OSSForks/pytorch
1
0
Fork 0
mirror of https://github.com/zebrajr/pytorch.git synced 2025-12-06 12:20:52 +01:00
Code Issues Actions 38 Packages Projects Releases Wiki Activity
6fea7499c2
pytorch/test/test_ops.py
Richard Zou 6fea7499c2 CompositeImplicitAutograd compliance testing (#65819) 
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/65819

Related to #61669.

Functions registered as CompositeImplicitAutograd MUST work for most, if
not all, backends. This includes Tensor subclasses.

To achieve this, we (PyTorch) impose a set of constraints on how a
CompositeImplicitAutograd function can be written.

Concretely, this PR adds tests for all OpInfos that checks for
compliance. The things that get tested in this PR apply to composite
ops and are that:
- the op does not change the metadata of a Tensor without performing
dispatches
- the op does not call set_ or resize_
- the op does not directly access the data ptr

The mechanism for the test is to create a new __torch_dispatch__
object, CompositeCompliantTensor. For each operator, we wrap all inputs
in CompositeCompliantTensor, turn on python mode for it,
and send it through the operator.

Non-CompositeImplicitAutograd operators will pass the test because they
perform a dispatch to backend code. Here's how CompositeCompliantTensor
catches problems:

- If it sees set_ or resize_ getting called, it will directly error
out
- After each operation, CompositeCompliantTensor checks to make sure
that its metadata is consistent with that of the thing it is wrapping.
If the CompositeImplicitAutograd op modifes the metadata directly
(through e.g. the TensorImpl API) then the metadata will go out of sync.
- If data_ptr gets called, that returns a nice error (because the
storage is meta).

CompositeCompliantTensor is written in an interesting way. First off,
if a view operation occurs (e.g. `B = A.view_op(...)`), then B.storage()
must alias A.storage() where B.storage() is CompositeCompliantTensor's
storage, NOT the storage of the tensor it is wrapping. This is an
invariant in autograd, see #62182 for details. To handle
this we replay the view on A's storage and set it as B's storage.

Secondly, there are cases where the metadata is allowed to go out of
sync. I believe this is only possible with in-place view functions, like
transpose_, t_, squeeze_, unsqueeze_. Those are special cased.

Finally, I added a new section to aten/src/ATen/native/README.md about
what it means to be CompositeImplicitAutograd Compliant

Test Plan: - run tests

Reviewed By: ezyang, bdhirsh

Differential Revision: D31268369

Pulled By: zou3519

fbshipit-source-id: 31634b1cbe1778ab30196013cfc376ef9bd2e8b1
2021-11-30 07:35:22 -08:00

1171 lines
58 KiB
Python
Raw Blame History

# Owner(s): ["high priority"]
from collections.abc import Sequence
from functools import partial, wraps
import warnings
import unittest
import torch
from torch.testing import FileCheck, make_tensor
from torch.testing._internal.common_dtype import floating_and_complex_types_and, get_all_dtypes
from torch.testing._internal.common_utils import \
(TestCase, is_iterable_of_tensors, run_tests, IS_SANDCASTLE, clone_input_helper,
gradcheck, gradgradcheck, IS_IN_CI, suppress_warnings, noncontiguous_like,
TEST_WITH_ASAN, IS_WINDOWS, IS_FBCODE)
from torch.testing._internal.common_methods_invocations import \
(op_db, _NOTHING, UnaryUfuncInfo, ReductionOpInfo, SpectralFuncInfo)
from torch.testing._internal.common_device_type import \
(deviceCountAtLeast, instantiate_device_type_tests, ops, onlyCPU,
onlyCUDA, onlyNativeDeviceTypes, skipCUDAIfRocm, OpDTypes, skipMeta)
from torch.testing._internal.common_jit import JitCommonTestCase, check_against_reference
from torch.testing._internal.jit_metaprogramming_utils import create_script_fn, create_traced_fn, \
check_alias_annotation
from torch.testing._internal.jit_utils import disable_autodiff_subgraph_inlining
import torch.testing._internal.opinfo_helper as opinfo_helper
from torch.testing._internal.composite_compliance import _check_composite_compliance
# variant testing is only done with torch.float and torch.cfloat to avoid
# excessive test times and maximize signal to noise ratio
_variant_ops = partial(ops, dtypes=OpDTypes.supported,
allowed_dtypes=(torch.float, torch.cfloat))
# Get names of all the operators which have ref in their entry in OpInfo (testing infra)
# except for Unary Ufuncs (separately implemented in test/test_unary_ufuncs.py)
# and Spectral Functions (separately implemented for only 1D as of now, in test/test_spectral_ops.py)
_ref_test_ops = list(filter(lambda op: not isinstance(op, (UnaryUfuncInfo, ReductionOpInfo,
SpectralFuncInfo)) and op.ref is not None and op.ref is not _NOTHING, op_db))
# Tests that apply to all operators and aren't related to any particular
# system
class TestCommon(TestCase):
exact_dtype = True
# Verifies, on teardown, that no OpInfo is still using dynamic dtypes in CI
@classmethod
def tearDownClass(cls):
super().tearDownClass()
if IS_IN_CI:
err_msg = ("The operator(s) below is(are) using dynamic_dtypes in the OpInfo entries."
"This is OK for testing, but be sure to set the dtypes manually before landing your PR!")
# Assure no opinfo entry has dynamic_dtypes
filtered_ops = list(filter(opinfo_helper.is_dynamic_dtype_set, op_db))
for op in filtered_ops:
fmt_str = opinfo_helper.str_format_dynamic_dtype(op)
err_msg += "\n" + fmt_str
assert len(filtered_ops) == 0, err_msg
# Validates that each OpInfo specifies its forward and backward dtypes
# correctly for CPU and CUDA devices
@skipMeta
@skipCUDAIfRocm
@onlyNativeDeviceTypes
@ops(op_db, dtypes=OpDTypes.none)
def test_dtypes(self, device, op):
# dtypes to try to backward in
allowed_backward_dtypes = floating_and_complex_types_and(torch.bfloat16, torch.float16)
# lists for (un)supported dtypes
supported_dtypes = []
unsupported_dtypes = []
supported_backward_dtypes = []
unsupported_backward_dtypes = []
def unsupported(dtype):
unsupported_dtypes.append(dtype)
if dtype in allowed_backward_dtypes:
unsupported_backward_dtypes.append(dtype)
for dtype in get_all_dtypes():
# tries to acquire samples - failure indicates lack of support
requires_grad = (dtype in allowed_backward_dtypes and op.supports_autograd)
try:
samples = op.sample_inputs(device, dtype, requires_grad=requires_grad)
except Exception as e:
unsupported(dtype)
continue
# Counts number of successful backward attempts
# NOTE: This exists as a kludge because this only understands how to
# request a gradient if the output is a tensor or a sequence with
# a tensor as its first element.
num_backward_successes = 0
for sample in samples:
# tries to call operator with the sample - failure indicates
# lack of support
try:
result = op(sample.input, *sample.args, **sample.kwargs)
except Exception as e:
# NOTE: some ops will fail in forward if their inputs
# require grad but they don't support computing the gradient
# in that type! This is a bug in the op!
unsupported(dtype)
# Short-circuits testing this dtype -- it doesn't work
if dtype in unsupported_dtypes:
break
# Short-circuits if the dtype isn't a backward dtype or
# it's already identified as not supported
if dtype not in allowed_backward_dtypes or dtype in unsupported_backward_dtypes:
continue
# Checks for backward support in the same dtype
try:
result = sample.output_process_fn_grad(result)
if isinstance(result, torch.Tensor):
backward_tensor = result
elif isinstance(result, Sequence) and isinstance(result[0], torch.Tensor):
backward_tensor = result[0]
else:
continue
# Note: this grad may not have the same dtype as dtype
# For functions like complex (float -> complex) or abs
# (complex -> float) the grad tensor will have a
# different dtype than the input.
# For simplicity, this is still modeled as these ops
# supporting grad in the input dtype.
grad = torch.randn_like(backward_tensor)
backward_tensor.backward(grad)
num_backward_successes += 1
except Exception as e:
unsupported_backward_dtypes.append(dtype)
if dtype not in unsupported_dtypes:
supported_dtypes.append(dtype)
if num_backward_successes > 0 and dtype not in unsupported_backward_dtypes:
supported_backward_dtypes.append(dtype)
# Checks that dtypes are listed correctly and generates an informative
# error message
device_type = torch.device(device).type
claimed_supported = set(op.supported_dtypes(device_type))
supported_dtypes = set(supported_dtypes)
supported_but_unclaimed = supported_dtypes - claimed_supported
claimed_but_unsupported = claimed_supported - supported_dtypes
msg = """The supported dtypes for {0} on {1} according to its OpInfo are
{2}, but the detected supported dtypes are {3}.
""".format(op.name, device_type, claimed_supported, supported_dtypes)
if len(supported_but_unclaimed) > 0:
msg += "The following dtypes should be added to the OpInfo: {0}. ".format(supported_but_unclaimed)
if len(claimed_but_unsupported) > 0:
msg += "The following dtypes should be removed from the OpInfo: {0}.".format(claimed_but_unsupported)
self.assertEqual(supported_dtypes, claimed_supported, msg=msg)
# Checks that backward dtypes are listed correctly and generates an
# informative error message
# NOTE: this code is nearly identical to the check + msg generation
claimed_backward_supported = set(op.supported_backward_dtypes(device_type))
supported_backward_dtypes = set(supported_backward_dtypes)
supported_but_unclaimed = supported_backward_dtypes - claimed_backward_supported
claimed_but_unsupported = claimed_backward_supported - supported_backward_dtypes
msg = """The supported backward dtypes for {0} on {1} according to its OpInfo are
{2}, but the detected supported backward dtypes are {3}.
""".format(op.name, device_type, claimed_backward_supported, supported_backward_dtypes)
if len(supported_but_unclaimed) > 0:
msg += "The following backward dtypes should be added to the OpInfo: {0}. ".format(supported_but_unclaimed)
if len(claimed_but_unsupported) > 0:
msg += "The following backward dtypes should be removed from the OpInfo: {0}.".format(claimed_but_unsupported)
self.assertEqual(supported_backward_dtypes, claimed_backward_supported, msg=msg)
# Validates that each OpInfo works correctly on different CUDA devices
@skipCUDAIfRocm
@onlyCUDA
@deviceCountAtLeast(2)
@ops(op_db, allowed_dtypes=(torch.float32, torch.long))
def test_multiple_devices(self, devices, dtype, op):
for cuda_device_str in devices:
cuda_device = torch.device(cuda_device_str)
# NOTE: only tests on first sample
samples = op.sample_inputs(cuda_device, dtype)
sample = samples[0]
result = op(sample.input, *sample.args, **sample.kwargs)
if isinstance(result, torch.Tensor):
self.assertTrue(result.device == cuda_device)
elif is_iterable_of_tensors(result):
self.assertTrue(all(map(lambda t: t.device == cuda_device, result)))
else:
self.skipTest("Skipped! Only supports single tensor or iterable of tensor outputs.")
# Tests that the function and its (ndarray-accepting) reference produce the same
# values on the tensors from sample_inputs func for the corresponding op.
@onlyNativeDeviceTypes
@suppress_warnings
@ops(_ref_test_ops, allowed_dtypes=(torch.float32, torch.long, torch.complex64))
def test_reference_testing(self, device, dtype, op):
sample_inputs = op.sample_inputs(device, dtype)
for sample_input in sample_inputs:
self.compare_with_reference(op, op.ref, sample_input)
@skipMeta
@onlyNativeDeviceTypes
@ops([op for op in op_db if op.error_inputs_func is not None], dtypes=OpDTypes.none)
def test_errors(self, device, op):
error_inputs = op.error_inputs(device)
for ei in error_inputs:
si = ei.sample_input
with self.assertRaisesRegex(ei.error_type, ei.error_regex):
op(si.input, *si.args, **si.kwargs)
# Tests that the function produces the same result when called with
# noncontiguous tensors.
# TODO: get working with Windows by addressing failing operators
# TODO: get working with ASAN by addressing failing operators
@unittest.skipIf(IS_WINDOWS, "Skipped under Windows")
@unittest.skipIf(TEST_WITH_ASAN, "Skipped under ASAN")
@onlyNativeDeviceTypes
@suppress_warnings
@ops(op_db, allowed_dtypes=(torch.float32, torch.long, torch.complex64))
def test_noncontiguous_samples(self, device, dtype, op):
test_grad = dtype in op.supported_backward_dtypes(torch.device(device).type)
sample_inputs = op.sample_inputs(device, dtype, requires_grad=test_grad)
for sample_input in sample_inputs:
t_inp, t_args, t_kwargs = sample_input.input, sample_input.args, sample_input.kwargs
n_inp, n_args, n_kwargs = sample_input.noncontiguous()
# Verifies sample input tensors should have no grad or history
sample_tensor = t_inp if isinstance(t_inp, torch.Tensor) else t_inp[0]
assert sample_tensor.grad is None
assert sample_tensor.grad_fn is None
# validates forward
expected = op(t_inp, *t_args, **t_kwargs)
actual = op(n_inp, *n_args, **n_kwargs)
self.assertEqual(actual, expected)
# validates backward
# NOTE: only handles single tensor outputs and the first tensor
# of ops that output a sequence
# Short-circuits if the op doesn't support grad in this device x dtype
if not test_grad:
continue
if isinstance(expected, torch.Tensor):
expected_backward_tensor = expected
actual_backward_tensor = actual
elif isinstance(expected, Sequence) and isinstance(expected[0], torch.Tensor):
expected_backward_tensor = expected[0]
actual_backward_tensor = actual[0]
else:
continue
grad_for_expected = torch.randn_like(expected_backward_tensor)
grad_for_actual = noncontiguous_like(grad_for_expected)
expected_backward_tensor.backward(grad_for_expected)
actual_backward_tensor.backward(grad_for_actual)
# Acquires grad (which may be on the first element in a list)
expected_grad = t_inp.grad if isinstance(t_inp, torch.Tensor) else t_inp[0].grad
actual_grad = n_inp.grad if isinstance(n_inp, torch.Tensor) else n_inp[0].grad
# TODO: FIXME: only validates grad on first tensor input
self.assertEqual(actual_grad, expected_grad)
# Validates ops implement the correct out= behavior
# See https://github.com/pytorch/pytorch/wiki/Developer-FAQ#how-does-out-work-in-pytorch
# for a description of the correct behavior
# TODO: operations that support out= but don't support float
# are not covered by this test.
@ops(op_db, allowed_dtypes=(torch.float,))
def test_out(self, device, dtype, op):
# TODO: verify the op doesn't support the out= kwarg
if not op.supports_out:
self.skipTest("Skipped! Op doesn't support out= kwarg.")
# NOTE: only tests on first sample
samples = op.sample_inputs(device, dtype)
sample = samples[0]
# calls it normally to get the expected result
expected = op(sample.input, *sample.args, **sample.kwargs)
op_out = partial(op, sample.input, *sample.args, **sample.kwargs)
# Short-circuits if output is not a single tensor or an
# iterable of tensors
if not isinstance(expected, torch.Tensor) and not is_iterable_of_tensors(expected, include_empty=True):
self.skipTest("Skipped! Only supports single tensor or iterable of tensor outputs.")
# A wrapper around map that works with single tensors and always
# instantiates the map. Used below to apply transforms to
# single tensor and iterable tensor outputs.
def _apply_out_transform(fn, out):
if isinstance(out, torch.Tensor):
return fn(out)
# assumes (see above) that out is an iterable of tensors
return tuple(map(fn, out))
# Case 0: out= with the correct shape, dtype, and device
# but NaN values for floating point and complex tensors, and
# maximum values for integer tensors.
# Expected behavior: out= values have no effect on the computation.
def _case_zero_transform(t):
try:
info = torch.iinfo(t.dtype)
return torch.full_like(t, info.max)
except TypeError as te:
# for non-integer types fills with NaN
return torch.full_like(t, float('nan'))
out = _apply_out_transform(_case_zero_transform, expected)
result = op_out(out=out)
self.assertEqual(expected, out)
# Checks that the returned value shares storage with out
# NOTE: only checks on the CPU and CUDA device types since some
# device types don't have storage
if self.device_type == 'cpu' or self.device_type == 'cuda':
if isinstance(out, torch.Tensor):
self.assertEqual(out.storage().data_ptr(), result.storage().data_ptr())
else:
for out_t, result_t in zip(out, result):
self.assertEqual(out_t.storage().data_ptr(), result_t.storage().data_ptr())
# Case 1: out= with the correct shape, dtype, and device,
# but noncontiguous.
# Expected behavior: strides are respected and `out` storage is not changed.
def _case_one_transform(t):
return make_tensor(t.shape,
dtype=t.dtype,
device=t.device,
noncontiguous=True)
# Extracts strides from a tensor or iterable of tensors into a tuple
def _extract_strides(out):
if isinstance(out, torch.Tensor):
return (out.stride(),)
# assumes (see above) that out is an iterable of tensors
return tuple(map(lambda t: t.stride(), out))
def _extract_data_ptrs(out):
if isinstance(out, torch.Tensor):
return (out.data_ptr(),)
# assumes (see above) that out is an iterable of tensors
return tuple(map(lambda t: t.data_ptr(), out))
out = _apply_out_transform(_case_one_transform, expected)
original_strides = _extract_strides(out)
original_ptrs = _extract_data_ptrs(out)
op_out(out=out)
final_strides = _extract_strides(out)
final_ptrs = _extract_data_ptrs(out)
self.assertEqual(expected, out)
self.assertEqual(original_strides, final_strides)
self.assertEqual(original_ptrs, final_ptrs)
# Case 2: out= with the correct dtype and device, but the wrong shape
# Expected behavior: resize with a warning.
def _case_two_transform(t):
wrong_shape = list(t.shape)
if len(wrong_shape) == 0:
# Handles scalar tensor case (empty list)
wrong_shape = [2]
else:
wrong_shape[-1] = wrong_shape[-1] + 1
return make_tensor(wrong_shape, dtype=t.dtype, device=t.device)
out = _apply_out_transform(_case_two_transform, expected)
msg_fail = "Resized a non-empty tensor but did not warn about it."
with self.assertWarnsRegex(UserWarning, "An output with one or more elements", msg=msg_fail):
op_out(out=out)
self.assertEqual(expected, out)
# Case 3: out= with the correct dtype and device, but an empty
# tensor.
# Expected behavior: resize without warning.
def _case_three_transform(t):
return make_tensor((0,),
dtype=t.dtype,
device=t.device)
out = _apply_out_transform(_case_three_transform, expected)
with warnings.catch_warnings(record=True) as caught:
warnings.simplefilter("always")
op_out(out=out)
# Verifies no warning is a resize warning
for w in caught:
if "An output with one or more elements" in str(w.message):
self.fail("Resizing an out= argument with no elements threw a resize warning!")
self.assertEqual(expected, out)
# Case 4: out= with correct shape and dtype, but wrong device.
wrong_device = None
if torch.device(device).type != 'cpu':
wrong_device = 'cpu'
elif torch.cuda.is_available():
wrong_device = 'cuda'
if wrong_device is not None:
def _case_four_transform(t):
return make_tensor(t.shape, dtype=t.dtype, device=wrong_device)
out = _apply_out_transform(_case_four_transform, expected)
msg_fail = f"Expected RuntimeError when calling with input.device={device} and out.device={wrong_device}"
with self.assertRaises(RuntimeError, msg=msg_fail):
op_out(out=out)
# Case 5: out= with correct shape and device, but a dtype
# that output cannot be "safely" cast to (long).
# Expected behavior: error.
# NOTE: this case is filtered by dtype since some ops produce
# bool tensors, for example, which can be safely cast to any
# dtype. It is applied when single tensors are floating point or complex
# dtypes, or if an op returns multiple tensors when at least one such
# tensor is a floating point or complex dtype.
_dtypes = floating_and_complex_types_and(torch.float16, torch.bfloat16)
if (isinstance(expected, torch.Tensor) and expected.dtype in _dtypes or
(not isinstance(expected, torch.Tensor) and any(t.dtype in _dtypes for t in expected))):
def _case_five_transform(t):
return make_tensor(t.shape, dtype=torch.long, device=t.device)
out = _apply_out_transform(_case_five_transform, expected)
msg_fail = "" if not isinstance(expected, torch.Tensor) else \
("Expected RuntimeError when doing an unsafe cast from a result of dtype "
f"{expected.dtype} into an out= with dtype torch.long")
with self.assertRaises(RuntimeError, msg=msg_fail):
op_out(out=out)
# Tests that the forward and backward passes of operations produce the
# same values for the cross-product of op variants (method, inplace)
# against eager's gold standard op function variant
@_variant_ops(op_db)
def test_variant_consistency_eager(self, device, dtype, op):
# Acquires variants (method variant, inplace variant, aliases)
method = op.method_variant
inplace = op.inplace_variant
# list of all inplace ops: inplace variant + alias inplace variants if exist
inplace_ops = [inplace, ]
variants = [method, inplace]
for a_op in op.aliases:
variants.append(a_op.op)
variants.append(a_op.method_variant)
variants.append(a_op.inplace_variant)
inplace_ops.append(a_op.inplace_variant)
inplace_variants = tuple(filter(None, inplace_ops))
variants = tuple(filter(None, variants))
_requires_grad = (op.supports_autograd and
(dtype.is_floating_point or op.supports_complex_autograd(torch.device(device).type)))
include_conjugated_inputs = op.test_conjugated_samples and dtype.is_complex
samples = op.sample_inputs(device, dtype, requires_grad=_requires_grad, include_conjugated_inputs=include_conjugated_inputs)
def _test_consistency_helper(samples, variants):
for sample in samples:
# TODO: Check grad for all Tensors requiring grad if sample.input is TensorList
tensor = sample.input if isinstance(sample.input, torch.Tensor) else sample.input[0]
# Computes function forward and backward values
tensor.grad = None
expected_forward = op(sample.input, *sample.args, **sample.kwargs)
expected_grad = None
output_process_fn_grad = sample.output_process_fn_grad if sample.output_process_fn_grad \
else lambda x: x
# Skips inplace variants if the output dtype is not the same as
# the input dtype
skip_inplace = False
if (isinstance(expected_forward, torch.Tensor) and
expected_forward.dtype is not tensor.dtype):
skip_inplace = True
# TODO: backward consistency only supported for single tensor outputs
# TODO: backward consistency only checked on sample.input, not all
# tensor inputs
# TODO: update to handle checking grads of all tensor inputs as
# derived from each tensor output
if (op.supports_autograd and isinstance(expected_forward, torch.Tensor)
and (dtype.is_floating_point or op.supports_complex_autograd(torch.device(device).type))):
output_process_fn_grad(expected_forward).sum().backward()
expected_grad = tensor.grad
# Test eager consistency
for variant in variants:
# Skips inplace ops
if variant in inplace_ops and skip_inplace:
continue
# Compares variant's forward
# Note: copies the to-be-modified input when testing the inplace variant
tensor.grad = None
cloned = clone_input_helper(sample.input) if variant in inplace_ops else sample.input
if variant in inplace_ops and sample.broadcasts_input:
with self.assertRaises(RuntimeError,
msg=('inplace variant either incorrectly allowed '
'resizing or you have marked the sample {}'
' incorrectly with `broadcasts_self=True'.format(sample.summary()))):
variant_forward = variant(cloned,
*sample.args,
**sample.kwargs)
continue
variant_forward = variant(cloned,
*sample.args,
**sample.kwargs)
self.assertEqual(expected_forward, variant_forward)
# Compares variant's backward
if expected_grad is not None and \
(variant not in inplace_ops or op.supports_inplace_autograd):
output_process_fn_grad(variant_forward).sum().backward()
self.assertEqual(expected_grad, tensor.grad)
_test_consistency_helper(samples, variants)
def _test_inplace_preserve_storage(samples, variants):
for sample in samples:
# Skips inplace variants if the output dtype is not the same as
# the input dtype
expected_forward = op(sample.input, *sample.args, **sample.kwargs)
tensor = sample.input if isinstance(sample.input, torch.Tensor) else sample.input[0]
skip_inplace = False
if (isinstance(expected_forward, torch.Tensor) and
expected_forward.dtype is not tensor.dtype):
skip_inplace = True
if skip_inplace:
return
for variant in variants:
cloned = clone_input_helper(sample.input) if variant in inplace_ops else sample.input
inp_tensor = cloned if isinstance(cloned, torch.Tensor) else cloned[0]
data_ptr = inp_tensor.data_ptr()
variant_forward = variant(cloned,
*sample.args,
**sample.kwargs)
# TODO Support non-tensor outputs if they exist for inplace ops
if (isinstance(variant_forward, torch.Tensor)):
self.assertEqual(data_ptr, variant_forward.data_ptr(), atol=0, rtol=0)
else:
self.assertTrue(False, "Non-tensor outputs for inplace ops are not supported")
if len(inplace_ops) > 0:
inplace_samples = list(filter(lambda sample: not sample.broadcasts_input, samples))
_test_inplace_preserve_storage(inplace_samples, inplace_variants)
# Checks if the operator (if it is composite) is written to support most
# backends and Tensor subclasses. See "CompositeImplicitAutograd Compliance"
# in aten/src/ATen/native/README.md for more details
#
# NB: onlyCPU because CompositeImplicitAutograd ops go through the same
# codepath on all devices. Ideally we'd use a meta device here but coverage
# for that is not good yet.
@unittest.skipIf(IS_FBCODE or IS_SANDCASTLE, '__torch_dispatch__ does not work in fbcode')
@onlyCPU
@ops(op_db, allowed_dtypes=(torch.float,))
def test_composite_compliance(self, device, dtype, op):
samples = op.sample_inputs(device, dtype, requires_grad=False)
for sample in samples:
args = [sample.input] + list(sample.args)
kwargs = sample.kwargs
_check_composite_compliance(op, args, kwargs)
# gradcheck requires double precision
_gradcheck_ops = partial(ops, dtypes=OpDTypes.supported,
allowed_dtypes=[torch.double, torch.cdouble])
class TestGradients(TestCase):
exact_dtype = True
# Copies inputs to inplace operations to avoid inplace modifications
# to leaves requiring gradient
def _get_safe_inplace(self, inplace_variant):
@wraps(inplace_variant)
def _fn(t, *args, **kwargs):
return inplace_variant(t.clone(), *args, **kwargs)
return _fn
def _check_helper(self, device, dtype, op, variant, check, *, check_forward_ad=False, check_backward_ad=True,
check_undefined_grad=True, check_batched_grad=None, check_batched_forward_grad=False):
# NB: check_backward_ad does not affect gradgradcheck (always True)
if variant is None:
self.skipTest("Skipped! Variant not implemented.")
if not op.supports_dtype(dtype, torch.device(device).type):
self.skipTest(f"Skipped! {op.name} does not support dtype {str(dtype)}")
def is_inplace(variant):
if hasattr(variant, "__wrapped__"):
return variant.__wrapped__ is op.get_inplace()
return variant is op.get_inplace()
include_conjugated_inputs = op.test_conjugated_samples and dtype.is_complex
samples = op.sample_inputs(device, dtype, requires_grad=True, include_conjugated_inputs=include_conjugated_inputs)
for sample in samples:
if sample.broadcasts_input and is_inplace(variant):
continue
# Note on TensorList inputs
#
# gradcheck does not support TensorList inputs so here we pass TensorList
# inputs of size n as n single Tensor inputs to gradcheck and wrap the op
# in a function that puts the n Tensor inputs back into a TensorList
def fn(*inputs):
# Put tensors back into TensorList since we splat them when passing to gradcheck
if is_iterable_of_tensors(sample.input):
n = len(sample.input)
inputs = (inputs[:n], *inputs[n:])
output = op.gradcheck_wrapper(variant, *inputs, **sample.kwargs)
if sample.output_process_fn_grad is not None:
return sample.output_process_fn_grad(output)
return output
# Splat TensorList inputs into single Tensor inputs
gradcheck_args = (sample.input,) if isinstance(sample.input, torch.Tensor) else tuple(sample.input)
gradcheck_args += sample.args
if check == 'gradcheck':
if check_batched_grad is None:
check_batched_grad = op.check_batched_grad
self.assertTrue(gradcheck(fn, gradcheck_args,
check_batched_grad=check_batched_grad,
check_grad_dtypes=True,
nondet_tol=op.gradcheck_nondet_tol,
fast_mode=op.gradcheck_fast_mode,
check_forward_ad=check_forward_ad,
check_backward_ad=check_backward_ad,
check_undefined_grad=check_undefined_grad,
check_batched_forward_grad=check_batched_forward_grad))
elif check == 'gradgradcheck':
self.assertFalse(check_forward_ad, msg="Cannot run forward AD check for gradgradcheck")
self.assertTrue(gradgradcheck(fn, gradcheck_args,
gen_non_contig_grad_outputs=False,
check_batched_grad=op.check_batched_gradgrad,
check_grad_dtypes=True,
nondet_tol=op.gradcheck_nondet_tol,
fast_mode=op.gradcheck_fast_mode))
self.assertTrue(gradgradcheck(fn, gradcheck_args,
gen_non_contig_grad_outputs=True,
check_batched_grad=op.check_batched_gradgrad,
check_grad_dtypes=True,
nondet_tol=op.gradcheck_nondet_tol,
fast_mode=op.gradcheck_fast_mode))
else:
self.assertTrue(False, msg="Unknown check requested!")
def _grad_test_helper(self, device, dtype, op, variant, *, check_forward_ad=False, check_backward_ad=True,
check_undefined_grad=True, check_batched_grad=None, check_batched_forward_grad=False):
return self._check_helper(device, dtype, op, variant, 'gradcheck', check_forward_ad=check_forward_ad,
check_backward_ad=check_backward_ad, check_undefined_grad=check_undefined_grad,
check_batched_grad=check_batched_grad,
check_batched_forward_grad=check_batched_forward_grad)
def _gradgrad_test_helper(self, device, dtype, op, variant):
return self._check_helper(device, dtype, op, variant, 'gradgradcheck')
def _skip_helper(self, op, device, dtype):
if not op.supports_autograd and not op.supports_forward_ad:
self.skipTest("Skipped! autograd not supported.")
if not op.supports_complex_autograd(torch.device(device).type) and dtype.is_complex:
self.skipTest("Skipped! Complex autograd not supported.")
# Tests that gradients are computed correctly
@_gradcheck_ops(op_db)
def test_fn_grad(self, device, dtype, op):
self._skip_helper(op, device, dtype)
self._grad_test_helper(device, dtype, op, op.get_op())
# Method grad (and gradgrad, see below) tests are disabled since they're
# costly and redundant with function grad (and gradgad) tests
# @_gradcheck_ops(op_db)
# def test_method_grad(self, device, dtype, op):
# self._skip_helper(op, device, dtype)
# self._grad_test_helper(device, dtype, op, op.get_method())
@_gradcheck_ops(op_db)
def test_inplace_grad(self, device, dtype, op):
self._skip_helper(op, device, dtype)
if not op.inplace_variant or not op.supports_inplace_autograd:
self.skipTest("Skipped! Operation does not support inplace autograd.")
self._grad_test_helper(device, dtype, op, self._get_safe_inplace(op.get_inplace()))
# Test that gradients of gradients are computed correctly
@_gradcheck_ops(op_db)
def test_fn_gradgrad(self, device, dtype, op):
self._skip_helper(op, device, dtype)
if not op.supports_gradgrad:
self.skipTest("Skipped! Operation does not support gradgrad")
self._gradgrad_test_helper(device, dtype, op, op.get_op())
# Test that gradients of gradients are properly raising
@_gradcheck_ops(op_db)
def test_fn_fail_gradgrad(self, device, dtype, op):
self._skip_helper(op, device, dtype)
if op.supports_gradgrad:
self.skipTest("Skipped! Operation does support gradgrad")
err_msg = r"derivative for .* is not implemented"
with self.assertRaisesRegex(RuntimeError, err_msg):
self._gradgrad_test_helper(device, dtype, op, op.get_op())
# Method gradgrad (and grad, see above) tests are disabled since they're
# costly and redundant with function gradgrad (and grad) tests
# @_gradcheck_ops(op_db)
# def test_method_gradgrad(self, device, dtype, op):
# self._skip_helper(op, device, dtype)
# self._gradgrad_test_helper(device, dtype, op, op.get_method())
@_gradcheck_ops(op_db)
def test_inplace_gradgrad(self, device, dtype, op):
self._skip_helper(op, device, dtype)
if not op.inplace_variant or not op.supports_inplace_autograd:
self.skipTest("Skipped! Operation does not support inplace autograd.")
self._gradgrad_test_helper(device, dtype, op, self._get_safe_inplace(op.get_inplace()))
def _forward_grad_helper(self, device, dtype, op, variant):
# TODO: clean up how attributes are passed to gradcheck from OpInfos
def call_grad_test_helper():
self._grad_test_helper(device, dtype, op, variant, check_forward_ad=True, check_backward_ad=False,
check_undefined_grad=False, check_batched_grad=False,
check_batched_forward_grad=op.check_batched_forward_grad)
if op.supports_forward_ad:
call_grad_test_helper()
else:
err_msg = r"Trying to use forward AD with .* that does not support it\."
hint_msg = ("Running forward AD for an OP that has does not support it did not "
"raise any error. If your op supports forward AD, you should set supports_forward_ad=True")
with self.assertRaisesRegex(NotImplementedError, err_msg, msg=hint_msg):
call_grad_test_helper()
@_gradcheck_ops(op_db)
def test_forward_mode_AD(self, device, dtype, op):
self._skip_helper(op, device, dtype)
self._forward_grad_helper(device, dtype, op, op.get_op())
@_gradcheck_ops(op_db)
def test_inplace_forward_mode_AD(self, device, dtype, op):
self._skip_helper(op, device, dtype)
if not op.inplace_variant or not op.supports_inplace_autograd:
self.skipTest("Skipped! Operation does not support inplace autograd.")
self._forward_grad_helper(device, dtype, op, self._get_safe_inplace(op.get_inplace()))
# Functions that do not support autograd should not fail in forward mode
# Inplace functions (such as "resize_") are expected to fail in forward mode and should be skipped
# Test only when supports_autograd=False and for double dtype
@ops(filter(lambda op: not op.supports_autograd, op_db), dtypes=OpDTypes.supported, allowed_dtypes=(torch.double,))
def test_nondifferentiable(self, device, dtype, op):
# Expecting no errors
samples = op.sample_inputs(device, dtype, requires_grad=True)
sample = samples[0]
result = op(sample.input, *sample.args, **sample.kwargs)
# types.LambdaType gave false positives
def is_lambda(lamb):
LAMBDA = lambda: 0 # noqa: E731
return isinstance(lamb, type(LAMBDA)) and lamb.__name__ == LAMBDA.__name__
# Tests operators for consistency between JIT and eager, also checks
# correctness of JIT specific alias schemas and intended
# autodifferentiation behavior.
# Inherits from JitCommonTestCase instead of TestCase directly to share
# functionality with original test_jit.py method operator tests
class TestJit(JitCommonTestCase):
exact_dtype = True
# Tests that the forward and backward passes of operations produce the
# same values for the cross-product of op variants (function, method, inplace)
# and runtimes (eager, traced, scripted).
# TODO WARNING: inplace x {traced, scripted} not currently tested
@_variant_ops(op_db)
def test_variant_consistency_jit(self, device, dtype, op):
_requires_grad = op.supports_autograd and (dtype.is_floating_point or
op.supports_complex_autograd(torch.device(device).type))
include_conjugated_inputs = op.test_conjugated_samples and dtype.is_complex
samples = op.sample_inputs(device, dtype, requires_grad=_requires_grad, include_conjugated_inputs=include_conjugated_inputs)
# Acquires variants to test
func = op.get_op()
method = op.get_method()
variants = {
# TODO: inplace tests currently fail, fix and add inplace variant
'function': func, 'method': method,
}
# TODO: find better way to standardize on op registration itself..
has_fake_function = op.name in ["resize_", 'resize_as_']
if has_fake_function:
variants = {'method': getattr(torch.Tensor, op.name)}
samples = op.sample_inputs(device, dtype, requires_grad=False)
support_script = op.supports_scripting
tested = False
for sample in samples:
# Test traced and scripted consistency
for func_type, variant in variants.items():
if variant is None:
continue
# scripting and check_alias_analysis do not work with lambdas
# lambdas are typically used as a way to simulate methods without
# functional variants, so rely on the other variant for testing
# for now
if is_lambda(variant):
continue
tested = True
# Create accessor for script function variant
name = op.name + '_' if func_type == 'inplace' else op.name
# run with disable_autodiff_subgraph_inlining(True) to test
# autodiff support. Context manager forces the graph to contain
# DifferentiableGraph nodes if they are present
with disable_autodiff_subgraph_inlining():
# Check scripted forward, grad, and grad grad
if support_script:
script_fn = create_script_fn(self, name, func_type)
def out_fn(output):
# Processes the output for autograd
if sample.output_process_fn_grad is not None:
return sample.output_process_fn_grad(output)
return output
def get_sample():
return clone_input_helper(sample.input) if op.name[-1] == '_' else sample.input
if support_script:
check_against_reference(self,
script_fn,
func,
out_fn,
(get_sample(),) + sample.args,
sample.kwargs,
no_grad=not _requires_grad, no_gradgrad=not op.supports_gradgrad)
# Check traced forward, grad, and grad grad
# TODO: fix tracing here
supports_tracing = not has_fake_function
if op.assert_jit_shape_analysis:
self.assertTrue(supports_tracing)
if supports_tracing:
traced_fn = create_traced_fn(self, variant)
check_against_reference(self,
traced_fn,
func,
out_fn,
(get_sample(),) + sample.args,
sample.kwargs,
no_grad=not _requires_grad, no_gradgrad=not op.supports_gradgrad)
# Check alias annotation schema for correctness (make
# sure inputs that aren't supposed to be modified aren't)
# Note: only runs in float32 because schema isn't affected by dtype,
# so running it on all dtypes is would be excessive
if dtype == torch.float32:
# TODO: no reason why we cant run this with tracing graph
if support_script and op.name != "rsub":
check_alias_annotation(name, (get_sample(),) + sample.args, sample.kwargs,
func_type=func_type, aten_name=op.aten_name)
# TODO: use script graph as well
checked_shape_analysis = False
if supports_tracing:
out = variant(get_sample(), *sample.args, **sample.kwargs)
# right now, tuple of outputs and tensor output supported
# TODO: list of tensor outputs
tuple_of_tensors = isinstance(out, tuple) and all([isinstance(elem, torch.Tensor) for elem in out])
if isinstance(out, torch.Tensor) or tuple_of_tensors:
if tuple_of_tensors:
sizes = [elem.size() for elem in out]
else:
sizes = out.size()
self.checkShapeAnalysis(sizes, traced_fn.graph, op.assert_jit_shape_analysis)
checked_shape_analysis = True
if op.assert_jit_shape_analysis:
self.assertTrue(checked_shape_analysis)
# Check autodifferentiation of nodes for traced and scripted graphs, only need to check once per sample
if dtype is torch.float32:
# Sandcastle doesn't fuse nodes
if IS_SANDCASTLE:
# fusible nodes are expected to be found in FusionGroups in the DifferentiableGraphs
nonfusible_nodes = op.autodiff_nonfusible_nodes + op.autodiff_fusible_nodes
fusible_nodes = []
else:
nonfusible_nodes = op.autodiff_nonfusible_nodes
fusible_nodes = op.autodiff_fusible_nodes
if supports_tracing:
self.assertAutodiffNode(traced_fn.last_graph, op.assert_autodiffed, nonfusible_nodes, fusible_nodes)
if support_script:
self.assertAutodiffNode(script_fn.last_graph, op.assert_autodiffed, nonfusible_nodes, fusible_nodes)
assert tested, "JIT Test does not execute any logic"
# alias testing is only done with torch.float for the same reason
_alias_ops = partial(ops, dtypes=OpDTypes.supported,
allowed_dtypes=(torch.float,))
@_alias_ops((op for op in op_db if op.aliases))
def test_jit_alias_remapping(self, device, dtype, op):
# Required to avoid undefined value: tensor error in JIT compilation of the function template
tensor = torch.tensor
samples = op.sample_inputs(device, dtype, requires_grad=True)
if len(samples) == 0:
self.skipTest("Skipped! No sample inputs!")
# NOTE: only tests on first sample
sample = samples[0]
# [Scripting Data Preparation]
# Prepare data for test scripting
# Below we prepare strings of args/kwargs with and without type annotations.
# These strings are inserted into function template strings which is then torch scripted.
# - args string is ["t0"] corresponding to the "input" tensor required by the op
# - args_kw is the value of args and strings of kwargs used to call the op (without type annotations), for example,
# ["to", "1.0", "(1,)", "True", "tensor(1.0)"] -> def fn(t0): return variant(t0, 1.0, (1,), True, tensor(1.0))
args = ["t0"]
def quote_strs(v):
if isinstance(v, str):
return f"'{v}'"
return str(v)
args_kw = args + \
[f"{v}" for v in sample.args] + \
[f"{k}={quote_strs(v)}" for k, v in sample.kwargs.items()]
# Prepare data for test tracing
sample_args_kwargs = ()
if len(sample.args) > 0:
sample_args_kwargs += (sample.args, )
if len(sample.kwargs) > 0:
sample_args_kwargs += (sample.kwargs, )
original_name = op.aten_name
original_name_inplace = original_name + "_"
expected_dtype = op(sample.input, *sample.args, **sample.kwargs).dtype
for a_op in op.aliases:
inplace = a_op.inplace_variant
method_or_inplace = [a_op.inplace_variant, a_op.method_variant]
variants = (v for v in (a_op.op, a_op.method_variant, a_op.inplace_variant) if v is not None)
# Test scripting:
for variant in variants:
variant_name = variant.__name__
op_name = original_name_inplace if variant is inplace else original_name
if variant in method_or_inplace:
fn_template = '''
def _fn(t0{c}):
return t0.{alias_name}({args_kw})
'''
# remove the first input tensor
script = fn_template.format(
c=", " if len(args_kw[1:]) > 1 else "",
args_kw=", ".join(args_kw[1:]),
alias_name=variant_name,
)
else:
fn_template = '''
def _fn({args}):
return variant({args_kw})
'''
script = fn_template.format(
args=", ".join(args),
args_kw=", ".join(args_kw),
)
scripted = torch.jit.CompilationUnit(script)._fn
if (variant is inplace and not torch.can_cast(expected_dtype, dtype)):
try:
inp = clone_input_helper(sample.input)
scripted(inp)
except Exception as e:
continue
self.fail("Inplace operation on integer tensor that should be promoted to float didn't fail!")
inp = clone_input_helper(sample.input)
scripted(inp)
inp = clone_input_helper(sample.input)
graph = scripted.graph_for(inp)
FileCheck().check(op.aten_name).check_not(variant_name).run(graph)
# Test tracing:
for variant in variants:
variant_name = variant.__name__
op_name = original_name_inplace if variant is inplace else original_name
def _fn(*sample_args, **sample_kwargs):
return variant(*sample_args, **sample_kwargs)
inp = (clone_input_helper(sample.input),) + sample_args_kwargs
traced = torch.jit.trace(_fn, *inp)
inp = (clone_input_helper(sample.input),) + sample_args_kwargs
traced(*inp)
inp = (clone_input_helper(sample.input),) + sample_args_kwargs
graph = traced.graph_for(*inp)
FileCheck().check(op_name).check_not(variant_name).run(graph)
class TestMathBits(TestCase):
# Tests that
# 1. The operator's output for physically conjugated/negated tensors and conjugate/negative view tensors
# produces the same value
# 2. The gradients are same in both cases mentioned in (1)
# 3. If the operator's inplace variant is supported, tests that the inplace operation
# produces the correct value when called on a conjugate/negative view tensor and that the output
# has its conj/neg bit set to true
# This test only runs for C -> R and C -> C functions
# TODO: add tests for `R->C` functions
# Note: This test runs for functions that take both tensors and tensorlists as input.
def _test_math_view(self, device, dtype, op, _requires_grad, math_op_physical, math_op_view, is_bit_set, out_type):
samples = op.sample_inputs(device, dtype, requires_grad=_requires_grad)
inplace_variant = op.inplace_variant
# helper function to physically conjugate/negate the tensor
def math_physical(input):
if isinstance(input, torch.Tensor):
tensor_requires_grad = input.requires_grad
with torch.no_grad():
input = math_op_physical(input)
return input.requires_grad_(tensor_requires_grad)
if isinstance(input, Sequence):
out = list(map(clone_input_helper, input))
out[0] = math_physical(out[0])
return tuple(out)
# helper function to clone and conjugate/negate the input if its a tensor
# else clone the sequence and conjugate/negate the first element in the sequence
# If a requires_grad argument is provided the tensor being conjugated/negated will
# have its requires_grad set to that value.
def clone_and_perform_view(input, **kwargs):
if isinstance(input, torch.Tensor):
requires_grad = kwargs.get('requires_grad', input.requires_grad)
with torch.no_grad():
input = input.clone()
# Note: .conj() is not called under no_grad mode since it's not allowed to modify a
# view created in no_grad mode. Here it's ok to do so, so as a workaround we call conj
# before resetting the requires_grad field for input
input = math_op_view(input)
assert input.is_leaf
return input.requires_grad_(requires_grad)
if isinstance(input, Sequence):
out = list(map(clone_input_helper, input))
out[0] = clone_and_perform_view(out[0])
return tuple(out)
for sample in samples:
tensor = sample.input if isinstance(sample.input, torch.Tensor) else sample.input[0]
cloned1 = clone_and_perform_view(sample.input)
sample.input = math_physical(sample.input)
# Computes function forward value with a physically conjugated/negated tensor and
# a conj/neg view tensor and verifies that the output in both case are equal.
expected_forward = op(sample.input, *sample.args, **sample.kwargs)
forward_with_mathview = op(cloned1, *sample.args, **sample.kwargs)
self.assertEqual(expected_forward, forward_with_mathview)
# If the op has an inplace variant, and the input doesn't require broadcasting
# and has the same dtype as output, verify that the inplace operation on a conjugated/negated
# input produces correct output, and the output tensor has the conj/neg bit set to True
if inplace_variant is not None and not sample.broadcasts_input:
cloned2 = clone_and_perform_view(tensor, requires_grad=False)
if (isinstance(expected_forward, torch.Tensor) and
expected_forward.dtype is tensor.dtype):
inplace_forward = inplace_variant(cloned2, *sample.args, **sample.kwargs)
self.assertTrue(is_bit_set(inplace_forward))
self.assertEqual(inplace_forward, expected_forward)
# TODO: backward consistency only supported for single tensor outputs
# TODO: backward consistency only checked on sample.input, not all
# tensor inputs
# TODO: update to handle checking grads of all tensor inputs as
# derived from each tensor output
if isinstance(expected_forward, torch.Tensor) and expected_forward.requires_grad:
tensor = sample.input if isinstance(sample.input, torch.Tensor) else sample.input[0]
expected_forward.sum().backward(retain_graph=True)
forward_with_mathview.sum().backward(retain_graph=True)
if tensor.grad is not None:
cloned1_tensor = cloned1 if isinstance(cloned1, torch.Tensor) else cloned1[0]
self.assertEqual(tensor.grad, cloned1_tensor.grad)
tensor.grad, cloned1_tensor.grad = None, None
# a repeat of the above test if output is not complex valued
if (out_type(expected_forward)):
grad = torch.randn_like(expected_forward)
expected_forward.backward(math_op_physical(grad))
forward_with_mathview.backward(math_op_view(grad))
self.assertEqual(tensor.grad, cloned1_tensor.grad)
@ops(op_db, allowed_dtypes=(torch.cfloat,))
def test_conj_view(self, device, dtype, op):
if not op.test_conjugated_samples:
self.skipTest("Operation doesn't support conjugated inputs.")
math_op_physical = torch.conj_physical
math_op_view = torch.conj
_requires_grad = (op.supports_autograd and op.supports_complex_autograd(torch.device(device).type))
is_bit_set = torch.is_conj
self._test_math_view(device, dtype, op, _requires_grad, math_op_physical, math_op_view, is_bit_set, torch.is_complex)
@ops(op_db, allowed_dtypes=(torch.double,))
def test_neg_view(self, device, dtype, op):
if not op.test_neg_view:
self.skipTest("Operation not tested with tensors with negative bit.")
# The view op here is an identity, but math_op_physical's output is
# modified inplace, so we must at least clone
math_op_physical = torch.clone
def math_op_view(x):
return torch.conj(x * -1j).imag
_requires_grad = (op.supports_autograd and op.supports_complex_autograd(torch.device(device).type))
is_bit_set = torch.is_neg
self._test_math_view(device, dtype, op, _requires_grad, math_op_physical, math_op_view, is_bit_set,
lambda x: not torch.is_complex(x))
instantiate_device_type_tests(TestCommon, globals())
instantiate_device_type_tests(TestGradients, globals())
instantiate_device_type_tests(TestJit, globals())
instantiate_device_type_tests(TestMathBits, globals())
if __name__ == '__main__':
run_tests()
Reference in New Issue View Git Blame Copy Permalink