使用 TPU 在 GCP 上进行 Keras/Tensorflow 培训

Question

I am trying to train a model on GCP with keras and tensorflow 1.15. From now my code is similar to what I could do on colab, namely:

# TPUs
import tensorflow as tf
print(tf.__version__)
cluster_resolver = tf.distribute.cluster_resolver.TPUClusterResolver("tpu-name")
tf.config.experimental_connect_to_cluster(cluster_resolver)
tf.tpu.experimental.initialize_tpu_system(cluster_resolver)
tpu_strategy = tf.distribute.experimental.TPUStrategy(cluster_resolver)
print("Number of accelerators: ", tpu_strategy.num_replicas_in_sync)


import numpy as np


np.random.seed(123)  # for reproducibility
import tensorflow as tf
from tensorflow.keras.models import Sequential
from tensorflow.keras.layers import Dense, Dropout, Activation, Flatten
from tensorflow.keras.layers import Convolution2D, MaxPooling2D, Input
from tensorflow.keras import utils
from tensorflow.keras.datasets import mnist, cifar10
from tensorflow.keras.models import Model

# 4. Load data into train and test sets
(X_train, y_train) = load_data(sets="gs://BUCKETS/dogscats/train/",target_size=img_size)
(X_test, y_test) =  load_data(sets="gs://BUCKETS/dogscats/valid/",target_size=img_size)
print(X_train.shape, X_test.shape)

# 5. Preprocess input data
#X_train = X_train.reshape(X_train.shape[0], 28, 28, 1)
#X_test = X_test.reshape(X_test.shape[0], 28, 28,1)
X_train = X_train.astype('float32')
X_test = X_test.astype('float32')
X_train /= 255.0
X_test /= 255.0

print(y_train.shape, y_test.shape)
# 6. Preprocess class labels One hot encoding
Y_train = utils.to_categorical(y_train, 2)
Y_test = utils.to_categorical(y_test, 2)
print(Y_train.shape, Y_test.shape)

with tpu_strategy.scope():
  model = make_model((img_size, img_size, 3))
  # 8. Compile model
  model.compile(loss='categorical_crossentropy',
                optimizer="sgd",
                metrics=['accuracy'])

model.summary()

batch_size = 1250 * tpu_strategy.num_replicas_in_sync
# 9. Fit model on training data
model.fit(X_train, Y_train, steps_per_epoch=len(X_train)//batch_size,  
            epochs=5, verbose=1)

But my data is on the bucket and my code is on an VM. So what I have to do? I tried to load my data using "gs://BUCKETS" but it does not work. What should I do? EDIT: I add my code to load data, I forgot it sorry.

def load_data(sets="dogcats/train/", k = 5000, target_size=250):
  # define location of dataset
  folder = sets
  photos, labels = list(), list()
  # determine class
  output = 0.0
  for i, dog in enumerate(listdir(folder + "dogs/")):
    if i >= k:
      break
    # load image
    photo = load_img(folder + "dogs/" +dog, target_size=(target_size, target_size))
    # convert to numpy array
    photo = img_to_array(photo)
    # store
    photos.append(photo)
    labels.append(output)

  output = 1.0

  for i, cat in enumerate(listdir(folder + "cats/") ):
    if i >= k:
      break
    # load image
    photo = load_img(folder + "cats/"+cat, target_size=(target_size, target_size))
    # convert to numpy array
    photo = img_to_array(photo)
    # store
    photos.append(photo)
    labels.append(output)

  # convert to a numpy arrays
  photos = asarray(photos)
  labels = asarray(labels)
  print(photos.shape, labels.shape)
  photos, labels = shuffle(photos, labels, random_state=0)
  return photos, labels

EDIT2: To complete the answer of @daudnadeem in case some other people are in the same case.

My goal was to get images from a bucket, so the code works well and allowed to get byte object. To transform it into image you just need to use PIL library:

from PIL import Image
from io import BytesIO
import numpy as np

from google.cloud import storage
client = storage.Client()
bucket = client.get_bucket("BUCKETS")
blob = bucket.get_blob('dogscats/train/<you-will-need-to-point-to-a-file-and-not-a-directory>')
data = blob.download_as_string()

img = Image.open(BytesIO(data))
img = np.array(img)

Answer 1

(X_train, y_train) = load_data(sets="gs://BUCKETS/dogscats/train/",target_size=img_size)
(X_test, y_test) =  load_data(sets="gs://BUCKETS/dogscats/valid/",target_size=img_size)

This obviously won't work since essentially all you've done is given sets a string. What you need to do is download this data as a string, and then use that.

First install the package pip install google-cloud-storage or pip3 install google-cloud-storage

pip -> Python

pip3 -> Python3

Have a look at this , you will need a service account to interact with GCP from your code. For authentication purposes.

When you get your service account as a json, you need to do one of two things:

Set it as an env variable: export GOOGLE_APPLICATION_CREDENTIALS="/home/user/Downloads/[FILE_NAME].json"

or my preferrable workaround

gcloud auth activate-service-account \
  <repalce-with-email-from-json-file> \
          --key-file=<path/to/your/json/file> --project=<name-of-your-gcp-project>

Now lets look at how you can use google-cloud-storage library to download your file as a string:

from google.cloud import storage
client = storage.Client()
bucket = client.get_bucket("BUCKETS")
blob = bucket.get_blob('/dogscats/train/<you-will-need-to-point-to-a-file-and-not-a-directory>')
data = blob.download_as_string()

Now that you have your data as a string, you can simply pass data into load data like so (X_train, y_train) = load_data(sets=data,target_size=img_size)

It sounds complex but heres a quick psuedo layout:

1. Install google-cloud-storage
2. Go to Google Cloud Platform Console -> IAM & Admin -> Service Accounts
3. Create service account with relative permissions (google-cloud-storage)
4. Download the (JSON) file, and remember location.
5. Activate service account
6. Download file as String and pass that string to your load_data(data)

Hope that helps!