Transfer learning and fine-tuning

View on TensorFlow.org Run in Google Colab View source on GitHub Download notebook

In this tutorial, you will learn how to classify images of cats and dogs by using transfer learning from a pre-trained network.

A pre-trained model is a saved network that was previously trained on a large dataset, typically on a large-scale image-classification task. You either use the pretrained model as is or use transfer learning to customize this model to a given task.

The intuition behind transfer learning for image classification is that if a model is trained on a large and general enough dataset, this model will effectively serve as a generic model of the visual world. You can then take advantage of these learned feature maps without having to start from scratch by training a large model on a large dataset.

In this notebook, you will try two ways to customize a pretrained model:

  1. Feature Extraction: Use the representations learned by a previous network to extract meaningful features from new samples. You simply add a new classifier, which will be trained from scratch, on top of the pretrained model so that you can repurpose the feature maps learned previously for the dataset.

    You do not need to (re)train the entire model. The base convolutional network already contains features that are generically useful for classifying pictures. However, the final, classification part of the pretrained model is specific to the original classification task, and subsequently specific to the set of classes on which the model was trained.

  2. Fine-Tuning: Unfreeze a few of the top layers of a frozen model base and jointly train both the newly-added classifier layers and the last layers of the base model. This allows us to "fine-tune" the higher-order feature representations in the base model in order to make them more relevant for the specific task.

You will follow the general machine learning workflow.

  1. Examine and understand the data
  2. Build an input pipeline, in this case using Keras ImageDataGenerator
  3. Compose the model
    • Load in the pretrained base model (and pretrained weights)
    • Stack the classification layers on top
  4. Train the model
  5. Evaluate model
import matplotlib.pyplot as plt
import numpy as np
import os
import tensorflow as tf
2024-07-19 05:09:44.190884: E external/local_xla/xla/stream_executor/cuda/cuda_fft.cc:485] Unable to register cuFFT factory: Attempting to register factory for plugin cuFFT when one has already been registered
2024-07-19 05:09:44.211659: E external/local_xla/xla/stream_executor/cuda/cuda_dnn.cc:8454] Unable to register cuDNN factory: Attempting to register factory for plugin cuDNN when one has already been registered
2024-07-19 05:09:44.218008: E external/local_xla/xla/stream_executor/cuda/cuda_blas.cc:1452] Unable to register cuBLAS factory: Attempting to register factory for plugin cuBLAS when one has already been registered

Data preprocessing

Data download

In this tutorial, you will use a dataset containing several thousand images of cats and dogs. Download and extract a zip file containing the images, then create a tf.data.Dataset for training and validation using the tf.keras.utils.image_dataset_from_directory utility. You can learn more about loading images in this tutorial.

_URL = 'https://storage.googleapis.com/mledu-datasets/cats_and_dogs_filtered.zip'
path_to_zip = tf.keras.utils.get_file('cats_and_dogs.zip', origin=_URL, extract=True)
PATH = os.path.join(os.path.dirname(path_to_zip), 'cats_and_dogs_filtered')

train_dir = os.path.join(PATH, 'train')
validation_dir = os.path.join(PATH, 'validation')

BATCH_SIZE = 32
IMG_SIZE = (160, 160)

train_dataset = tf.keras.utils.image_dataset_from_directory(train_dir,
                                                            shuffle=True,
                                                            batch_size=BATCH_SIZE,
                                                            image_size=IMG_SIZE)
Downloading data from https://storage.googleapis.com/mledu-datasets/cats_and_dogs_filtered.zip
68606236/68606236 ━━━━━━━━━━━━━━━━━━━━ 0s 0us/step
Found 2000 files belonging to 2 classes.
WARNING: All log messages before absl::InitializeLog() is called are written to STDERR
I0000 00:00:1721365787.906248   77561 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355
I0000 00:00:1721365787.910125   77561 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355
I0000 00:00:1721365787.913670   77561 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355
I0000 00:00:1721365787.917204   77561 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355
I0000 00:00:1721365787.928345   77561 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355
I0000 00:00:1721365787.931794   77561 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355
I0000 00:00:1721365787.935107   77561 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355
I0000 00:00:1721365787.938371   77561 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355
I0000 00:00:1721365787.941804   77561 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355
I0000 00:00:1721365787.945255   77561 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355
I0000 00:00:1721365787.948518   77561 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355
I0000 00:00:1721365787.951787   77561 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355
I0000 00:00:1721365789.208759   77561 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355
I0000 00:00:1721365789.211011   77561 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355
I0000 00:00:1721365789.212985   77561 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355
I0000 00:00:1721365789.214995   77561 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355
I0000 00:00:1721365789.217223   77561 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355
I0000 00:00:1721365789.219298   77561 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355
I0000 00:00:1721365789.221151   77561 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355
I0000 00:00:1721365789.223049   77561 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355
I0000 00:00:1721365789.225201   77561 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355
I0000 00:00:1721365789.227274   77561 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355
I0000 00:00:1721365789.229143   77561 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355
I0000 00:00:1721365789.231052   77561 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355
I0000 00:00:1721365789.270034   77561 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355
I0000 00:00:1721365789.272703   77561 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355
I0000 00:00:1721365789.274620   77561 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355
I0000 00:00:1721365789.276600   77561 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355
I0000 00:00:1721365789.279130   77561 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355
I0000 00:00:1721365789.281240   77561 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355
I0000 00:00:1721365789.283134   77561 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355
I0000 00:00:1721365789.285078   77561 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355
I0000 00:00:1721365789.287052   77561 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355
I0000 00:00:1721365789.289496   77561 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355
I0000 00:00:1721365789.291822   77561 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355
I0000 00:00:1721365789.294147   77561 cuda_executor.cc:1015] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355
validation_dataset = tf.keras.utils.image_dataset_from_directory(validation_dir,
                                                                 shuffle=True,
                                                                 batch_size=BATCH_SIZE,
                                                                 image_size=IMG_SIZE)
Found 1000 files belonging to 2 classes.

Show the first nine images and labels from the training set:

class_names = train_dataset.class_names

plt.figure(figsize=(10, 10))
for images, labels in train_dataset.take(1):
  for i in range(9):
    ax = plt.subplot(3, 3, i + 1)
    plt.imshow(images[i].numpy().astype("uint8"))
    plt.title(class_names[labels[i]])
    plt.axis("off")

png

As the original dataset doesn't contain a test set, you will create one. To do so, determine how many batches of data are available in the validation set using tf.data.experimental.cardinality, then move 20% of them to a test set.

val_batches = tf.data.experimental.cardinality(validation_dataset)
test_dataset = validation_dataset.take(val_batches // 5)
validation_dataset = validation_dataset.skip(val_batches // 5)
print('Number of validation batches: %d' % tf.data.experimental.cardinality(validation_dataset))
print('Number of test batches: %d' % tf.data.experimental.cardinality(test_dataset))
Number of validation batches: 26
Number of test batches: 6

Configure the dataset for performance

Use buffered prefetching to load images from disk without having I/O become blocking. To learn more about this method see the data performance guide.

AUTOTUNE = tf.data.AUTOTUNE

train_dataset = train_dataset.prefetch(buffer_size=AUTOTUNE)
validation_dataset = validation_dataset.prefetch(buffer_size=AUTOTUNE)
test_dataset = test_dataset.prefetch(buffer_size=AUTOTUNE)

Use data augmentation

When you don't have a large image dataset, it's a good practice to artificially introduce sample diversity by applying random, yet realistic, transformations to the training images, such as rotation and horizontal flipping. This helps expose the model to different aspects of the training data and reduce overfitting. You can learn more about data augmentation in this tutorial.

data_augmentation = tf.keras.Sequential([
  tf.keras.layers.RandomFlip('horizontal'),
  tf.keras.layers.RandomRotation(0.2),
])

Let's repeatedly apply these layers to the same image and see the result.

for image, _ in train_dataset.take(1):
  plt.figure(figsize=(10, 10))
  first_image = image[0]
  for i in range(9):
    ax = plt.subplot(3, 3, i + 1)
    augmented_image = data_augmentation(tf.expand_dims(first_image, 0))
    plt.imshow(augmented_image[0] / 255)
    plt.axis('off')

png

Rescale pixel values

In a moment, you will download tf.keras.applications.MobileNetV2 for use as your base model. This model expects pixel values in [-1, 1], but at this point, the pixel values in your images are in [0, 255]. To rescale them, use the preprocessing method included with the model.

preprocess_input = tf.keras.applications.mobilenet_v2.preprocess_input
rescale = tf.keras.layers.Rescaling(1./127.5, offset=-1)

Create the base model from the pre-trained convnets

You will create the base model from the MobileNet V2 model developed at Google. This is pre-trained on the ImageNet dataset, a large dataset consisting of 1.4M images and 1000 classes. ImageNet is a research training dataset with a wide variety of categories like jackfruit and syringe. This base of knowledge will help us classify cats and dogs from our specific dataset.

First, you need to pick which layer of MobileNet V2 you will use for feature extraction. The very last classification layer (on "top", as most diagrams of machine learning models go from bottom to top) is not very useful. Instead, you will follow the common practice to depend on the very last layer before the flatten operation. This layer is called the "bottleneck layer". The bottleneck layer features retain more generality as compared to the final/top layer.

First, instantiate a MobileNet V2 model pre-loaded with weights trained on ImageNet. By specifying the include_top=False argument, you load a network that doesn't include the classification layers at the top, which is ideal for feature extraction.

# Create the base model from the pre-trained model MobileNet V2
IMG_SHAPE = IMG_SIZE + (3,)
base_model = tf.keras.applications.MobileNetV2(input_shape=IMG_SHAPE,
                                               include_top=False,
                                               weights='imagenet')
Downloading data from https://storage.googleapis.com/tensorflow/keras-applications/mobilenet_v2/mobilenet_v2_weights_tf_dim_ordering_tf_kernels_1.0_160_no_top.h5
9406464/9406464 ━━━━━━━━━━━━━━━━━━━━ 0s 0us/step

This feature extractor converts each 160x160x3 image into a 5x5x1280 block of features. Let's see what it does to an example batch of images:

image_batch, label_batch = next(iter(train_dataset))
feature_batch = base_model(image_batch)
print(feature_batch.shape)
W0000 00:00:1721365794.301174   77561 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365794.356931   77561 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365794.358316   77561 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365794.359739   77561 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365794.361139   77561 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365794.370214   77561 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365794.371827   77561 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365794.373454   77561 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365794.375161   77561 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365794.378969   77561 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365794.382734   77561 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365794.384618   77561 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365794.386676   77561 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365794.388881   77561 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365794.409993   77561 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
(32, 5, 5, 1280)

Feature extraction

In this step, you will freeze the convolutional base created from the previous step and to use as a feature extractor. Additionally, you add a classifier on top of it and train the top-level classifier.

Freeze the convolutional base

It is important to freeze the convolutional base before you compile and train the model. Freezing (by setting layer.trainable = False) prevents the weights in a given layer from being updated during training. MobileNet V2 has many layers, so setting the entire model's trainable flag to False will freeze all of them.

base_model.trainable = False

Important note about BatchNormalization layers

Many models contain tf.keras.layers.BatchNormalization layers. This layer is a special case and precautions should be taken in the context of fine-tuning, as shown later in this tutorial.

When you set layer.trainable = False, the BatchNormalization layer will run in inference mode, and will not update its mean and variance statistics.

When you unfreeze a model that contains BatchNormalization layers in order to do fine-tuning, you should keep the BatchNormalization layers in inference mode by passing training = False when calling the base model. Otherwise, the updates applied to the non-trainable weights will destroy what the model has learned.

For more details, see the Transfer learning guide.

# Let's take a look at the base model architecture
base_model.summary()

Add a classification head

To generate predictions from the block of features, average over the spatial 5x5 spatial locations, using a tf.keras.layers.GlobalAveragePooling2D layer to convert the features to a single 1280-element vector per image.

global_average_layer = tf.keras.layers.GlobalAveragePooling2D()
feature_batch_average = global_average_layer(feature_batch)
print(feature_batch_average.shape)
(32, 1280)

Apply a tf.keras.layers.Dense layer to convert these features into a single prediction per image. You don't need an activation function here because this prediction will be treated as a logit, or a raw prediction value. Positive numbers predict class 1, negative numbers predict class 0.

prediction_layer = tf.keras.layers.Dense(1, activation='sigmoid')
prediction_batch = prediction_layer(feature_batch_average)
print(prediction_batch.shape)
(32, 1)

Build a model by chaining together the data augmentation, rescaling, base_model and feature extractor layers using the Keras Functional API. As previously mentioned, use training=False as our model contains a BatchNormalization layer.

inputs = tf.keras.Input(shape=(160, 160, 3))
x = data_augmentation(inputs)
x = preprocess_input(x)
x = base_model(x, training=False)
x = global_average_layer(x)
x = tf.keras.layers.Dropout(0.2)(x)
outputs = prediction_layer(x)
model = tf.keras.Model(inputs, outputs)
model.summary()

The 8+ million parameters in MobileNet are frozen, but there are 1.2 thousand trainable parameters in the Dense layer. These are divided between two tf.Variable objects, the weights and biases.

len(model.trainable_variables)
2
tf.keras.utils.plot_model(model, show_shapes=True)

png

Compile the model

Compile the model before training it. Since there are two classes and a sigmoid oputput, use the BinaryAccuracy.

base_learning_rate = 0.0001
model.compile(optimizer=tf.keras.optimizers.Adam(learning_rate=base_learning_rate),
              loss=tf.keras.losses.BinaryCrossentropy(),
              metrics=[tf.keras.metrics.BinaryAccuracy(threshold=0.5, name='accuracy')])

Train the model

After training for 10 epochs, you should see ~96% accuracy on the validation set.

initial_epochs = 10

loss0, accuracy0 = model.evaluate(validation_dataset)
W0000 00:00:1721365796.587523   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.589095   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.590474   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.591882   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.593363   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.595133   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.596800   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.598298   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.599858   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.601381   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.602897   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.604669   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.622305   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.624052   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.626068   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.631918   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.634038   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.645601   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.656249   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.658035   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.659771   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.661537   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.663293   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.665094   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.666876   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.668635   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.670393   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.672768   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.675294   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.679165   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.681457   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.683653   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.685923   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.688269   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.697596   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.700666   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.701952   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.703240   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.704518   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.705810   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.707113   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.708428   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.709796   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.711170   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.712601   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.713995   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.715573   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.717192   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.718821   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.720635   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.722697   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.726909   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.736581   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.738018   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.739412   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.740848   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.742442   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.743961   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.745381   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.746828   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.748284   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.749953   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.751658   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.753898   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.755615   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.757188   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.758845   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.760468   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.768452   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.769843   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.771206   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.772549   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.773891   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.775263   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.776640   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.778025   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.779500   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.780976   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.782511   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.784035   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.785808   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.787638   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.789503   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.791804   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.794298   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.798522   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.811157   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.812014   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.812875   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.813716   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.814549   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.815394   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.816239   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.817102   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.817940   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.818789   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.819637   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.820503   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.821424   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.822321   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.823268   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.824180   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.825226   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.826811   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.833946   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.834813   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.835648   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.836568   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.837484   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.838499   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.839356   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.840215   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.841137   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.841990   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.842844   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.843780   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.844615   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.845552   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.847315   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.848178   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.849276   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.855362   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.856234   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.857093   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.857928   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.858775   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.859632   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.860466   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.861301   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.862144   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.862985   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.863836   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.864706   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.865657   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.866587   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.867606   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.868556   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.869682   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.871522   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.878558   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.879406   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.880220   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.881047   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.881859   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.882673   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.883489   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.884300   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.885151   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.885989   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.886823   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.887677   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.888489   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.889299   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.890127   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.891053   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.897642   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.898505   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.899330   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.900199   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.901026   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.901862   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.902780   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.903621   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.904471   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.905348   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.906160   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.906970   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.907856   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.908707   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.909584   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.910488   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.917380   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.918274   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.919131   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.919974   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.920859   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.921701   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.922584   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.923441   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.924294   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.925162   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.925999   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.926837   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.927743   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.928620   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.929493   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.930504   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.938109   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.939006   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.939883   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.940745   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.941599   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.942449   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.943302   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.944186   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.945052   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.945921   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.946798   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.947667   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.948539   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.949414   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.950322   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.951362   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.958873   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.959782   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.960698   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.961594   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.962594   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.963454   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.964316   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.965210   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.966106   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.966971   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.967872   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.968816   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.969774   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.970709   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.971654   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.972652   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.980835   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.981776   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.982677   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.983578   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.984471   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.985375   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.986281   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.987246   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.988157   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.989085   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.990009   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.990923   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.991950   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.992986   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.994063   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365796.995174   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365797.003647   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365797.004517   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365797.005364   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365797.006200   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365797.007104   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365797.008005   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365797.008877   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365797.009671   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365797.010403   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365797.011163   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365797.011917   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365797.012687   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365797.013465   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365797.014320   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365797.015079   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365797.015854   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365797.022852   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365797.023604   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365797.024337   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365797.025105   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365797.025827   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365797.026605   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365797.027379   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365797.028110   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365797.028838   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365797.029582   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365797.030303   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365797.031022   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365797.031767   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365797.032506   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365797.033257   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365797.034074   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365797.041509   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365797.042278   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365797.043012   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365797.043751   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365797.044549   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365797.045360   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365797.046156   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365797.046910   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365797.047707   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365797.048505   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365797.049300   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365797.050121   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365797.050937   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365797.051812   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365797.052620   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365797.053433   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365797.061516   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365797.062344   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365797.063179   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365797.064003   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365797.064815   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365797.065611   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365797.066369   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365797.067175   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365797.067938   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365797.068786   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365797.069609   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365797.070423   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365797.071231   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365797.072033   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365797.072843   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365797.073794   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365797.082616   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365797.083452   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365797.084310   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365797.085179   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365797.085984   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365797.086932   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365797.087870   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365797.088715   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365797.089559   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365797.090438   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365797.091256   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365797.092072   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365797.092953   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365797.093861   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365797.094788   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365797.095820   77759 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
25/26 ━━━━━━━━━━━━━━━━━━━━ 0s 28ms/step - accuracy: 0.2656 - loss: 1.0196
W0000 00:00:1721365797.973659   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365797.974554   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365797.975377   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365797.976217   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365797.977059   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365797.977861   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365797.978652   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365797.979403   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365797.980169   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365797.980921   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365797.981729   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365797.982522   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365797.983312   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365797.984103   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365797.985029   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365797.992526   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365797.993322   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365797.994073   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365797.994807   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365797.995541   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365797.996280   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365797.997012   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365797.997747   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365797.998458   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365797.999169   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365797.999891   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.000642   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.001428   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.002247   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.002990   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.003746   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.004623   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.005593   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.012666   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.013511   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.014311   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.015109   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.015910   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.016707   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.017504   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.018266   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.019021   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.019975   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.020915   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.021983   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.022853   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.023709   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.024585   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.025483   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.026585   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.033248   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.033947   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.034615   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.035292   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.035974   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.036658   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.037350   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.038032   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.038723   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.039417   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.040094   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.040763   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.041508   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.042211   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.042925   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.043666   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.044483   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.045444   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.051654   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.052383   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.053081   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.053772   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.054464   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.055171   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.055878   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.056538   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.057240   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.058003   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.058756   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.059561   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.060292   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.060991   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.061706   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.062435   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.063285   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.069691   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.070419   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.071092   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.071777   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.072452   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.073143   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.073822   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.074508   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.075193   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.075895   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.076577   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.077286   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.078062   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.078775   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.079509   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.080294   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.081179   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.082271   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.088310   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.088996   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.089648   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.090304   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.090959   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.091616   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.092276   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.092934   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.093598   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.094253   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.094903   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.095553   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.096225   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.096959   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.097638   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.098357   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.104197   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.104880   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.105539   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.106186   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.106855   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.107541   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.108197   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.108851   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.109513   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.110176   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.110837   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.111496   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.112152   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.112816   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.113618   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.114267   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.114992   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.120304   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.120998   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.121660   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.122325   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.122987   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.123655   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.124325   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.124988   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.125662   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.126329   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.126999   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.127667   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.128334   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.129080   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.129766   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.130492   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.136394   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.137114   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.137791   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.138442   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.139086   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.139735   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.140397   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.141072   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.141744   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.142414   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.143070   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.143742   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.144409   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.145074   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.145804   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.151967   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.152647   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.153295   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.153943   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.154599   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.155254   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.155900   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.156547   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.157213   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.157887   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.158537   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.159192   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.159830   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.160474   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.161159   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.161817   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.167731   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.168448   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.169122   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.169783   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.170436   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.171081   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.171759   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.172449   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.173143   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.173840   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.174519   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
26/26 ━━━━━━━━━━━━━━━━━━━━ 3s 41ms/step - accuracy: 0.2662 - loss: 1.0181
W0000 00:00:1721365798.175219   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.175926   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.176615   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.177351   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.178140   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.184380   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.185082   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.185740   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.186410   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.187078   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.187745   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.188443   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.189133   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.189812   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.190506   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.191182   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.191879   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.192577   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.193270   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.193971   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.194743   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.200985   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.201667   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.202319   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.202978   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.203644   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.204300   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.204955   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.205620   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.206298   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.206981   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.207652   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.208321   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.208998   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.209662   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.210326   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.211027   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.216589   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.217329   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.218004   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.218673   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.219355   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.220040   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.220758   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.221459   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.222169   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.222885   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.223601   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.224330   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.225074   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.225867   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.226591   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.232444   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.233163   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.233828   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.234493   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.235193   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.235894   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.236607   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.237348   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.238066   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.238796   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.239522   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.240199   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.240881   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.241603   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.242363   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.248512   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.249215   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.249876   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.250530   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.251185   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.251841   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.252501   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.253176   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.253837   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.254492   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.255143   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.255804   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.256471   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.257133   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.257800   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.258519   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.264026   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.264752   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.265444   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.266124   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.266860   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.267605   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.268387   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.269158   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.269924   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.270711   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.271501   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.272220   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.272935   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.273724   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.274512   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.280635   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.281375   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.282067   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.282763   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.283481   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.284205   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.284979   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.285718   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.286455   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.287250   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.288024   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.288829   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.289623   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.290415   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.291205   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.298355   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.299085   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.299782   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.300470   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.301153   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.301828   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.302507   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.303207   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.303912   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.304597   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.305290   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.305978   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.306675   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.307382   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.308081   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365798.308881   77761 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
print("initial loss: {:.2f}".format(loss0))
print("initial accuracy: {:.2f}".format(accuracy0))
initial loss: 1.00
initial accuracy: 0.27
history = model.fit(train_dataset,
                    epochs=initial_epochs,
                    validation_data=validation_dataset)
Epoch 1/10
61/63 ━━━━━━━━━━━━━━━━━━━━ 0s 31ms/step - accuracy: 0.4405 - loss: 0.8813
W0000 00:00:1721365805.142475   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.143426   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.144325   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.145198   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.146071   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.146973   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.147954   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.148840   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.149732   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.150633   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.151651   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.152647   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.153638   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.154660   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.156944   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.163783   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.164614   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.165410   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.166199   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.166986   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.167817   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.168638   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.169476   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.170279   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.171077   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.171885   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.172750   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.173775   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.174709   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.175553   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.176630   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.177498   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.180023   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.186365   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.187345   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.188313   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.189260   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.190201   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.191108   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.192025   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.192908   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.193788   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.195041   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.196260   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.197748   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.198852   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.199927   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.201032   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.202181   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.209118   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.209853   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.210560   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.211269   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.211976   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.212691   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.213406   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.214146   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.214854   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.215571   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.216287   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.217019   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.217856   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.218617   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.219420   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.220396   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.221182   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.222448   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.229042   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.229837   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.230609   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.231402   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.232179   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.232954   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.233710   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.234431   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.235166   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.236063   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.236943   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.237922   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.238757   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.239558   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.240372   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.241215   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.242264   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.249333   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.250092   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.250801   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.251523   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.252236   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.252965   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.253691   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.254434   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.255148   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.255870   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.256596   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.257343   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.258253   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.259017   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.259887   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.260986   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.261818   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.263351   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.270761   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.271454   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.272113   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.272768   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.273441   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.274106   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.274773   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.275430   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.276117   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.276801   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.277489   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.278171   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.278885   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.279569   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.280257   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.280996   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.281766   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.282823   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.289013   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.289718   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.290382   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.291073   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.291764   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.292505   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.293178   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.293845   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.294536   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.295208   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.295877   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.296582   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.297252   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.297977   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.298966   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.299652   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.300424   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.306402   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.307100   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.307769   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.308472   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.309177   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.309846   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.310522   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.311197   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.311862   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.312545   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.313223   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.313903   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.314630   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.315333   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.316048   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.316841   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.317648   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.318868   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.325163   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.325853   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.326508   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.327160   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.327814   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.328495   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.329169   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.329823   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.330489   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.331155   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.331831   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.332512   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.333184   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.333847   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.334534   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.335272   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.341037   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.341731   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.342407   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.343099   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.343762   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
63/63 ━━━━━━━━━━━━━━━━━━━━ 0s 36ms/step - accuracy: 0.4419 - loss: 0.8791
W0000 00:00:1721365805.344426   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.345099   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.345766   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.346437   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.347104   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.347778   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.348452   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.349157   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.349829   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.355942   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.356678   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.357391   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.358102   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.358779   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.359552   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.360246   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.360915   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.361592   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.362298   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.362961   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.363622   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.364315   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.365021   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.365732   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.366515   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.373211   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.373938   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.374628   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.375320   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.376012   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.376712   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.377403   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.378102   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.378764   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.379468   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.380157   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.380843   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.381544   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.382249   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.382952   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.383749   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.390596   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.391309   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.392005   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.392695   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.393431   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.394104   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.394765   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.395449   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.396135   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.396821   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.397511   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.398221   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.398931   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.399629   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.400328   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.401069   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.407847   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.408627   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.409368   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.410097   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.410820   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.411544   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.412248   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.412933   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.413620   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.414355   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.415069   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.415781   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.416523   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.417258   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.418016   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.418840   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.425276   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.425990   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.426661   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.427329   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.428038   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.428738   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.429459   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.430174   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.430882   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.431618   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.432353   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.433079   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.433796   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.434523   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.435306   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.441857   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.442575   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.443264   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.443948   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.444621   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.445324   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.446013   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.446684   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.447370   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.448070   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.448750   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.449429   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.450112   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.450795   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.451486   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.452239   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.458438   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.459187   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.459892   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.460596   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.461349   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.462096   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.462865   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.463643   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.464415   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.465214   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.466005   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.466794   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.467576   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.468390   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.469221   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.476169   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.476942   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.477692   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.478409   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.479196   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.479976   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.480755   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.481485   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.482280   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.483070   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.483852   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.484638   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.485374   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.486179   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.486980   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.487834   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.495237   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.495984   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.496715   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.497475   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.498191   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.498963   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.499740   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.500464   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.501190   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.501928   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.502643   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.503360   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.504102   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.504859   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.505612   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365805.506472   77764 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
63/63 ━━━━━━━━━━━━━━━━━━━━ 8s 49ms/step - accuracy: 0.4426 - loss: 0.8780 - val_accuracy: 0.6027 - val_loss: 0.6711
Epoch 2/10
63/63 ━━━━━━━━━━━━━━━━━━━━ 3s 41ms/step - accuracy: 0.6678 - loss: 0.6226 - val_accuracy: 0.8280 - val_loss: 0.4723
Epoch 3/10
63/63 ━━━━━━━━━━━━━━━━━━━━ 3s 41ms/step - accuracy: 0.7808 - loss: 0.4865 - val_accuracy: 0.9196 - val_loss: 0.3519
Epoch 4/10
63/63 ━━━━━━━━━━━━━━━━━━━━ 3s 42ms/step - accuracy: 0.8440 - loss: 0.4126 - val_accuracy: 0.9406 - val_loss: 0.2800
Epoch 5/10
63/63 ━━━━━━━━━━━━━━━━━━━━ 3s 42ms/step - accuracy: 0.8631 - loss: 0.3474 - val_accuracy: 0.9480 - val_loss: 0.2357
Epoch 6/10
63/63 ━━━━━━━━━━━━━━━━━━━━ 3s 42ms/step - accuracy: 0.9017 - loss: 0.2901 - val_accuracy: 0.9592 - val_loss: 0.1975
Epoch 7/10
63/63 ━━━━━━━━━━━━━━━━━━━━ 3s 42ms/step - accuracy: 0.9151 - loss: 0.2656 - val_accuracy: 0.9604 - val_loss: 0.1786
Epoch 8/10
63/63 ━━━━━━━━━━━━━━━━━━━━ 3s 42ms/step - accuracy: 0.9019 - loss: 0.2516 - val_accuracy: 0.9666 - val_loss: 0.1571
Epoch 9/10
63/63 ━━━━━━━━━━━━━━━━━━━━ 3s 42ms/step - accuracy: 0.9017 - loss: 0.2445 - val_accuracy: 0.9691 - val_loss: 0.1456
Epoch 10/10
63/63 ━━━━━━━━━━━━━━━━━━━━ 3s 43ms/step - accuracy: 0.9240 - loss: 0.2176 - val_accuracy: 0.9728 - val_loss: 0.1344

Learning curves

Let's take a look at the learning curves of the training and validation accuracy/loss when using the MobileNetV2 base model as a fixed feature extractor.

acc = history.history['accuracy']
val_acc = history.history['val_accuracy']

loss = history.history['loss']
val_loss = history.history['val_loss']

plt.figure(figsize=(8, 8))
plt.subplot(2, 1, 1)
plt.plot(acc, label='Training Accuracy')
plt.plot(val_acc, label='Validation Accuracy')
plt.legend(loc='lower right')
plt.ylabel('Accuracy')
plt.ylim([min(plt.ylim()),1])
plt.title('Training and Validation Accuracy')

plt.subplot(2, 1, 2)
plt.plot(loss, label='Training Loss')
plt.plot(val_loss, label='Validation Loss')
plt.legend(loc='upper right')
plt.ylabel('Cross Entropy')
plt.ylim([0,1.0])
plt.title('Training and Validation Loss')
plt.xlabel('epoch')
plt.show()

png

To a lesser extent, it is also because training metrics report the average for an epoch, while validation metrics are evaluated after the epoch, so validation metrics see a model that has trained slightly longer.

Fine tuning

In the feature extraction experiment, you were only training a few layers on top of an MobileNetV2 base model. The weights of the pre-trained network were not updated during training.

One way to increase performance even further is to train (or "fine-tune") the weights of the top layers of the pre-trained model alongside the training of the classifier you added. The training process will force the weights to be tuned from generic feature maps to features associated specifically with the dataset.

Also, you should try to fine-tune a small number of top layers rather than the whole MobileNet model. In most convolutional networks, the higher up a layer is, the more specialized it is. The first few layers learn very simple and generic features that generalize to almost all types of images. As you go higher up, the features are increasingly more specific to the dataset on which the model was trained. The goal of fine-tuning is to adapt these specialized features to work with the new dataset, rather than overwrite the generic learning.

Un-freeze the top layers of the model

All you need to do is unfreeze the base_model and set the bottom layers to be un-trainable. Then, you should recompile the model (necessary for these changes to take effect), and resume training.

base_model.trainable = True
# Let's take a look to see how many layers are in the base model
print("Number of layers in the base model: ", len(base_model.layers))

# Fine-tune from this layer onwards
fine_tune_at = 100

# Freeze all the layers before the `fine_tune_at` layer
for layer in base_model.layers[:fine_tune_at]:
  layer.trainable = False
Number of layers in the base model:  154

Compile the model

As you are training a much larger model and want to readapt the pretrained weights, it is important to use a lower learning rate at this stage. Otherwise, your model could overfit very quickly.

model.compile(loss=tf.keras.losses.BinaryCrossentropy(),
              optimizer = tf.keras.optimizers.RMSprop(learning_rate=base_learning_rate/10),
              metrics=[tf.keras.metrics.BinaryAccuracy(threshold=0.5, name='accuracy')])
model.summary()
len(model.trainable_variables)
56

Continue training the model

If you trained to convergence earlier, this step will improve your accuracy by a few percentage points.

fine_tune_epochs = 10
total_epochs =  initial_epochs + fine_tune_epochs

history_fine = model.fit(train_dataset,
                         epochs=total_epochs,
                         initial_epoch=len(history.epoch),
                         validation_data=validation_dataset)
Epoch 11/20
W0000 00:00:1721365839.567237   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.570013   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.571919   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.573808   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.577852   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.579596   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.582731   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.584836   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.586729   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.588692   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.590881   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.593094   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.604647   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.606373   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.608074   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.609853   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.611629   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.613139   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.622870   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.625410   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.628272   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.632212   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.664287   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.680566   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.682272   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.683922   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.685576   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.687460   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.689311   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.691027   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.692691   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.694626   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.696544   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.698351   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.700140   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.707915   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.709335   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.710897   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.712460   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.713891   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.715352   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.716885   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.718877   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.721316   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.724492   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.730185   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.742107   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.743723   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.745167   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.746625   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.748089   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.749584   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.751105   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.752648   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.754150   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.755675   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.757155   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.758769   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.764232   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.765816   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.767358   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
1/63 ━━━━━━━━━━━━━━━━━━━━ 9:33 9s/step - accuracy: 0.9062 - loss: 0.3488
W0000 00:00:1721365839.768913   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.770510   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.771573   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.772718   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.773727   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.775050   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.776714   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.779522   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.785869   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.786942   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.787985   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.789024   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.790023   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.791073   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.792066   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.793126   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.794154   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.795176   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.796187   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.797192   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.801207   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.802154   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.803071   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.804004   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.804902   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.805825   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.806774   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.807871   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.809052   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.810453   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.812809   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.819843   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.821727   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.822675   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.823629   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.824557   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.825526   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.826520   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.827589   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.828527   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.829488   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.830445   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.831407   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.834908   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.835805   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.836699   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.837587   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.838470   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.839328   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.840201   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.841253   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.842269   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.843445   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.851619   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.852631   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.853604   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.854561   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.855653   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.856894   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.858011   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.859389   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.860416   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.862044   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.863047   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.864972   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.869931   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.870923   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.871881   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.872831   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.873779   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.874812   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.875843   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.876986   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.878013   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.879123   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.881967   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.886446   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.887451   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.888423   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.889441   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.890512   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.891526   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.893099   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.894942   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.896026   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.897095   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.898303   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.899729   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.905117   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.906084   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.906993   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.907891   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.908809   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.909812   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.910881   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.911922   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.912926   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.914131   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365839.916067   77757 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
63/63 ━━━━━━━━━━━━━━━━━━━━ 0s 53ms/step - accuracy: 0.8102 - loss: 0.4081
W0000 00:00:1721365843.020209   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.021380   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.022427   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.023447   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.024361   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.025311   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.026339   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.027376   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.028338   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.029342   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.030391   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.031436   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.035947   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.036952   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.038021   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.039073   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.040123   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.041241   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.042211   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.043230   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.044367   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.045776   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.050191   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.051166   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.052109   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.053052   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.053977   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.054898   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.055774   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.056650   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.057589   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.058523   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.059423   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.060307   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.064145   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.064966   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.065759   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.066539   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.067364   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.068188   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.069003   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.069904   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.070896   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.072067   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.075981   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.076826   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.077637   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.078435   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.079224   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.080013   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.080816   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.081619   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.082385   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.083144   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.083896   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.084673   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.087013   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.087903   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.088768   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.089565   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.090322   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.091165   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.092084   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.093186   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.096239   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.097066   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.097855   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.098635   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.099406   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.100177   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.100932   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.101725   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.102518   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.103284   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.104049   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.104813   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.107809   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.108548   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.109266   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.109994   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.110729   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.111460   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.112183   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.112974   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.113776   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.114656   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.119196   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.119964   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.120695   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.121456   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.122178   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.122909   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.123641   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.124404   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.125114   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.125827   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.126531   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.127241   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.130063   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.130795   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.131504   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.132206   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.132915   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.133615   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.134326   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.135080   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.135841   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.136646   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.139627   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.140390   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.141124   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.141861   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.142643   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.143450   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.144234   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.145099   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.145846   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.146810   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.147549   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.148616   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.151788   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.152566   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.153334   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.154098   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.154858   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.155667   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.156421   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.157208   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.157961   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.158759   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.162059   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.162834   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.163564   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.164290   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.165051   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.165785   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.166719   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.167755   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.168518   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.169279   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.170075   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.170964   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.174139   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.174875   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.175571   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.176272   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.176991   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.177742   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.178487   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.179251   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.179993   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
W0000 00:00:1721365843.180818   77756 gpu_timer.cc:114] Skipping the delay kernel, measurement accuracy will be reduced
63/63 ━━━━━━━━━━━━━━━━━━━━ 14s 82ms/step - accuracy: 0.8106 - loss: 0.4074 - val_accuracy: 0.9752 - val_loss: 0.0909
Epoch 12/20
63/63 ━━━━━━━━━━━━━━━━━━━━ 4s 62ms/step - accuracy: 0.8888 - loss: 0.2716 - val_accuracy: 0.9827 - val_loss: 0.0736
Epoch 13/20
63/63 ━━━━━━━━━━━━━━━━━━━━ 4s 62ms/step - accuracy: 0.9197 - loss: 0.2129 - val_accuracy: 0.9851 - val_loss: 0.0635
Epoch 14/20
63/63 ━━━━━━━━━━━━━━━━━━━━ 4s 62ms/step - accuracy: 0.9223 - loss: 0.1830 - val_accuracy: 0.9851 - val_loss: 0.0588
Epoch 15/20
63/63 ━━━━━━━━━━━━━━━━━━━━ 4s 62ms/step - accuracy: 0.9414 - loss: 0.1553 - val_accuracy: 0.9851 - val_loss: 0.0573
Epoch 16/20
63/63 ━━━━━━━━━━━━━━━━━━━━ 4s 62ms/step - accuracy: 0.9359 - loss: 0.1579 - val_accuracy: 0.9876 - val_loss: 0.0497
Epoch 17/20
63/63 ━━━━━━━━━━━━━━━━━━━━ 4s 61ms/step - accuracy: 0.9471 - loss: 0.1291 - val_accuracy: 0.9814 - val_loss: 0.0536
Epoch 18/20
63/63 ━━━━━━━━━━━━━━━━━━━━ 4s 61ms/step - accuracy: 0.9557 - loss: 0.1239 - val_accuracy: 0.9827 - val_loss: 0.0517
Epoch 19/20
63/63 ━━━━━━━━━━━━━━━━━━━━ 4s 61ms/step - accuracy: 0.9577 - loss: 0.1197 - val_accuracy: 0.9839 - val_loss: 0.0502
Epoch 20/20
63/63 ━━━━━━━━━━━━━━━━━━━━ 4s 61ms/step - accuracy: 0.9556 - loss: 0.1202 - val_accuracy: 0.9851 - val_loss: 0.0473

Let's take a look at the learning curves of the training and validation accuracy/loss when fine-tuning the last few layers of the MobileNetV2 base model and training the classifier on top of it. The validation loss is much higher than the training loss, so you may get some overfitting.

You may also get some overfitting as the new training set is relatively small and similar to the original MobileNetV2 datasets.

After fine tuning the model nearly reaches 98% accuracy on the validation set.

acc += history_fine.history['accuracy']
val_acc += history_fine.history['val_accuracy']

loss += history_fine.history['loss']
val_loss += history_fine.history['val_loss']
plt.figure(figsize=(8, 8))
plt.subplot(2, 1, 1)
plt.plot(acc, label='Training Accuracy')
plt.plot(val_acc, label='Validation Accuracy')
plt.ylim([0.8, 1])
plt.plot([initial_epochs-1,initial_epochs-1],
          plt.ylim(), label='Start Fine Tuning')
plt.legend(loc='lower right')
plt.title('Training and Validation Accuracy')

plt.subplot(2, 1, 2)
plt.plot(loss, label='Training Loss')
plt.plot(val_loss, label='Validation Loss')
plt.ylim([0, 1.0])
plt.plot([initial_epochs-1,initial_epochs-1],
         plt.ylim(), label='Start Fine Tuning')
plt.legend(loc='upper right')
plt.title('Training and Validation Loss')
plt.xlabel('epoch')
plt.show()

png

Evaluation and prediction

Finally you can verify the performance of the model on new data using test set.

loss, accuracy = model.evaluate(test_dataset)
print('Test accuracy :', accuracy)
6/6 ━━━━━━━━━━━━━━━━━━━━ 0s 29ms/step - accuracy: 0.9904 - loss: 0.0288
Test accuracy : 0.9895833134651184

And now you are all set to use this model to predict if your pet is a cat or dog.

# Retrieve a batch of images from the test set
image_batch, label_batch = test_dataset.as_numpy_iterator().next()
predictions = model.predict_on_batch(image_batch).flatten()
predictions = tf.where(predictions < 0.5, 0, 1)

print('Predictions:\n', predictions.numpy())
print('Labels:\n', label_batch)

plt.figure(figsize=(10, 10))
for i in range(9):
  ax = plt.subplot(3, 3, i + 1)
  plt.imshow(image_batch[i].astype("uint8"))
  plt.title(class_names[predictions[i]])
  plt.axis("off")
Predictions:
 [1 1 1 1 0 0 1 1 0 1 1 1 1 0 0 1 1 0 0 0 1 0 0 0 1 1 0 0 1 1 1 1]
Labels:
 [1 1 1 1 0 0 1 1 0 1 1 1 1 0 0 1 1 0 0 0 1 0 0 0 1 1 0 0 1 1 1 1]

png

Summary

  • Using a pre-trained model for feature extraction: When working with a small dataset, it is a common practice to take advantage of features learned by a model trained on a larger dataset in the same domain. This is done by instantiating the pre-trained model and adding a fully-connected classifier on top. The pre-trained model is "frozen" and only the weights of the classifier get updated during training. In this case, the convolutional base extracted all the features associated with each image and you just trained a classifier that determines the image class given that set of extracted features.

  • Fine-tuning a pre-trained model: To further improve performance, one might want to repurpose the top-level layers of the pre-trained models to the new dataset via fine-tuning. In this case, you tuned your weights such that your model learned high-level features specific to the dataset. This technique is usually recommended when the training dataset is large and very similar to the original dataset that the pre-trained model was trained on.

To learn more, visit the Transfer learning guide.

# MIT License
#
# Copyright (c) 2017 François Chollet                                                                                                                    # IGNORE_COPYRIGHT: cleared by OSS licensing
#
# Permission is hereby granted, free of charge, to any person obtaining a
# copy of this software and associated documentation files (the "Software"),
# to deal in the Software without restriction, including without limitation
# the rights to use, copy, modify, merge, publish, distribute, sublicense,
# and/or sell copies of the Software, and to permit persons to whom the
# Software is furnished to do so, subject to the following conditions:
#
# The above copyright notice and this permission notice shall be included in
# all copies or substantial portions of the Software.
#
# THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
# IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
# FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
# THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
# LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
# FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
# DEALINGS IN THE SOFTWARE.