Skip to content

Instantly share code, notes, and snippets.

@prerakmody
Last active March 8, 2019 08:53
Show Gist options
  • Star 0 You must be signed in to star a gist
  • Fork 0 You must be signed in to fork a gist
  • Save prerakmody/e960089cfd4770a8140039490a626b5c to your computer and use it in GitHub Desktop.
Save prerakmody/e960089cfd4770a8140039490a626b5c to your computer and use it in GitHub Desktop.
Tensorflow GPU Check (Ubuntu) (Jupyter Notebook)
import pytorch
use_gpu = torch.cuda.is_available()
class Net(nn.Module):
def __init__(self):
super(Net, self).__init__()
self.conv1 = nn.Conv2d(1, 6, 3, padding=1)
self.pool = nn.MaxPool2d(2, 2)
self.conv2 = nn.Conv2d(6, 16, 3, padding=1)
self.pool = nn.MaxPool2d(2, 2)
self.fc1 = nn.Linear(16 * 7 * 7, 120)
self.fc2 = nn.Linear(120, 84)
self.fc3 = nn.Linear(84, 10)
def forward(self, x):
return x
net = Net()
if use_gpu:
net = net.cuda()
from torchsummary import summary
summary(net, (1, 28, 28))
"""
Nvidia GeForce 940M = Supports only CUDA 8.
CUDA 8 needs cudnn 7 to work.
CUDA 8 only supports tensorflow-gpu 1.4.1 and older.
pip install tensorflow-gpu==1.4.1
pip install tensorboard==1.6.0
pip list | grep tensor
"""
def check_gpu(verbose=1):
import os
if verbose:
try:
print('\n =========== NVIDIA =========== ')
print('0. Nvidia Hardware (lspci | grep -i nvidia) : ')
tmp = [each for each in os.popen('lspci | grep -i nvidia')]
if len(tmp): print(' ---> ', tmp[0])
else: print(' ---> No Graphics Card Found')
tmp = [each for each in os.popen('cat /proc/driver/nvidia/version')]
print('0. Nvidia Driver (cat /proc/driver/nvidia/version) : ')
if len(tmp): print(' ---> ', tmp[0])
else: print(' ---> No Nvidia driver found ...')
print('0. Nvidia Driver (ls /usr/lib | grep nvidia-) : ')
tmp = [each.strip('\n') for each in os.popen('ls /usr/lib | grep nvidia-')]
if len(tmp): print(' ---> ', tmp)
else: print(' ---> No Nvidia driver found ...')
print('\n0. Nvidia Packages : (dpkg --get-selections | grep nvidia) : ')
tmp = [each for each in os.popen('dpkg --get-selections | grep nvidia')]
if len(tmp):
for each in os.popen('dpkg --get-selections | grep nvidia'):
print(' ---> ', each.strip('\n'))
else:
print(' ---> No Nvidia packages found ...')
print('\n0. Nvidia Control Panel : nvidia-smi')
tmp = [each.strip('\n').strip('|') for each in os.popen('nvidia-smi')]
if len(tmp):
print(' ---> ', tmp[2])
else:
print(' ---> No Nvidia Control Panel Detected')
print('\n =========== CUDA =========== ')
print('1. Cuda Version : nvcc -V')
tmp = [each.strip('\n') for each in os.popen('nvcc -V')]
if len(tmp): print(' ---> ', tmp[-1])
try:
print('1. Envs : PATH (containing cuda)', [each for each in os.environ['PATH'].split(':') if each.find('cuda') > -1])
print('1. Envs : CUDA_HOME', os.environ['CUDA_HOME'])
print('1. Envs : CUDA_ROOT', os.environ['CUDA_ROOT'])
print('1. Envs : LD_LIBRARY_PATH:', os.environ['LD_LIBRARY_PATH'])
tmp_cmd = 'find ' + os.environ['CUDA_HOME'] + ' -name *dnn*'
print('1. CUDnn :', [ each.replace('\n', '') for each in os.popen(tmp_cmd)])
except:
print ('Error : COuld not find local CUDA envs')
print('\n =========== CUDA & NVIDIA ========== ')
try:
tmp_path = os.environ['CUDA_HOME'] + '/samples/1_Utilities/deviceQuery/deviceQuery'
print('2. ', tmp_path)
tmp = [each.strip('\n') for each in os.popen(tmp_path)]
if len(tmp):
print(' ---> ', tmp[4])
print(' ---> ', tmp[6])
print(' ---> ', tmp[-1])
print(' ---> CUDA successfully detected Nvidia drivers')
else:
print(' ---> CUDA did not detect Nvidia drivers')
except:
print (' ---> Some error')
except:
print('Error:')
print('\n =========== nvidia-smi ========== ')
tmp = [each.strip('\n') for each in os.popen('nvidia-smi --query-compute-apps=pid --format=csv,noheader')]
print ('Query for existing PIDs using GPU : nvidia-smi --query-compute-apps=pid --format=csv,noheader')
print (' ---> ', tmp)
if len(tmp):
print ('Damn son! You gotta kill the PIDS - {0} and then run nvidia-smi -r under root'.format(tmp))
print (' ---> Then come back and run this script again')
else:
import tensorflow as tf
from tensorflow.python.client import device_lib
devices = device_lib.list_local_devices()
for device in devices:
print ('TensorFlow Devices:', str(device.name).replace('\n',''))
if len(devices) > 1:
print ('\n')
return 1
else:
print (' --> Try resetting the Nvidia GPU')
print ('\n')
return 0
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment