Google Colab doesn't use GPUs for Theano code

Question

I'm trying to train a model based on Theano in Google Colab. I have set the runtime type to GPU and set the Theano flag device to cuda using:

import os
os.environ['THEANO_FLAGS'] = "device=cuda,force_device=True,floatX=float32"
import theano

But it doesn't seem to use GPUs, because the output of the following cell is always 'Used the cpu':

from theano import function, config, shared, tensor as tt
import numpy
import time

vlen = 10 * 30 * 768  # 10 x #cores x # threads per core
iters = 1000

rng = numpy.random.RandomState(22)
x = shared(numpy.asarray(rng.rand(vlen), config.floatX))
f = function([], tt.exp(x))
print(f.maker.fgraph.toposort())
t0 = time.time()
for i in range(iters):
    r = f()
t1 = time.time()
print("Looping %d times took %f seconds" % (iters, t1 - t0))
print("Result is %s" % (r,))
if numpy.any([isinstance(x.op, theano.tensor.elemwise.Elemwise) and
              ('Gpu' not in type(x.op).__name__)
              for x in f.maker.fgraph.toposort()]):
    print('Used the cpu')
else:
    print('Used the gpu')

I have installed PyGPU in conda-forge as suggested in https://discourse.pymc.io/t/pymc3-with-gpu-support-on-google-colab/1649/2 but it didn't change anything.

Can anybody help with this problem? Thanks in advance.

Google Colab doesn't use GPUs for Theano code

0 Answers