[Python] Tensorflow dataset crashing VsCode

So I'm trying to make a simple image classifier using a CNN and an already labeled data set called "places 365". However, whenever I run model.fit() my RAM usage maxes out and VS Code crashes or I restart the kernel.

In the past, I have gotten model.fit() to work just fine when I imported the cifar10 dataset for a previous model with tensorflow.keras.datasets.cifar10.load_data() so I am pretty sure I am missing something with how tfds.load() formats the dataset.

Here is my entire setup:

import tensorflow as tf
import tensorflow_datasets as tfds
from tensorflow.keras import datasets, layers, models
import matplotlib.pyplot as plt
import numpy as np

(ds_train, ds_test), ds_info = tfds.load(
'places365_small',
split=['train', 'test'],
shuffle_files=True,
as_supervised=True,
with_info=True,
)

def normalize_img(image, label , filename = ''):
"""Normalizes images: `uint8` -> `float32`."""
return tf.cast(image, tf.float32) / 255., label

ds_train = ds_train.map(
normalize_img, num_parallel_calls=tf.data.AUTOTUNE)
ds_train = ds_train.cache()
ds_train = ds_train.shuffle(ds_info.splits['train'].num_examples)
ds_train = ds_train.batch(128)
ds_train = ds_train.prefetch(tf.data.AUTOTUNE)

ds_test = ds_test.map(
normalize_img, num_parallel_calls=tf.data.AUTOTUNE)
ds_test = ds_test.batch(128)
ds_test = ds_test.cache()
ds_test = ds_test.prefetch(tf.data.AUTOTUNE)

model = tf.keras.models.Sequential()
model.add(tf.keras.layers.Conv2D(32, (3, 3), activation='relu', input_shape=(256, 256, 3)))
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.Conv2D(64, (3, 3), activation='relu'))

model.add(tf.keras.layers.Flatten())
model.add(tf.keras.layers.Dense(64, activation='relu'))
model.add(tf.keras.layers.Dense(365))

model.compile(
optimizer=tf.keras.optimizers.Adam(0.001),
loss=tf.keras.losses.SparseCategoricalCrossentropy(from_logits=True),
metrics=[tf.keras.metrics.SparseCategoricalAccuracy()],
)

model.fit(
ds_train,
epochs=6,
validation_data=ds_test,
)


I essentially followed the instructions here: Official TensorFlow documentation

All I really changed was:

• 
  loaded places365_small dataset instead of MNIST
  
  
• 
  added the filename arg to normalize_img since the dataset has 3 features (image, label, filename)
  
  
• 
  implemented CNN model
  
  • 
    changed input shape to match image dimensions for place365
    
    
  • 
    final dense layer changed to 365 (365 categories)
    

Everything else is the same, and all code blocks compile and run in Jupyter except for model.fit() so I have no idea why this stupid function is just gobbling my RAM and crashing VS Code.

Continue reading...