Using GPU in venv

Virtualenv

Create venv

virtualenv .\venv_directory

Activate venv

.\virtualenv\Scripts\activate

Install tensorflow

As we want to use the GPU we have to install

pip install tensorflow-gpu

Install VisualStudio 2017 Community

We need 2017 as we have to use CUDAv10.0 as TensorFlow is not yet compatible with CUDAv10.1, the installer is sadly only available with a free Microsoft Dev account and under the link for MyVisualStudio - Downloads.

Install VS2017 Community and the C++ Workload

Install CUDA and cuDNN

Get the CUDAv10.0 Installer from the CUDA Toolkit Archive and the cuDNN zip file from the cuDNN Archive.

Install CUDAv10.0 and copy the files in the cuDNN zip file to the install directory.

Check if the installer has correctly added a system environment variable called ‘CUDA_PATH’ pointing to the install directory, otherwise add it yourself.

Check if installation was successful

Check for GPU

import tensorflow as tf
print("Num GPUs Available: ", 
len(tf.config.experimental.list_physical_devices('GPU')))

Check if CUDA Toolkit is used correctly

from tensorflow.keras.datasets import mnist
from tensorflow.keras.models import Sequential
from tensorflow.keras.layers import Dense
from tensorflow.keras import utils
(X_train, y_train), (X_test, y_test) = mnist.load_data()
X_train = X_train.reshape(60000,784)
X_test = X_test.reshape(10000,784)
X_train = X_train.astype('float32')
X_test = X_test.astype('float32')
X_train /= 255.0
X_test /= 255.0
n_classes = 10
Y_train = utils.to_categorical(y_train)
Y_test = utils.to_categorical(y_test)
number_of_epochs = 10
batch_size = 128
dimension_input = X_train.shape[1]
model = Sequential()
model.add(Dense(n_classes, input_shape=(784,), activation='softmax'))
model.summary()
model.compile(optimizer='adam', loss='categorical_crossentropy', metrics=['accuracy'])
log = model.fit(X_train, Y_train, batch_size=batch_size, epochs=number_of_epochs, verbose=True, validation_data=(X_test, Y_test))

Enjoy!

Source: m4l4 on StackOverflow