def test_tensor_property_name():
string_io = io.StringIO()
fmt = torchsnooper.TensorFormat(property_name=True)
@torchsnooper.snoop(string_io, tensor_format=fmt)
def my_function():
x = torch.randn((5, 8))
return x
my_function()
output = string_io.getvalue()
print(output)
assert_output(output, (
CallEntry(),
LineEntry(),
VariableEntry(
'x',
'tensor<shape=(5, 8), dtype=float32, device=cpu, requires_grad=False>'
),
LineEntry(),
ReturnEntry(),
ReturnValueEntry(
'tensor<shape=(5, 8), dtype=float32, device=cpu, requires_grad=False>'
),
))
def test_tensor_property_name():
string_io = io.StringIO()
fmt = torchsnooper.TensorFormat(property_name=True)
@torchsnooper.snoop(string_io,
max_variable_length=100000,
tensor_format=fmt)
def my_function():
x = torch.randn((5, 8))
return x
my_function()
output = string_io.getvalue()
print(output)
assert_output(output, (
SourcePathEntry(),
CallEntry(),
LineEntry(),
VariableEntry(
'x',
'tensor<shape=(5, 8), dtype=float32, device=cpu, requires_grad=False, has_nan=False, has_inf=False, memory_format=torch.contiguous>'
),
LineEntry(),
ReturnEntry(),
ReturnValueEntry(
'tensor<shape=(5, 8), dtype=float32, device=cpu, requires_grad=False, has_nan=False, has_inf=False, memory_format=torch.contiguous>'
),
ElapsedTimeEntry(),
))
def test_tensor_property_selector():
string_io = io.StringIO()
fmt = torchsnooper.TensorFormat(properties=('shape', 'device',
'requires_grad'))
@torchsnooper.snoop(string_io, tensor_format=fmt)
def my_function():
x = torch.randn((5, 8))
return x
my_function()
output = string_io.getvalue()
print(output)
assert_output(output, (
CallEntry(),
LineEntry(),
VariableEntry('x', 'tensor<(5, 8), cpu>'),
LineEntry(),
ReturnEntry(),
ReturnValueEntry('tensor<(5, 8), cpu>'),
))
