def test_true_div(self):
# true_div's upcast policy is not exactly "upgrade_to_float",
# so the test is a little bit different
x_range = list(range(-127, 128))
y_range = list(range(-127, 0)) + list(range(1, 127))
xi = int8('xi')
yi = int8('yi')
xf = Scalar(theano.config.floatX)('xf')
yf = Scalar(theano.config.floatX)('yf')
ei = true_div(xi, yi)
fi = theano.function([xi, yi], ei)
ef = true_div(xf, yf)
ff = theano.function([xf, yf], ef)
for x_val in x_range:
for y_val in y_range:
outi = fi(x_val, y_val)
outf = ff(x_val, y_val)
assert outi.dtype == outf.dtype, 'incorrect dtype'
assert np.allclose(outi, outf), 'insufficient precision'
def local_gpu_elemwise(node, context_name):
op = node.op
scal_op = op.scalar_op
name = op.name
if name:
name = 'Gpu' + name
if len(node.outputs) > 1:
return
res = GpuElemwise(scal_op,
name=name,
inplace_pattern=copy.copy(op.inplace_pattern),
nfunc_spec=op.nfunc_spec)
# If the elemwise operation is a pow, casts might be required on the
# inputs and or outputs because only the (float, float)->float and
# (double, double)->double cases are implemented at the moment.
if isinstance(op.scalar_op, Pow):
# Only transfer the computation on the gpu if the output dtype is
# floating point. Else, give up on the transfer to the gpu.
out_dtype = node.outputs[0].dtype
if out_dtype not in ['float16', 'float32', 'float64']:
return
# Transfer the inputs on the GPU and cast them to the right dtype.
new_inputs = []
for inp in node.inputs:
if inp.dtype != out_dtype:
gpu_cast_op = GpuElemwise(Cast(Scalar(out_dtype)))
new_inputs.append(
gpu_cast_op(as_gpuarray_variable(inp, context_name)))
else:
new_inputs.append(as_gpuarray_variable(inp, context_name))
# Perform the exponent on the gpu and transfer the output back to the
# cpu.
gpu_output = res(*new_inputs)
cpu_output = host_from_gpu(gpu_output)
return [cpu_output]
else:
return res
def test_filter_float_subclass():
"""Make sure `Scalar.filter` can handle `float` subclasses."""
with config.change_flags(floatX="float64"):
test_type = Scalar("float64")
nan = np.array([np.nan], dtype="float64")[0]
assert isinstance(nan, float)
filtered_nan = test_type.filter(nan)
assert isinstance(filtered_nan, float)
with config.change_flags(floatX="float32"):
# Try again, except this time `nan` isn't a `float`
test_type = Scalar("float32")
nan = np.array([np.nan], dtype="float32")[0]
assert isinstance(nan, np.floating)
filtered_nan = test_type.filter(nan)
assert isinstance(filtered_nan, np.floating)