarseDropout(num_holes_range=(1, 8), hole_height_range=(0.0625, 0.125), hole_width_range=(0.0625, 0.125), p=0.3),
            A.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225]),
        ])
    else:
        return A.Compose([
            A.Resize(height=128, width=128),
            A.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225]),
        ])

train_transform = create_augmentation_pipeline(True)
val_transform = create_augmentation_pipeline(False)
print("✓ Augmentation pipelines created")

✓ Augmentation pipelines created

Load and Prepare Dataset 🔗

# Load dataset
print("Downloading Cats vs Dogs dataset...")
(ds_train, ds_val, ds_test), ds_info = tfds.load(
    'cats_vs_dogs',
    split=[config.train_split, config.val_split, config.test_split],
    with_info=True,
    as_supervised=True,
    shuffle_files=True
)
print(f"✓ Dataset loaded: &#123;ds_info.splits['train'].num_examples&#125; examples")

# Apply augmentations
def augment_image(image, label, transform):
    def apply(img):
        augmented = transform(image=img.numpy())
        return augmented['image'].astype(np.float32)
    
    aug_img = tf.py_function(apply, [image], tf.float32)
    aug_img.set_shape([128, 128, 3])
    return aug_img, label

# Prepare datasets
train_dataset = ds_train.map(
    lambda x, y: augment_image(x, y, train_transform),
    num_parallel_calls=tf.data.AUTOTUNE
).cache().shuffle(1000).batch(config.batch_size).prefetch(tf.data.AUTOTUNE)

val_dataset = ds_val.map(
    lambda x, y: augment_image(x, y, val_transform),
    num_parallel_calls=tf.data.AUTOTUNE
).cache().batch(config.batch_size).prefetch(tf.data.AUTOTUNE)

test_dataset = ds_test.map(
    lambda x, y: augment_image(x, y, val_transform),
    num_parallel_calls=tf.data.AUTOTUNE
).cache().batch(config.batch_size).prefetch(tf.data.AUTOTUNE)

print("✓ Data pipeline ready")

Downloading Cats vs Dogs dataset... ✓ Dataset loaded: 23262 examples ✓ Data pipeline ready

Build and Train Model 🔗

# Create model
model = models.Sequential([
    layers.Input(shape=config.input_shape),
    
    # Block 1
    layers.Conv2D(32, 3, padding='same', activation='relu'),
    layers.BatchNormalization(),
    layers.MaxPooling2D(2),
    layers.Dropout(0.25),
    
    # Block 2
    layers.Conv2D(64, 3, padding='same', activation='relu'),
    layers.BatchNormalization(),
    layers.MaxPooling2D(2),
    layers.Dropout(0.25),
    
    # Block 3
    layers.Conv2D(128, 3, padding='same', activation='relu'),
    layers.BatchNormalization(),
    layers.MaxPooling2D(2),
    layers.Dropout(0.25),
    
    # Block 4
    layers.Conv2D(256, 3, padding='same', activation='relu'),
    layers.BatchNormalization(),
    layers.GlobalAveragePooling2D(),
    
    # Dense layers
    layers.Dense(256, activation='relu'),
    layers.BatchNormalization(),
    layers.Dropout(0.5),
    layers.Dense(128, activation='relu'),
    layers.BatchNormalization(),
    layers.Dropout(0.5),
    
    # Output
    layers.Dense(config.num_classes, activation='sigmoid')
])

# Compile
model.compile(
    optimizer=optimizers.Adam(learning_rate=config.learning_rate),
    loss='binary_crossentropy',
    metrics=['accuracy', metrics.AUC(name='auc')]
)

print(f"✓ Model created with &#123;model.count_params():,&#125; parameters")
model.summary()

✓ Model created with 490,689 parameters

<pre style="white-space:pre;overflow-x:auto;line-height:normal;font-family:Menlo,'DejaVu Sans Mono',consolas,'Courier New',monospace">Model: "sequential" </pre>

<pre style="white-space:pre;overflow-x:auto;line-height:normal;font-family:Menlo,'DejaVu Sans Mono',consolas,'Courier New',monospace">┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓ ┃ Layer (type) ┃ Output Shape ┃ Param # ┃ ┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩ │ conv2d (Conv2D) │ (None, 128, 128, 32) │ 896 │ ├─────────────────────────────────┼────────────────────────┼───────────────┤ │ batch_normalization │ (None, 128, 128, 32) │ 128 │ │ (BatchNormalization) │ │ │ ├─────────────────────────────────┼────────────────────────┼───────────────┤ │ max_pooling2d (MaxPooling2D) │ (None, 64, 64, 32) │ 0 │ ├─────────────────────────────────┼────────────────────────┼───────────────┤ │ dropout (Dropout) │ (None, 64, 64, 32) │ 0 │ ├─────────────────────────────────┼────────────────────────┼───────────────┤ │ conv2d_1 (Conv2D) │ (None, 64, 64, 64) │ 18,496 │ ├─────────────────────────────────┼────────────────────────┼───────────────┤ │ batch_normalization_1 │ (None, 64, 64, 64) │ 256 │ │ (BatchNormalization) │ │ │ ├─────────────────────────────────┼────────────────────────┼───────────────┤ │ max_pooling2d_1 (MaxPooling2D) │ (None, 32, 32, 64) │ 0 │ ├─────────────────────────────────┼────────────────────────┼───────────────┤ │ dropout_1 (Dropout) │ (None, 32, 32, 64) │ 0 │ ├─────────────────────────────────┼────────────────────────┼───────────────┤ │ conv2d_2 (Conv2D) │ (None, 32, 32, 128) │ 73,856 │ ├─────────────────────────────────┼────────────────────────┼───────────────┤ │ batch_normalization_2 │ (None, 32, 32, 128) │ 512 │ │ (BatchNormalization) │ │ │ ├─────────────────────────────────┼────────────────────────┼───────────────┤ │ max_pooling2d_2 (MaxPooling2D) │ (None, 16, 16, 128) │ 0 │ ├─────────────────────────────────┼────────────────────────┼───────────────┤ │ dropout_2 (Dropout) │ (None, 16, 16, 128) │ 0 │ ├─────────────────────────────────┼────────────────────────┼───────────────┤ │ conv2d_3 (Conv2D) │ (None, 16, 16, 256) │ 295,168 │ ├─────────────────────────────────┼────────────────────────┼───────────────┤ │ batch_normalization_3 │ (None, 16, 16, 256) │ 1,024 │ │ (BatchNormalization) │ │ │ ├─────────────────────────────────┼────────────────────────┼───────────────┤ │ global_average_pooling2d │ (None, 256) │ 0 │ │ (GlobalAveragePooling2D) │ │ │ ├─────────────────────────────────┼────────────────────────┼───────────────┤ │ dense (Dense) │ (None, 256) │ 65,792 │ ├─────────────────────────────────┼────────────────────────┼───────────────┤ │ batch_normalization_4 │ (None, 256) │ 1,024 │ │ (BatchNormalization) │ │ │ ├─────────────────────────────────┼────────────────────────┼───────────────┤ │ dropout_3 (Dropout) │ (None, 256) │ 0 │ ├─────────────────────────────────┼────────────────────────┼───────────────┤ │ dense_1 (Dense) │ (None, 128) │ 32,896 │ ├─────────────────────────────────┼────────────────────────┼───────────────┤ │ batch_normalization_5 │ (None, 128) │ 512 │ │ (BatchNormalization) │ │ │ ├─────────────────────────────────┼────────────────────────┼───────────────┤ │ dropout_4 (Dropout) │ (None, 128) │ 0 │ ├─────────────────────────────────┼────────────────────────┼───────────────┤ │ dense_2 (Dense) │ (None, 1) │ 129 │ └─────────────────────────────────┴────────────────────────┴───────────────┘ </pre>

<pre style="white-space:pre;overflow-x:auto;line-height:normal;font-family:Menlo,'DejaVu Sans Mono',consolas,'Courier New',monospace"> Total params: 490,689 (1.87 MB) </pre>

<pre style="white-space:pre;overflow-x:auto;line-height:normal;font-family:Menlo,'DejaVu Sans Mono',consolas,'Courier New',monospace"> Trainable params: 488,961 (1.87 MB) </pre>

<pre style="white-space:pre;overflow-x:auto;line-height:normal;font-family:Menlo,'DejaVu Sans Mono',consolas,'Courier New',monospace"> Non-trainable params: 1,728 (6.75 KB) </pre>

# Train with callbacks
callbacks_list = [
    callbacks.ModelCheckpoint(
        'best_model.keras',
        monitor='val_auc',
        mode='max',
        save_best_only=True,
        verbose=1
    ),
    callbacks.EarlyStopping(
        monitor='val_loss',
        patience=5,
        restore_best_weights=True,
        verbose=1
    ),
    callbacks.ReduceLROnPlateau(
        monitor='val_loss',
        factor=0.5,
        patience=3,
        min_lr=1e-7,
        verbose=1
    )
]

print(f"Training on &#123;DEVICE&#125; for &#123;config.epochs&#125; epochs...")

history = model.fit(
    train_dataset,
    validation_data=val_dataset,
    epochs=config.epochs,
    callbacks=callbacks_list,
    verbose=1
)

print("✓ Training completed!")

Training on MPS for 10 epochs... Epoch 1/10

2025-09-17 18:53:16.505703: I tensorflow/core/kernels/data/tf_record_dataset_op.cc:387] The default buffer size is 262144, which is overridden by the user specified buffer_size of 8388608

[1m509/509[0m [32m━━━━━━━━━━━━━━━━━━━━[0m[37m[0m [1m0s[0m 122ms/step - accuracy: 0.5301 - auc: 0.5423 - loss: 0.8683 Epoch 1: val_auc improved from None to 0.63414, saving model to best_model.keras [1m509/509[0m [32m━━━━━━━━━━━━━━━━━━━━[0m[37m[0m [1m68s[0m 131ms/step - accuracy: 0.5438 - auc: 0.5613 - loss: 0.7959 - val_accuracy: 0.5682 - val_auc: 0.6341 - val_loss: 0.6740 - learning_rate: 0.0010 Epoch 2/10 [1m509/509[0m [32m━━━━━━━━━━━━━━━━━━━━[0m[37m[0m [1m0s[0m 119ms/step - accuracy: 0.5700 - auc: 0.5999 - loss: 0.6970 Epoch 2: val_auc improved from 0.63414 to 0.73070, saving model to best_model.keras [1m509/509[0m [32m━━━━━━━━━━━━━━━━━━━━[0m[37m[0m [1m65s[0m 127ms/step - accuracy: 0.5885 - auc: 0.6244 - loss: 0.6781 - val_accuracy: 0.6682 - val_auc: 0.7307 - val_loss: 0.6142 - learning_rate: 0.0010 Epoch 3/10 [1m509/509[0m [32m━━━━━━━━━━━━━━━━━━━━[0m[37m[0m [1m0s[0m 127ms/step - accuracy: 0.6304 - auc: 0.6839 - loss: 0.6387 Epoch 3: val_auc did not improve from 0.73070 [1m509/509[0m [32m━━━━━━━━━━━━━━━━━━━━[0m[37m[0m [1m69s[0m 135ms/step - accuracy: 0.6461 - auc: 0.7005 - loss: 0.6275 - val_accuracy: 0.6336 - val_auc: 0.7236 - val_loss: 0.6419 - learning_rate: 0.0010 Epoch 4/10 [1m509/509[0m [32m━━━━━━━━━━━━━━━━━━━━[0m[37m[0m [1m0s[0m 134ms/step - accuracy: 0.6720 - auc: 0.7341 - loss: 0.6019 Epoch 4: val_auc improved from 0.73070 to 0.77413, saving model to best_model.keras [1m509/509[0m [32m━━━━━━━━━━━━━━━━━━━━[0m[37m[0m [1m73s[0m 143ms/step - accuracy: 0.6768 - auc: 0.7429 - loss: 0.5959 - val_accuracy: 0.6884 - val_auc: 0.7741 - val_loss: 0.6048 - learning_rate: 0.0010 Epoch 5/10 [1m509/509[0m [32m━━━━━━━━━━━━━━━━━━━━[0m[37m[0m [1m0s[0m 131ms/step - accuracy: 0.6963 - auc: 0.7632 - loss: 0.5792 Epoch 5: val_auc improved from 0.77413 to 0.78378, saving model to best_model.keras [1m509/509[0m [32m━━━━━━━━━━━━━━━━━━━━[0m[37m[0m [1m72s[0m 141ms/step - accuracy: 0.6987 - auc: 0.7683 - loss: 0.5736 - val_accuracy: 0.6546 - val_auc: 0.7838 - val_loss: 0.6826 - learning_rate: 0.0010 Epoch 6/10 [1m509/509[0m [32m━━━━━━━━━━━━━━━━━━━━[0m[37m[0m [1m0s[0m 131ms/step - accuracy: 0.7208 - auc: 0.7952 - loss: 0.5488 Epoch 6: val_auc improved from 0.78378 to 0.84497, saving model to best_model.keras [1m509/509[0m [32m━━━━━━━━━━━━━━━━━━━━[0m[37m[0m [1m71s[0m 140ms/step - accuracy: 0.7255 - auc: 0.8010 - loss: 0.5413 - val_accuracy: 0.7148 - val_auc: 0.8450 - val_loss: 0.5432 - learning_rate: 0.0010 Epoch 7/10 [1m509/509[0m [32m━━━━━━━━━━━━━━━━━━━━[0m[37m[0m [1m0s[0m 134ms/step - accuracy: 0.7365 - auc: 0.8157 - loss: 0.5246 Epoch 7: val_auc improved from 0.84497 to 0.85142, saving model to best_model.keras [1m509/509[0m [32m━━━━━━━━━━━━━━━━━━━━[0m[37m[0m [1m72s[0m 142ms/step - accuracy: 0.7457 - auc: 0.8273 - loss: 0.5108 - val_accuracy: 0.7501 - val_auc: 0.8514 - val_loss: 0.5091 - learning_rate: 0.0010 Epoch 8/10 [1m509/509[0m [32m━━━━━━━━━━━━━━━━━━━━[0m[37m[0m [1m0s[0m 129ms/step - accuracy: 0.7755 - auc: 0.8554 - loss: 0.4731 Epoch 8: val_auc improved from 0.85142 to 0.89584, saving model to best_model.keras [1m509/509[0m [32m━━━━━━━━━━━━━━━━━━━━[0m[37m[0m [1m70s[0m 138ms/step - accuracy: 0.7790 - auc: 0.8621 - loss: 0.4629 - val_accuracy: 0.8023 - val_auc: 0.8958 - val_loss: 0.4355 - learning_rate: 0.0010 Epoch 9/10 [1m509/509[0m [32m━━━━━━━━━━━━━━━━━━━━[0m[37m[0m [1m0s[0m 127ms/step - accuracy: 0.8053 - auc: 0.8870 - loss: 0.4234 Epoch 9: val_auc improved from 0.89584 to 0.92848, saving model to best_model.keras [1m509/509[0m [32m━━━━━━━━━━━━━━━━━━━━[0m[37m[0m [1m69s[0m 135ms/step - accuracy: 0.8101 - auc: 0.8923 - loss: 0.4136 - val_accuracy: 0.8461 - val_auc: 0.9285 - val_loss: 0.3472 - learning_rate: 0.0010 Epoch 10/10 [1m509/509[0m [32m━━━━━━━━━━━━━━━━━━━━[0m[37m[0m [1m0s[0m 126ms/step - accuracy: 0.8309 - auc: 0.9112 - loss: 0.3786 Epoch 10: val_auc did not improve from 0.92848 [1m509/509[0m [32m━━━━━━━━━━━━━━━━━━━━[0m[37m[0m [1m69s[0m 135ms/step - accuracy: 0.8350 - auc: 0.9143 - loss: 0.3717 - val_accuracy: 0.8199 - val_auc: 0.9255 - val_loss: 0.3945 - learning_rate: 0.0010 Restoring model weights from the end of the best epoch: 9. ✓ Training completed!

Evaluate and Save Model 🔗

# Evaluate on test set
test_results = model.evaluate(test_dataset, verbose=1, return_dict=True)

print("\n✓ Test Results:")
for metric, value in test_results.items():
    print(f"  &#123;metric&#125;: &#123;value:.4f&#125;")

# Save model
model.save(config.model_path)
print(f"\n✓ Model saved as &#123;config.model_path&#125;")

# Convert to TFLite
try:
    converter = tf.lite.TFLiteConverter.from_keras_model(model)
    converter.optimizations = [tf.lite.Optimize.DEFAULT]
    tflite_model = converter.convert()
    
    tflite_path = config.model_path.replace('.keras', '.tflite')
    with open(tflite_path, 'wb') as f:
        f.write(tflite_model)
    print(f"✓ TFLite model saved as &#123;tflite_path&#125; (&#123;len(tflite_model)/1024:.2f&#125; KB)")
except Exception as e:
    print(f"TFLite conversion failed: &#123;e&#125;")

[1m73/73[0m [32m━━━━━━━━━━━━━━━━━━━━[0m[37m[0m [1m2s[0m 33ms/step - accuracy: 0.8547 - auc: 0.9369 - loss: 0.3232

✓ Test Results: accuracy: 0.8547 auc: 0.9369 loss: 0.3232

✓ Model saved as cats_dogs_mps.keras INFO:tensorflow:Assets written to: /var/folders/68/k137nch11m76w1plfrw320r00000gn/T/tmp0otru8h6/assets

INFO:tensorflow:Assets written to: /var/folders/68/k137nch11m76w1plfrw320r00000gn/T/tmp0otru8h6/assets

Saved artifact at '/var/folders/68/k137nch11m76w1plfrw320r00000gn/T/tmp0otru8h6'. The following endpoints are available:

Endpoint 'serve' args_0 (POSITIONAL_ONLY): TensorSpec(shape=(None, 128, 128, 3), dtype=tf.float32, name='keras_tensor') Output Type: TensorSpec(shape=(None, 1), dtype=tf.float32, name=None) Captures: 4705823808: TensorSpec(shape=(), dtype=tf.resource, name=None) 5583028288: TensorSpec(shape=(), dtype=tf.resource, name=None) 5583034096: TensorSpec(shape=(), dtype=tf.resource, name=None) 5583043776: TensorSpec(shape=(), dtype=tf.resource, name=None) 5583030224: TensorSpec(shape=(), dtype=tf.resource, name=None) 5583033744: TensorSpec(shape=(), dtype=tf.resource, name=None) 5583037440: TensorSpec(shape=(), dtype=tf.resource, name=None) 5583034448: TensorSpec(shape=(), dtype=tf.resource, name=None) 5583040256: TensorSpec(shape=(), dtype=tf.resource, name=None) 5583038848: TensorSpec(shape=(), dtype=tf.resource, name=None) 5583029168: TensorSpec(shape=(), dtype=tf.resource, name=None) 5583035680: TensorSpec(shape=(), dtype=tf.resource, name=None) 5583042016: TensorSpec(shape=(), dtype=tf.resource, name=None) 5583042896: TensorSpec(shape=(), dtype=tf.resource, name=None) 5583277568: TensorSpec(shape=(), dtype=tf.resource, name=None) 5583275280: TensorSpec(shape=(), dtype=tf.resource, name=None) 5583276688: TensorSpec(shape=(), dtype=tf.resource, name=None) 5583275632: TensorSpec(shape=(), dtype=tf.resource, name=None) 5583286544: TensorSpec(shape=(), dtype=tf.resource, name=None) 5583277392: TensorSpec(shape=(), dtype=tf.resource, name=None) 5583284256: TensorSpec(shape=(), dtype=tf.resource, name=None) 5583286368: TensorSpec(shape=(), dtype=tf.resource, name=None) 5583283552: TensorSpec(shape=(), dtype=tf.resource, name=None) 5583286896: TensorSpec(shape=(), dtype=tf.resource, name=None) 5583289008: TensorSpec(shape=(), dtype=tf.resource, name=None) 5583289712: TensorSpec(shape=(), dtype=tf.resource, name=None) 5583275984: TensorSpec(shape=(), dtype=tf.resource, name=None) 5583339760: TensorSpec(shape=(), dtype=tf.resource, name=None) 5583289536: TensorSpec(shape=(), dtype=tf.resource, name=None) 5583340464: TensorSpec(shape=(), dtype=tf.resource, name=None) 5583342224: TensorSpec(shape=(), dtype=tf.resource, name=None) 5583342928: TensorSpec(shape=(), dtype=tf.resource, name=None) 5583344336: TensorSpec(shape=(), dtype=tf.resource, name=None) 5583345744: TensorSpec(shape=(), dtype=tf.resource, name=None) 5583340288: TensorSpec(shape=(), dtype=tf.resource, name=None) 5583344688: TensorSpec(shape=(), dtype=tf.resource, name=None) 5583346800: TensorSpec(shape=(), dtype=tf.resource, name=None) 5583347504: TensorSpec(shape=(), dtype=tf.resource, name=None) ✓ TFLite model saved as cats_dogs_mps.tflite (501.87 KB)

WARNING: All log messages before absl::InitializeLog() is called are written to STDERR W0000 00:00:1758128695.974261 105836395 tf_tfl_flatbuffer_helpers.cc:365] Ignored output_format. W0000 00:00:1758128695.974270 105836395 tf_tfl_flatbuffer_helpers.cc:368] Ignored drop_control_dependency. 2025-09-17 19:04:55.974428: I tensorflow/cc/saved_model/reader.cc:83] Reading SavedModel from: /var/folders/68/k137nch11m76w1plfrw320r00000gn/T/tmp0otru8h6 2025-09-17 19:04:55.975394: I tensorflow/cc/saved_model/reader.cc:52] Reading meta graph with tags { serve } 2025-09-17 19:04:55.975398: I tensorflow/cc/saved_model/reader.cc:147] Reading SavedModel debug info (if present) from: /var/folders/68/k137nch11m76w1plfrw320r00000gn/T/tmp0otru8h6 I0000 00:00:1758128695.983613 105836395 mlir_graph_optimization_pass.cc:425] MLIR V1 optimization pass is not enabled 2025-09-17 19:04:55.985110: I tensorflow/cc/saved_model/loader.cc:236] Restoring SavedModel bundle. 2025-09-17 19:04:56.043013: I tensorflow/cc/saved_model/loader.cc:220] Running initialization op on SavedModel bundle at path: /var/folders/68/k137nch11m76w1plfrw320r00000gn/T/tmp0otru8h6 2025-09-17 19:04:56.059464: I tensorflow/cc/saved_model/loader.cc:471] SavedModel load for tags { serve }; Status: success: OK. Took 85037 microseconds. 2025-09-17 19:04:56.077965: I tensorflow/compiler/mlir/tensorflow/utils/dump_mlir_util.cc:269] disabling MLIR crash reproducer, set env var MLIR_CRASH_REPRODUCER_DIRECTORY to enable.

Visualize Results 🔗

# Plot training history
fig, axes = plt.subplots(1, 2, figsize=(12, 4))

# Loss
axes[0].plot(history.history['loss'], label='Train')
axes[0].plot(history.history['val_loss'], label='Validation')
axes[0].set_title('Model Loss')
axes[0].set_xlabel('Epoch')
axes[0].set_ylabel('Loss')
axes[0].legend()
axes[0].grid(True, alpha=0.3)

# Accuracy
axes[1].plot(history.history['accuracy'], label='Train')
axes[1].plot(history.history['val_accuracy'], label='Validation')
axes[1].set_title('Model Accuracy')
axes[1].set_xlabel('Epoch')
axes[1].set_ylabel('Accuracy')
axes[1].legend()
axes[1].grid(True, alpha=0.3)

plt.tight_layout()
plt.show()

print(f"\nFinal Results:")
print(f"  Test Accuracy: &#123;test_results['accuracy']:.3f&#125;")
print(f"  Test AUC: &#123;test_results['auc']:.3f&#125;")
print(f"  Device Used: &#123;DEVICE&#125;")
print(f"  Total Epochs: &#123;len(history.history['loss'])&#125;")
print("\n✓ Training complete! Model ready for deployment.")

Final Results: Test Accuracy: 0.855 Test AUC: 0.937 Device Used: MPS Total Epochs: 10

✓ Training complete! Model ready for deployment.