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

Revert "[BE] wrap deprecated function/class with typing_extensions.deprecated (#126898)"

Browse Source 
This reverts commit 749a132fb0.

Reverted https://github.com/pytorch/pytorch/pull/126898 on behalf of https://github.com/fbgheith due to switching typing-extensions=4.3.0 to 4.9.0 causes internal failure ([comment](https://github.com/pytorch/pytorch/pull/126898#issuecomment-2142884456))
This commit is contained in:
PyTorch MergeBot 2024-05-31 19:47:24 +00:00
parent ea13e9a097
commit 033e733021
117 changed files with 478 additions and 700 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
.github/requirements/conda-env-Linux-X64.txt vendored
View File

@ -6,4 +6,4 @@ numpy=1.23.3
pyyaml=6.0
requests=2.31.0
setuptools=68.2.2
typing-extensions=4.9.0
typing-extensions=4.3.0

2
.github/requirements/conda-env-iOS.txt vendored
View File

@ -5,4 +5,4 @@ numpy=1.23.3
pyyaml=6.0
requests=2.31.0
setuptools=68.2.2
typing-extensions=4.9.0
typing-extensions=4.3.0

2
.github/requirements/conda-env-macOS-ARM64 vendored
View File

@ -2,7 +2,7 @@ numpy=1.22.3
pyyaml=6.0
setuptools=61.2.0
cmake=3.22.*
typing-extensions=4.9.0
typing-extensions=4.3.0
dataclasses=0.8
pip=22.2.2
pillow=10.0.1

2
.github/requirements/conda-env-macOS-X64 vendored
View File

@ -4,7 +4,7 @@ numpy=1.21.2
pyyaml=5.3
setuptools=46.0.0
cmake=3.22.*
typing-extensions=4.9.0
typing-extensions=4.3.0
dataclasses=0.8
pip=22.2.2
pillow=10.0.1

2
test/distributed/_tensor/test_api.py
View File

@ -237,7 +237,7 @@ class DTensorAPITest(DTensorTestBase):
assert isinstance(outputs, DTensor)
return outputs.to_local()
with self.assertWarnsRegex(FutureWarning, "Deprecating"):
with self.assertWarnsRegex(UserWarning, "Deprecating"):
replica_module = distribute_module(
module_to_replicate,
device_mesh,

2
test/distributed/fsdp/test_fsdp_optim_state.py
View File

@ -1436,7 +1436,7 @@ class TestFSDPOptimState(FSDPTest):
def get_warning_context():
warning_regex = "`optim_input` argument is deprecated"
return self.assertWarnsRegex(
expected_warning=FutureWarning, expected_regex=warning_regex
expected_warning=UserWarning, expected_regex=warning_regex
)
self._run_on_all_optim_state_apis(

6
test/functorch/test_eager_transforms.py
View File

@ -3258,7 +3258,7 @@ class TestComposability(TestCase):
x = torch.randn(3, device=device)
# functorch version of the API is deprecated
with self.assertWarnsRegex(FutureWarning, "Please use `torch.vmap`"):
with self.assertWarnsRegex(UserWarning, "Please use torch.vmap"):
vmap(torch.sin)
# the non-functorch version is not deprecated
@ -3276,9 +3276,7 @@ class TestComposability(TestCase):
new_api = getattr(torch.func, transform)
# functorch version of the API is deprecated
with self.assertWarnsRegex(
FutureWarning, f"Please use `torch.func.{transform}`"
):
with self.assertWarnsRegex(UserWarning, f"Please use torch.func.{transform}"):
api(torch.sin)
# the non-functorch version is not deprecated

2
test/nn/test_init.py
View File

@ -521,7 +521,7 @@ class TestNNInit(TestCase):
init.normal(x)
with self.assertWarnsRegex(
FutureWarning,
UserWarning,
"deprecated",
msg="methods not suffixed with underscore should be deprecated",
):

9
test/nn/test_module_hooks.py
View File

@ -1387,8 +1387,7 @@ class TestModuleHookNN(NNTestCase):
m.register_backward_hook(noop)
with self.assertWarnsRegex(
FutureWarning,
"does not take as input a single Tensor or a tuple of Tensors",
UserWarning, "does not take as input a single Tensor or a tuple of Tensors"
):
m([a, b])
@ -1401,7 +1400,7 @@ class TestModuleHookNN(NNTestCase):
m.register_backward_hook(noop)
with self.assertWarnsRegex(
FutureWarning, "does not return a single Tensor or a tuple of Tensors"
UserWarning, "does not return a single Tensor or a tuple of Tensors"
):
m(a, b)
@ -1414,7 +1413,7 @@ class TestModuleHookNN(NNTestCase):
m.register_backward_hook(noop)
with self.assertWarnsRegex(
FutureWarning, "outputs are generated by different autograd Nodes"
UserWarning, "outputs are generated by different autograd Nodes"
):
m(a, b)
@ -1427,7 +1426,7 @@ class TestModuleHookNN(NNTestCase):
m.register_backward_hook(noop)
with self.assertWarnsRegex(
FutureWarning, "the forward contains multiple autograd Nodes"
UserWarning, "the forward contains multiple autograd Nodes"
):
m(a)

4
test/test_autocast.py
View File

@ -255,8 +255,8 @@ class TestAutocastCPU(TestCase):
def test_cpu_autocast_deprecated_warning(self):
with self.assertWarnsRegex(
FutureWarning,
r"`torch.cpu.amp.autocast\(args...\)` is deprecated. Please use `torch.amp.autocast\('cpu', args...\)` instead.",
DeprecationWarning,
r"torch.cpu.amp.autocast\(args...\) is deprecated. Please use torch.amp.autocast\('cpu', args...\) instead.",
):
with torch.cpu.amp.autocast():
_ = torch.ones(10)

10
test/test_autograd.py
View File

@ -154,7 +154,7 @@ class TestAutograd(TestCase):
def test_grad_mode_class_decoration(self):
# Decorating class is deprecated and should not be used
with self.assertWarnsRegex(FutureWarning, "Decorating classes is deprecated"):
with self.assertWarnsRegex(UserWarning, "Decorating classes is deprecated"):
@torch.no_grad()
class Foo:
@ -5937,13 +5937,13 @@ Done""",
b = torch.rand(2, 2, requires_grad=True, dtype=torch.float64)
with self.assertWarnsRegex(
FutureWarning, "`get_numerical_jacobian` was part of PyTorch's private API"
UserWarning, "get_numerical_jacobian was part of PyTorch's private API"
):
jacobian = get_numerical_jacobian(fn, (a, b), target=a, eps=1e-6)
self.assertEqual(jacobian[0], 2 * torch.eye(4, dtype=torch.double))
with self.assertWarnsRegex(
FutureWarning, "`get_numerical_jacobian` was part of PyTorch's private API"
UserWarning, "get_numerical_jacobian was part of PyTorch's private API"
):
jacobian = get_numerical_jacobian(fn, (a, b), eps=1e-6)
self.assertEqual(jacobian[0], 2 * torch.eye(4, dtype=torch.double))
@ -5963,7 +5963,7 @@ Done""",
outputs = fn(a, b)
with self.assertWarnsRegex(
FutureWarning, "`get_analytical_jacobian` was part of PyTorch's private API"
UserWarning, "get_analytical_jacobian was part of PyTorch's private API"
):
(
jacobians,
@ -5991,7 +5991,7 @@ Done""",
outputs = NonDetFunc.apply(a, 1e-6)
with self.assertWarnsRegex(
FutureWarning, "`get_analytical_jacobian` was part of PyTorch's private API"
UserWarning, "get_analytical_jacobian was part of PyTorch's private API"
):
(
jacobians,

8
test/test_cuda.py
View File

@ -1820,10 +1820,10 @@ torch.cuda.synchronize()
return grad, grad
self.assertRegex(
str(w[0].message), r"`torch.cuda.amp.custom_fwd\(args...\)` is deprecated."
str(w[0].message), r"torch.cuda.amp.custom_fwd\(args...\) is deprecated."
)
self.assertRegex(
str(w[1].message), r"`torch.cuda.amp.custom_bwd\(args...\)` is deprecated."
str(w[1].message), r"torch.cuda.amp.custom_bwd\(args...\) is deprecated."
)
mymm = MyMM.apply
@ -2016,8 +2016,8 @@ torch.cuda.synchronize()
def test_cuda_autocast_deprecated_warning(self):
with self.assertWarnsRegex(
FutureWarning,
r"`torch.cuda.amp.autocast\(args...\)` is deprecated. Please use `torch.amp.autocast\('cuda', args...\)` instead.",
DeprecationWarning,
r"torch.cuda.amp.autocast\(args...\) is deprecated. Please use torch.amp.autocast\('cuda', args...\) instead.",
):
with torch.cuda.amp.autocast():
_ = torch.ones(10)

2
test/test_optim.py
View File

@ -287,7 +287,7 @@ class TestOptimRenewed(TestCase):
inpt = torch.randn(5, device=device, dtype=dtype)
# avoid endless recompiles by wrapping LR in a tensor if we're compiling
lr = torch.tensor(0.01) if torch.compiler.is_compiling() else 0.01
lr = torch.tensor(0.01) if torch._utils.is_compiling() else 0.01
optimizer = optim_cls([{"params": [weight]}, {"params": [bias], "lr": lr}])
schedulers = [scheduler_c(optimizer) for scheduler_c in schedulers_c]

2
test/test_prims.py
View File

@ -338,7 +338,7 @@ $1: f32[2] = torch._ops.prims.sin.default($0)""")
prims.mul(torch.randn(2), 1 + 1j)
def test_check_deprecation_warning(self):
with self.assertWarnsRegex(FutureWarning, 'will be removed in the future'):
with self.assertWarnsRegex(DeprecationWarning, 'will be removed in the future'):
torch._prims_common.check(True, lambda: 'message')

2
test/test_pytree.py
View File

@ -723,7 +723,7 @@ class TestPythonPytree(TestCase):
self.y = y
with self.assertWarnsRegex(
FutureWarning, "torch.utils._pytree._register_pytree_node"
UserWarning, "torch.utils._pytree._register_pytree_node"
):
py_pytree._register_pytree_node(
DummyType,

2
test/test_stateless.py
View File

@ -901,7 +901,7 @@ exit(len(w))
m = torch.nn.Linear(1, 1)
params = dict(m.named_parameters())
x = torch.randn(3, 1)
with self.assertWarnsRegex(FutureWarning, "Please use `torch.func.functional_call`"):
with self.assertWarnsRegex(UserWarning, "Please use torch.func.functional_call"):
stateless.functional_call(m, params, x)
class TestPythonOptimizeMode(TestCase):

4
test/test_torch.py
View File

@ -6198,8 +6198,8 @@ else:
GradScaler = torch.cuda.amp.GradScaler if "cuda" == device.type else torch.cpu.amp.GradScaler
with self.assertWarnsRegex(
FutureWarning,
rf"`torch.{device.type}.amp.GradScaler\(args...\)` is deprecated.",
UserWarning,
rf"torch.{device.type}.amp.GradScaler\(args...\) is deprecated.",
):
_ = GradScaler(init_scale=2.0)

11
torch/__init__.py
View File

@ -1996,6 +1996,17 @@ from torch import func as func
from torch.func import vmap
# The function _sparse_coo_tensor_unsafe is removed from PyTorch
# Python API (v. 1.13), here we temporarily provide its replacement
# with a deprecation warning.
# TODO: remove the function for PyTorch v 1.15.
def _sparse_coo_tensor_unsafe(*args, **kwargs):
import warnings
warnings.warn('torch._sparse_coo_tensor_unsafe is deprecated, '
'use torch.sparse_coo_tensor(..., check_invariants=False) instead.')
kwargs['check_invariants'] = False
return torch.sparse_coo_tensor(*args, **kwargs)
# Register MPS specific decomps
torch.backends.mps._init()

3
torch/_dynamo/decorators.py
View File

@ -7,6 +7,7 @@ from . import trace_rules, variables
from .comptime import comptime
from .eval_frame import DisableContext, innermost_fn, RunOnlyContext
from .exc import IncorrectUsage
from .external_utils import is_compiling
if TYPE_CHECKING:
from torch._C._dynamo.eval_frame import ( # noqa: F401
@ -272,7 +273,7 @@ def mark_static(t, index=None):
Unlike mark_dynamic, this can be done inside a graph, in which case it
induces specialization on the tensor.
"""
if torch.compiler.is_compiling():
if is_compiling():
if index is None:
for s in t.size():
comptime.force_static(s)

8
torch/_dynamo/eval_frame.py
View File

@ -798,8 +798,7 @@ def explain(f, *extra_args, **extra_kwargs):
warnings.warn(
"explain(f, *args, **kwargs) is deprecated, use explain(f)(*args, **kwargs) instead. "
"If you don't migrate, we may break your explain call in the future if your user defined kwargs "
"conflict with future kwargs added to explain(f).",
FutureWarning,
"conflict with future kwargs added to explain(f)."
)
return inner(*extra_args, **extra_kwargs)
else:
@ -942,7 +941,7 @@ def check_signature_rewritable(graph):
tb = "".join(traceback.format_list(stack))
extra = ""
if len(user_stacks) > 1:
extra = f"(elided {len(user_stacks) - 1} more accesses)"
extra = f"(elided {len(user_stacks)-1} more accesses)"
msg = f"{source.name()}, accessed at:\n{tb}{extra}"
# TODO: option to print ALL of the stack traces at once
input_errors.append(msg)
@ -1477,8 +1476,7 @@ def export(
warnings.warn(
"export(f, *args, **kwargs) is deprecated, use export(f)(*args, **kwargs) instead. "
"If you don't migrate, we may break your export call in the future if your user defined kwargs "
"conflict with future kwargs added to export(f).",
FutureWarning,
"conflict with future kwargs added to export(f)."
)
return inner(*extra_args, **extra_kwargs)
else:

5
torch/_dynamo/external_utils.py
View File

@ -2,7 +2,6 @@
import functools
from typing import List
from typing_extensions import deprecated
import torch
import torch.utils._pytree as pytree
@ -13,10 +12,6 @@ except ModuleNotFoundError:
np = None # type: ignore[assignment]
@deprecated(
"`is_compiling` is deprecated. Use `torch.compiler.is_compiling()` instead.",
category=FutureWarning,
)
def is_compiling() -> bool:
"""
Indicates whether we are tracing/compiling with torch.compile() or torch.export().

6
torch/_functorch/apis.py
View File

@ -188,7 +188,7 @@ def vmap(
vmap does not provide general autobatching or handle variable-length
sequences out of the box.
"""
from torch.compiler import is_compiling
from torch._dynamo import is_compiling
_check_randomness_arg(randomness)
if not (chunk_size is None or chunk_size > 0):
@ -390,7 +390,7 @@ def grad(func: Callable, argnums: argnums_t = 0, has_aux: bool = False) -> Calla
"""
# To avoid cyclical dependency.
import torch._functorch.eager_transforms as eager_transforms
from torch.compiler import is_compiling
from torch._dynamo import is_compiling
def wrapper(*args, **kwargs):
return eager_transforms.grad_impl(func, argnums, has_aux, args, kwargs)
@ -432,8 +432,8 @@ def grad_and_value(
See :func:`grad` for examples
"""
from torch._dynamo import is_compiling
from torch._functorch import eager_transforms
from torch.compiler import is_compiling
def wrapper(*args, **kwargs):
return eager_transforms.grad_and_value_impl(

26
torch/_functorch/deprecated.py
View File

@ -1,12 +1,3 @@
"""
The APIs in this file are exposed as `functorch.*`. They are thin wrappers
around the torch.func.* APIs that have deprecation warnings -- we're trying
to move people to the torch.func.* equivalents.
NB: We don't use *args, **kwargs in the signatures because that changes the
documentation.
"""
import textwrap
import warnings
from typing import Any, Callable, Optional, Tuple, Union
@ -18,16 +9,25 @@ import torch.nn as nn
from torch._functorch.eager_transforms import argnums_t
from torch._functorch.vmap import in_dims_t, out_dims_t
"""
The APIs in this file are exposed as `functorch.*`. They are thin wrappers
around the torch.func.* APIs that have deprecation warnings -- we're trying
to move people to the torch.func.* equivalents.
NB: We don't use *args, **kwargs in the signatures because that changes the
documentation.
"""
def get_warning(api, new_api=None, replace_newlines=False):
if new_api is None:
new_api = f"torch.func.{api}"
warning = (
f"We've integrated functorch into PyTorch. As the final step of the \n"
f"integration, `functorch.{api}` is deprecated as of PyTorch \n"
f"integration, functorch.{api} is deprecated as of PyTorch \n"
f"2.0 and will be deleted in a future version of PyTorch >= 2.3. \n"
f"Please use `{new_api}` instead; see the PyTorch 2.0 release notes \n"
f"and/or the `torch.func` migration guide for more details \n"
f"Please use {new_api} instead; see the PyTorch 2.0 release notes \n"
f"and/or the torch.func migration guide for more details \n"
f"https://pytorch.org/docs/main/func.migrating.html"
)
if replace_newlines:
@ -37,7 +37,7 @@ def get_warning(api, new_api=None, replace_newlines=False):
def warn_deprecated(api, new_api=None):
warning = get_warning(api, new_api, replace_newlines=True)
warnings.warn(warning, FutureWarning, stacklevel=2)
warnings.warn(warning, stacklevel=2)
def setup_docs(functorch_api, torch_func_api=None, new_api_name=None):

4
torch/_functorch/eager_transforms.py
View File

@ -765,7 +765,7 @@ def jacrev(
# Dynamo does not support HOP composition if their inner function is
# annotated with @functools.wraps(...). We circumvent this issue by applying
# wraps only if we're not tracing with dynamo.
if not torch.compiler.is_compiling():
if not torch._dynamo.is_compiling():
wrapper_fn = wraps(func)(wrapper_fn)
return wrapper_fn
@ -1346,7 +1346,7 @@ def jacfwd(
# Dynamo does not support HOP composition if their inner function is
# annotated with @functools.wraps(...). We circumvent this issue by applying
# wraps only if we're not tracing with dynamo.
if not torch.compiler.is_compiling():
if not torch._dynamo.is_compiling():
wrapper_fn = wraps(func)(wrapper_fn)
return wrapper_fn

5
torch/_functorch/pytree_hacks.py
View File

@ -16,8 +16,7 @@ __all__ = ["tree_map_", "treespec_pprint"]
with warnings.catch_warnings():
warnings.simplefilter("always")
warnings.warn(
"`torch._functorch.pytree_hacks` is deprecated and will be removed in a future release. "
"Please `use torch.utils._pytree` instead.",
"torch._functorch.pytree_hacks is deprecated and will be removed in a future release. "
"Please use torch.utils._pytree instead.",
DeprecationWarning,
stacklevel=2,
)

2
torch/_higher_order_ops/associative_scan.py
View File

@ -76,7 +76,7 @@ def associative_scan(
assert callable(combine_fn), "combine_fn must be a callable, but got {combine_fn}"
assert isinstance(dim, int), "dim must be an int, but got {type(dim)}"
if not torch.compiler.is_compiling():
if not torch._dynamo.is_compiling():
with _set_compilation_env(), torch._dynamo.utils.disable_cache_limit():
return torch.compile(associative_scan, fullgraph=True)(
combine_fn, input, dim

3
torch/_inductor/ir.py
View File

@ -2750,8 +2750,7 @@ class FixedLayout(Layout):
"""A closure containing math to read a given element"""
def indexer(index):
assert len(index) == len(self.stride)
assert len(index) == len(self.size)
assert len(index) == len(self.stride) == len(self.size)
result = self.offset
for idx, stride, sz in zip(index, self.stride, self.size):
if sz != 1:

9
torch/_library/abstract_impl.py
View File

@ -1,7 +1,7 @@
import contextlib
import functools
import warnings
from typing import Callable, Optional
from typing_extensions import deprecated
import torch
from torch._library.utils import Kernel, RegistrationHandle
@ -124,11 +124,10 @@ class AbstractImplCtx:
self._shape_env = _fake_mode.shape_env
self._op = _op
@deprecated(
"`create_unbacked_symint` is deprecated, please use `new_dynamic_size` instead",
category=FutureWarning,
)
def create_unbacked_symint(self, *, min=2, max=None) -> torch.SymInt:
warnings.warn(
"create_unbacked_symint is deprecated, please use new_dynamic_size instead"
)
return self.new_dynamic_size(min=min, max=max)
def new_dynamic_size(self, *, min=0, max=None) -> torch.SymInt:

13
torch/_prims_common/__init__.py
View File

@ -21,7 +21,7 @@ from typing import (
TYPE_CHECKING,
Union,
)
from typing_extensions import deprecated, TypeAlias
from typing_extensions import TypeAlias
if TYPE_CHECKING:
@ -1789,11 +1789,6 @@ def check_in_bounds_for_storage(
# NOTE: This function should ideally be removed, but some Meta internal models
# packaged with `torch.package` are using it, so it will have to be removed
# at some point in the future when those models no longer use this function.
@deprecated(
"`torch._prims_common.check` is deprecated and will be removed in the future. "
"Please use `torch._check*` functions instead.",
category=FutureWarning,
)
def check(
b: bool, s: Callable[[], str], exc_type: Type[Exception] = RuntimeError
) -> None:
@ -1806,6 +1801,12 @@ def check(
.. note:: This function is planned for removal in the future. Please use
`torch._check*` functions instead.
"""
warnings.warn(
DeprecationWarning(
"'torch._prims_common.check' will be removed in the future. Please use "
"'torch._check*' functions instead"
)
)
torch._check_with(exc_type, b, s)

6
torch/_utils.py
View File

@ -7,7 +7,7 @@ import traceback
import warnings
from collections import defaultdict
from typing import Any, Callable, DefaultDict, Generic, List, Optional
from typing_extensions import deprecated, ParamSpec
from typing_extensions import ParamSpec
import torch
@ -852,10 +852,6 @@ def classproperty(func):
return _ClassPropertyDescriptor(func)
@deprecated(
"`is_compiling` is deprecated. Use `torch.compiler.is_compiling()` instead.",
category=FutureWarning,
)
def is_compiling() -> bool:
"""
Indicates whether we are tracing/compiling with torch.compile() or torch.export().

10
torch/_vmap_internals.py
View File

@ -1,6 +1,6 @@
import functools
import warnings
from typing import Any, Callable, List, Optional, Tuple, Union
from typing_extensions import deprecated
import torch
from torch import Tensor
@ -190,14 +190,14 @@ def _get_name(func: Callable):
# vmap(func)(inputs) wraps all Tensor inputs to be batched in BatchedTensors,
# sends those into func, and then unwraps the output BatchedTensors. Operations
# on BatchedTensors perform the batched operations that the user is asking for.
@deprecated(
"Please use `torch.vmap` instead of `torch._vmap_internals.vmap`.",
category=FutureWarning,
)
def vmap(func: Callable, in_dims: in_dims_t = 0, out_dims: out_dims_t = 0) -> Callable:
"""
Please use torch.vmap instead of this API.
"""
warnings.warn(
"Please use torch.vmap instead of torch._vmap_internals.vmap. ",
stacklevel=2,
)
return _vmap(func, in_dims, out_dims)

11
torch/ao/nn/quantizable/modules/activation.py
View File

@ -224,6 +224,7 @@ class MultiheadAttention(nn.MultiheadAttention):
return fp
@classmethod
def from_observed(cls, other):
# The whole flow is float -> observed -> quantized
@ -335,10 +336,7 @@ class MultiheadAttention(nn.MultiheadAttention):
if attn_mask is not None:
if attn_mask.dtype == torch.uint8:
warnings.warn(
"Byte tensor for `attn_mask` in `nn.MultiheadAttention` is deprecated. "
"Use bool tensor instead.",
)
warnings.warn("Byte tensor for attn_mask in nn.MultiheadAttention is deprecated. Use bool tensor instead.")
attn_mask = attn_mask.to(torch.bool)
assert attn_mask.is_floating_point() or attn_mask.dtype == torch.bool, \
f'Only float and bool types are supported for attn_mask, not {attn_mask.dtype}'
@ -356,10 +354,7 @@ class MultiheadAttention(nn.MultiheadAttention):
# convert ByteTensor key_padding_mask to bool
if key_padding_mask is not None and key_padding_mask.dtype == torch.uint8:
warnings.warn(
"Byte tensor for `key_padding_mask` in `nn.MultiheadAttention` is deprecated. "
"Use bool tensor instead.",
)
warnings.warn("Byte tensor for key_padding_mask in nn.MultiheadAttention is deprecated. Use bool tensor instead.")
key_padding_mask = key_padding_mask.to(torch.bool)
if self.bias_k is not None and self.bias_v is not None:
if static_k is None and static_v is None:

6
torch/ao/nn/quantized/dynamic/modules/rnn.py
View File

@ -1,6 +1,5 @@
import numbers
import warnings
from typing_extensions import deprecated
import torch
import torch.nn as nn
@ -17,11 +16,8 @@ def _apply_permutation(tensor: Tensor, permutation: Tensor, dim: int = 1) -> Ten
return tensor.index_select(dim, permutation)
@deprecated(
"`apply_permutation` is deprecated, please use `tensor.index_select(dim, permutation)` instead",
category=FutureWarning,
)
def apply_permutation(tensor: Tensor, permutation: Tensor, dim: int = 1) -> Tensor:
warnings.warn("apply_permutation is deprecated, please use tensor.index_select(dim, permutation) instead")
return _apply_permutation(tensor, permutation, dim)

18
torch/ao/quantization/fx/convert.py
View File

@ -949,30 +949,24 @@ def convert(
if convert_custom_config is None:
convert_custom_config = ConvertCustomConfig()
if isinstance(convert_custom_config, dict):
if isinstance(convert_custom_config, Dict):
warnings.warn(
"Passing a convert_custom_config_dict to convert is deprecated and will not be supported "
"in a future version. Please pass in a ConvertCustomConfig instead.",
FutureWarning,
)
"in a future version. Please pass in a ConvertCustomConfig instead.")
convert_custom_config = ConvertCustomConfig.from_dict(convert_custom_config)
if isinstance(qconfig_mapping, dict):
if isinstance(qconfig_mapping, Dict):
warnings.warn(
"Passing a QConfig dictionary to convert is deprecated and will not be supported "
"in a future version. Please pass in a QConfigMapping instead.",
FutureWarning,
)
"in a future version. Please pass in a QConfigMapping instead.")
qconfig_mapping = QConfigMapping.from_dict(qconfig_mapping) if qconfig_mapping else None
qconfig_mapping = copy.deepcopy(qconfig_mapping)
assert qconfig_mapping is None or isinstance(qconfig_mapping, QConfigMapping)
if isinstance(backend_config, dict):
if isinstance(backend_config, Dict):
warnings.warn(
"Passing a backend_config_dict to prepare is deprecated and will not be supported "
"in a future version. Please pass in a BackendConfig instead.",
FutureWarning,
)
"in a future version. Please pass in a BackendConfig instead.")
backend_config = BackendConfig.from_dict(backend_config)
if backend_config is None:

12
torch/ao/quantization/fx/fuse.py
View File

@ -52,20 +52,16 @@ def fuse(
if fuse_custom_config is None:
fuse_custom_config = FuseCustomConfig()
if isinstance(fuse_custom_config, dict):
if isinstance(fuse_custom_config, Dict):
warnings.warn(
"Passing a fuse_custom_config_dict to fuse is deprecated and will not be supported "
"in a future version. Please pass in a FuseCustomConfig instead.",
FutureWarning,
)
"in a future version. Please pass in a FuseCustomConfig instead.")
fuse_custom_config = FuseCustomConfig.from_dict(fuse_custom_config)
if isinstance(backend_config, dict):
if isinstance(backend_config, Dict):
warnings.warn(
"Passing a backend_config_dict to prepare is deprecated and will not be supported "
"in a future version. Please pass in a BackendConfig instead.",
FutureWarning,
)
"in a future version. Please pass in a BackendConfig instead.")
backend_config = BackendConfig.from_dict(backend_config)
named_modules = dict(model.named_modules())

24
torch/ao/quantization/fx/prepare.py
View File

@ -1749,36 +1749,28 @@ def prepare(
if _equalization_config is None:
_equalization_config = QConfigMapping()
if isinstance(qconfig_mapping, dict):
if isinstance(qconfig_mapping, Dict):
warnings.warn(
"Passing a QConfig dictionary to prepare is deprecated and will not be supported "
"in a future version. Please pass in a QConfigMapping instead.",
FutureWarning,
)
"in a future version. Please pass in a QConfigMapping instead.")
qconfig_mapping = QConfigMapping.from_dict(qconfig_mapping)
if isinstance(_equalization_config, dict):
if isinstance(_equalization_config, Dict):
warnings.warn(
"Passing a QConfig dictionary to prepare for equalization is deprecated and will not "
"be supported in a future version. Please pass in a QConfigMapping instead.",
FutureWarning,
)
"be supported in a future version. Please pass in a QConfigMapping instead.")
_equalization_config = QConfigMapping.from_dict(_equalization_config)
if isinstance(prepare_custom_config, dict):
if isinstance(prepare_custom_config, Dict):
warnings.warn(
"Passing a prepare_custom_config_dict to prepare is deprecated and will not be supported "
"in a future version. Please pass in a PrepareCustomConfig instead.",
FutureWarning,
)
"in a future version. Please pass in a PrepareCustomConfig instead.")
prepare_custom_config = PrepareCustomConfig.from_dict(prepare_custom_config)
if isinstance(backend_config, dict):
if isinstance(backend_config, Dict):
warnings.warn(
"Passing a backend_config_dict to prepare is deprecated and will not be supported "
"in a future version. Please pass in a BackendConfig instead.",
FutureWarning,
)
"in a future version. Please pass in a BackendConfig instead.")
backend_config = BackendConfig.from_dict(backend_config)
assert isinstance(qconfig_mapping, QConfigMapping)

22
torch/ao/quantization/qconfig.py
View File

@ -1,6 +1,5 @@
from collections import namedtuple
from typing import Optional, Any, Union, Type
from typing_extensions import deprecated
import torch
import torch.nn as nn
@ -107,10 +106,6 @@ class QConfig(namedtuple('QConfig', ['activation', 'weight'])):
return super().__new__(cls, activation, weight)
@deprecated(
"`QConfigDynamic` is going to be deprecated in PyTorch 1.12, please use `QConfig` instead",
category=FutureWarning,
)
class QConfigDynamic(namedtuple('QConfigDynamic', ['activation', 'weight'])):
"""
Describes how to dynamically quantize a layer or a part of the network by providing
@ -132,6 +127,7 @@ class QConfigDynamic(namedtuple('QConfigDynamic', ['activation', 'weight'])):
if isinstance(weight, nn.Module):
raise ValueError("QConfigDynamic received observer instance, please pass observer class instead. " +
"Use MyObserver.with_args(x=1) to override arguments to constructor if needed")
warnings.warn("QConfigDynamic is going to be deprecated in PyTorch 1.12, please use QConfig instead")
return super().__new__(cls, activation, weight)
@ -426,20 +422,16 @@ _default_quint8_placeholder_qconfig = QConfig(
weight=None,
)
@deprecated(
"`torch.ao.quantization.get_default_qconfig_dict` is deprecated and will be removed in "
"a future version. Please use `torch.ao.quantization.get_default_qconfig_mapping` instead.",
category=FutureWarning,
)
def get_default_qconfig_dict(backend='x86', version=0):
warnings.warn(
"torch.ao.quantization.get_default_qconfig_dict is deprecated and will be removed in "
"a future version. Please use torch.ao.quantization.get_default_qconfig_mapping instead.")
return torch.ao.quantization.get_default_qconfig_mapping(backend, version).to_dict()
@deprecated(
"`torch.ao.quantization.get_default_qat_qconfig_dict` is deprecated and will be removed in "
"a future version. Please use `torch.ao.quantization.get_default_qat_qconfig_mapping` instead.",
category=FutureWarning,
)
def get_default_qat_qconfig_dict(backend='x86', version=1):
warnings.warn(
"torch.ao.quantization.get_default_qat_qconfig_dict is deprecated and will be removed in "
"a future version. Please use torch.ao.quantization.get_default_qat_qconfig_mapping instead.")
return torch.ao.quantization.get_default_qat_qconfig_mapping(backend, version).to_dict()
def _assert_valid_qconfig(qconfig: Optional[QConfig],

18
torch/ao/quantization/quantize_fx.py
View File

@ -117,12 +117,10 @@ forward graph of the parent module,
if _equalization_config is None:
_equalization_config = QConfigMapping()
if isinstance(prepare_custom_config, dict):
if isinstance(prepare_custom_config, Dict):
warnings.warn(
"Passing a prepare_custom_config_dict to prepare is deprecated and will not be supported "
"in a future version. Please pass in a PrepareCustomConfig instead.",
FutureWarning,
)
"in a future version. Please pass in a PrepareCustomConfig instead.")
prepare_custom_config = PrepareCustomConfig.from_dict(prepare_custom_config)
# swap FloatFunctional with FXFloatFunctional
@ -224,12 +222,10 @@ def fuse_fx(
if fuse_custom_config is None:
fuse_custom_config = FuseCustomConfig()
if isinstance(fuse_custom_config, dict):
if isinstance(fuse_custom_config, Dict):
warnings.warn(
"Passing a fuse_custom_config_dict to fuse is deprecated and will not be supported "
"in a future version. Please pass in a FuseCustomConfig instead.",
FutureWarning,
)
"in a future version. Please pass in a FuseCustomConfig instead.")
fuse_custom_config = FuseCustomConfig.from_dict(fuse_custom_config)
torch._C._log_api_usage_once("quantization_api.quantize_fx.fuse_fx")
@ -515,12 +511,10 @@ def _convert_fx(
if convert_custom_config is None:
convert_custom_config = ConvertCustomConfig()
if isinstance(convert_custom_config, dict):
if isinstance(convert_custom_config, Dict):
warnings.warn(
"Passing a convert_custom_config_dict to convert is deprecated and will not be supported "
"in a future version. Please pass in a ConvertCustomConfig instead.",
FutureWarning,
)
"in a future version. Please pass in a ConvertCustomConfig instead.")
convert_custom_config = ConvertCustomConfig.from_dict(convert_custom_config)
_check_is_graph_module(graph_module)

16
torch/autograd/__init__.py
View File

@ -252,20 +252,17 @@ def backward(
)
if grad_variables is not None:
warnings.warn(
"`grad_variables` is deprecated. Use `grad_tensors` instead.",
FutureWarning,
)
warnings.warn("'grad_variables' is deprecated. Use 'grad_tensors' instead.")
if grad_tensors is None:
grad_tensors = grad_variables
else:
raise RuntimeError(
"`grad_tensors` and `grad_variables` (deprecated) "
"arguments both passed to `backward()`. Please only "
"use `grad_tensors`."
"'grad_tensors' and 'grad_variables' (deprecated) "
"arguments both passed to backward(). Please only "
"use 'grad_tensors'."
)
if inputs is not None and len(inputs) == 0:
raise RuntimeError("`inputs` argument to `backward()` cannot be empty.")
raise RuntimeError("'inputs' argument to backward() cannot be empty.")
tensors = (tensors,) if isinstance(tensors, torch.Tensor) else tuple(tensors)
inputs = (
@ -398,8 +395,7 @@ def grad(
warnings.warn(
"only_inputs argument is deprecated and is ignored now "
"(defaults to True). To accumulate gradient for other "
"parts of the graph, please use torch.autograd.backward.",
FutureWarning,
"parts of the graph, please use torch.autograd.backward."
)
grad_outputs_ = _tensor_or_tensors_to_tuple(grad_outputs, len(t_outputs))

11
torch/autograd/_functions/tensor.py
View File

@ -1,6 +1,6 @@
import operator
import warnings
from functools import reduce
from typing_extensions import deprecated
import torch
import torch._utils
@ -9,12 +9,11 @@ from ..function import Function
class Type(Function):
@staticmethod
@deprecated(
"`torch.autograd._functions.Type` is deprecated as of PyTorch 2.1, "
"please use `torch.tensor.to(dtype=dtype)` instead.",
category=FutureWarning,
)
def forward(ctx, i, dest_type):
warnings.warn(
"torch.autograd._functions.Type is deprecated as of PyTorch 2.1, please use "
"torch.tensor.to(dtype=dtype) instead."
)
ctx.input_type = type(i)
ctx.input_device = -1 if not i.is_cuda else i.get_device()
return i.type(dest_type)

16
torch/autograd/function.py
View File

@ -4,7 +4,6 @@ import itertools
import warnings
from collections import OrderedDict
from typing import Any, List, Optional, Tuple
from typing_extensions import deprecated
import torch
import torch._C as _C
@ -180,14 +179,12 @@ class FunctionCtx:
"""
self.dirty_tensors = args
@deprecated(
"`mark_shared_storage` is deprecated. "
"Tensors with shared storages are automatically tracked. "
"Note that calls to `set_()` are not tracked",
category=FutureWarning,
)
def mark_shared_storage(self, *pairs):
pass
warnings.warn(
"mark_shared_storage is deprecated. "
"Tensors with shared storages are automatically tracked. Note "
"that calls to `set_()` are not tracked"
)
def mark_non_differentiable(self, *args: torch.Tensor):
r"""Mark outputs as non-differentiable.
@ -494,8 +491,9 @@ class Function(_SingleLevelFunction):
"""
def __init__(self, *args, **kwargs):
cls = self.__class__
warnings.warn(
f"{self.__class__} should not be instantiated. Methods on autograd functions"
f"{cls} should not be instantiated. Methods on autograd functions"
"are all static, so you should invoke them on the class itself. "
"Instantiating an autograd function will raise an "
"error in a future version of PyTorch.",

31
torch/autograd/gradcheck.py
View File

@ -3,7 +3,6 @@ import functools
import warnings
from itertools import product
from typing import Callable, Dict, Iterable, List, Optional, Tuple, Union
from typing_extensions import deprecated
import torch
import torch.testing
@ -307,14 +306,6 @@ def _get_numerical_jacobian(
return jacobians
@deprecated(
"`get_numerical_jacobian` was part of PyTorch's private API and not "
"meant to be exposed. We are deprecating it and it will be removed "
"in a future version of PyTorch. If you have a specific use for "
"this or feature request for this to be a stable API, please file "
"us an issue at https://github.com/pytorch/pytorch/issues/new",
category=FutureWarning,
)
def get_numerical_jacobian(fn, inputs, target=None, eps=1e-3, grad_out=1.0):
"""Compute the numerical Jacobian for a given fn and its inputs.
@ -334,6 +325,13 @@ def get_numerical_jacobian(fn, inputs, target=None, eps=1e-3, grad_out=1.0):
Note that `target` may not even be part of `input` to `fn`, so please be
**very careful** in this to not clone `target`.
"""
warnings.warn(
"get_numerical_jacobian was part of PyTorch's private API and not "
"meant to be exposed. We are deprecating it and it will be removed "
"in a future version of PyTorch. If you have a specific use for "
"this or feature request for this to be a stable API, please file "
"us an issue at https://github.com/pytorch/pytorch/issues/new"
)
if (
grad_out != 1.0
): # grad_out param is only kept for backward compatibility reasons
@ -820,17 +818,16 @@ def _get_analytical_vJu_backward_mode(
return reduced_jacobians
@deprecated(
"`get_analytical_jacobian` was part of PyTorch's private API and not "
"meant to be exposed. We are deprecating it and it will be removed "
"in a future version of PyTorch. If you have a specific use for "
"this or feature request for this to be a stable API, please file "
"us an issue at https://github.com/pytorch/pytorch/issues/new",
category=FutureWarning,
)
def get_analytical_jacobian(inputs, output, nondet_tol=0.0, grad_out=1.0):
# Replicates the behavior of the old get_analytical_jacobian before the refactor
# This shares much of its code with _check_analytical_jacobian_attributes
warnings.warn(
"get_analytical_jacobian was part of PyTorch's private API and not "
"meant to be exposed. We are deprecating it and it will be removed "
"in a future version of PyTorch. If you have a specific use for "
"this or feature request for this to be a stable API, please file "
"us an issue at https://github.com/pytorch/pytorch/issues/new"
)
if (
grad_out != 1.0
): # grad_out param is only kept for backward compatibility reasons

4
torch/autograd/profiler.py
View File

@ -213,9 +213,7 @@ class profile:
self.use_cuda = use_cuda
if self.use_cuda:
warn(
"The attribute `use_cuda` will be deprecated soon, "
"please use ``use_device = 'cuda'`` instead.",
FutureWarning,
"The attribute `use_cuda` will be deprecated soon, please use ``use_device = 'cuda'`` instead."
)
self.use_device: Optional[str] = "cuda"
else:

10
torch/autograd/profiler_legacy.py
View File

@ -1,6 +1,5 @@
import itertools
import warnings
from typing_extensions import deprecated
from warnings import warn
import torch
import torch.cuda
@ -24,11 +23,6 @@ from torch.autograd.profiler_util import (
__all__ = ["profile"]
@deprecated(
"`torch.autograd.profiler_legacy.profile` is deprecated and will be removed in a future release. "
"Please use `torch.profiler` instead.",
category=None, # TODO: change to `FutureWarning`
)
class profile:
"""DEPRECATED: use torch.profiler instead."""
@ -57,7 +51,7 @@ class profile:
self.with_modules = with_modules
if self.use_cuda and not torch.cuda.is_available():
warnings.warn("CUDA is not available, disabling CUDA profiling")
warn("CUDA is not available, disabling CUDA profiling")
self.use_cuda = False
if self.use_cuda:

23
torch/autograd/profiler_util.py
View File

@ -6,7 +6,6 @@ from collections import defaultdict, namedtuple
from operator import attrgetter
from typing import Any, Dict, List, Optional, Tuple
from typing_extensions import deprecated
import torch
from torch.autograd import DeviceType
@ -416,10 +415,6 @@ class FormattedTimesMixin:
return 0.0 if self.count == 0 else 1.0 * self.device_time_total / self.count # type: ignore[attr-defined]
@property
@deprecated(
"`cuda_time` is deprecated, please use `device_time` instead.",
category=FutureWarning,
)
def cuda_time(self): # To be deprecated
return self.device_time
@ -543,12 +538,8 @@ class FunctionEvent(FormattedTimesMixin):
)
@property
@deprecated(
"`self_cuda_memory_usage` is deprecated. Use `self_device_memory_usage` instead.",
category=FutureWarning,
)
def self_cuda_memory_usage(self): # To be deprecated
return self.self_device_memory_usage
self.self_device_memory_usage
@property
def cpu_time_total(self):
@ -583,12 +574,8 @@ class FunctionEvent(FormattedTimesMixin):
return self.time_range.elapsed_us()
@property
@deprecated(
"`cuda_time_total` is deprecated. Use `device_time_total` instead.",
category=FutureWarning,
)
def cuda_time_total(self): # To be deprecated
return self.device_time_total
self.device_time_total
@property
def self_device_time_total(self):
@ -603,12 +590,8 @@ class FunctionEvent(FormattedTimesMixin):
return self.device_time_total
@property
@deprecated(
"`self_cuda_time_total` is deprecated. Use `self_device_time_total` instead.",
category=FutureWarning,
)
def self_cuda_time_total(self): # To be deprecated
return self.self_device_time_total
self.self_device_time_total
@property
def key(self):

20
torch/backends/cuda/__init__.py
View File

@ -1,7 +1,7 @@
import contextlib
import warnings
from typing import Union
from typing_extensions import deprecated
import torch
@ -377,15 +377,6 @@ def enable_cudnn_sdp(enabled: bool):
@contextlib.contextmanager
@deprecated(
(
"`torch.backends.cuda.sdp_kernel()` is deprecated. "
"In the future, this context manager will be removed. "
"Please see `torch.nn.attention.sdpa_kernel()` for the new context manager, "
"with updated signature."
),
category=FutureWarning,
)
def sdp_kernel(
enable_flash: bool = True,
enable_math: bool = True,
@ -398,6 +389,15 @@ def sdp_kernel(
This context manager can be used to temporarily enable or disable any of the three backends for scaled dot product attention.
Upon exiting the context manager, the previous state of the flags will be restored.
"""
warnings.warn(
(
"torch.backends.cuda.sdp_kernel() "
"is deprecated. In the future, this context manager will be removed. "
"Please see, torch.nn.attention.sdpa_kernel() for the new context manager, with updated "
"signature."
),
FutureWarning,
)
from torch.nn.attention import sdpa_kernel
backend_list = []

11
torch/cpu/amp/autocast_mode.py
View File

@ -1,5 +1,5 @@
import warnings
from typing import Any
from typing_extensions import deprecated
import torch
@ -12,11 +12,6 @@ class autocast(torch.amp.autocast_mode.autocast):
``torch.cpu.amp.autocast(args...)`` is deprecated. Please use ``torch.amp.autocast("cpu", args...)`` instead.
"""
@deprecated(
"`torch.cpu.amp.autocast(args...)` is deprecated. "
"Please use `torch.amp.autocast('cpu', args...)` instead.",
category=FutureWarning,
)
def __init__(
self,
enabled: bool = True,
@ -28,6 +23,10 @@ class autocast(torch.amp.autocast_mode.autocast):
self.device = "cpu"
self.fast_dtype = dtype
return
warnings.warn(
"torch.cpu.amp.autocast(args...) is deprecated. Please use torch.amp.autocast('cpu', args...) instead.",
DeprecationWarning,
)
super().__init__(
"cpu", enabled=enabled, dtype=dtype, cache_enabled=cache_enabled
)

10
torch/cpu/amp/grad_scaler.py
View File

@ -1,4 +1,4 @@
from typing_extensions import deprecated
import warnings
import torch
@ -11,11 +11,6 @@ class GradScaler(torch.amp.GradScaler):
``torch.cpu.amp.GradScaler(args...)`` is deprecated. Please use ``torch.amp.GradScaler("cpu", args...)`` instead.
"""
@deprecated(
"`torch.cpu.amp.GradScaler(args...)` is deprecated. "
"Please use `torch.amp.GradScaler('cpu', args...)` instead.",
category=FutureWarning,
)
def __init__(
self,
init_scale: float = 2.0**16,
@ -24,6 +19,9 @@ class GradScaler(torch.amp.GradScaler):
growth_interval: int = 2000,
enabled: bool = True,
) -> None:
warnings.warn(
"torch.cpu.amp.GradScaler(args...) is deprecated. Please use torch.amp.GradScaler('cpu', args...) instead."
)
super().__init__(
"cpu",
init_scale=init_scale,

6
torch/cuda/_memory_viz.py
View File

@ -145,8 +145,8 @@ def compare(before, after, format_flamegraph=format_flamegraph):
before_segs = {_seg_key(seg) for seg in before}
after_segs = {_seg_key(seg) for seg in after}
print(f'only_before = {[a for a, _ in (before_segs - after_segs)]}')
print(f'only_after = {[a for a, _ in (after_segs - before_segs)]}')
print(f'only_before = {[a for a,_ in (before_segs - after_segs)]}')
print(f'only_after = {[a for a,_ in (after_segs - before_segs)]}')
for seg in before:
if _seg_key(seg) not in after_segs:
@ -382,7 +382,7 @@ add_local_files(local_files, $VIZ_KIND)
def _format_viz(data, viz_kind, device):
if device is not None:
warnings.warn('device argument is deprecated, plots now contain all device', FutureWarning)
warnings.warn('device argument is deprecated, plots now contain all device')
buffer = pickle.dumps(data)
buffer += b'\x00' * (3 - len(buffer) % 3)
# Encode the buffer with base64

27
torch/cuda/amp/autocast_mode.py
View File

@ -1,6 +1,6 @@
import functools
import warnings
from typing import Any
from typing_extensions import deprecated
import torch
@ -13,11 +13,6 @@ class autocast(torch.amp.autocast_mode.autocast):
``torch.cuda.amp.autocast(args...)`` is deprecated. Please use ``torch.amp.autocast("cuda", args...)`` instead.
"""
@deprecated(
"`torch.cuda.amp.autocast(args...)` is deprecated. "
"Please use `torch.amp.autocast('cuda', args...)` instead.",
category=FutureWarning,
)
def __init__(
self,
enabled: bool = True,
@ -29,6 +24,10 @@ class autocast(torch.amp.autocast_mode.autocast):
self.device = "cuda"
self.fast_dtype = dtype
return
warnings.warn(
"torch.cuda.amp.autocast(args...) is deprecated. Please use torch.amp.autocast('cuda', args...) instead.",
DeprecationWarning,
)
super().__init__(
"cuda", enabled=enabled, dtype=dtype, cache_enabled=cache_enabled
)
@ -50,29 +49,25 @@ class autocast(torch.amp.autocast_mode.autocast):
return super().__call__(func)
@deprecated(
"`torch.cuda.amp.custom_fwd(args...)` is deprecated. "
"Please use `torch.amp.custom_fwd(args..., device_type='cuda')` instead.",
category=FutureWarning,
)
def custom_fwd(fwd=None, *, cast_inputs=None):
"""
``torch.cuda.amp.custom_fwd(args...)`` is deprecated. Please use
``torch.amp.custom_fwd(args..., device_type='cuda')`` instead.
"""
warnings.warn(
"torch.cuda.amp.custom_fwd(args...) is deprecated. Please use torch.amp.custom_fwd(args..., device_type='cuda') instead."
)
return functools.partial(torch.amp.custom_fwd, device_type="cuda")(
fwd=fwd, cast_inputs=cast_inputs
)
@deprecated(
"`torch.cuda.amp.custom_bwd(args...)` is deprecated. "
"Please use `torch.amp.custom_bwd(args..., device_type='cuda')` instead.",
category=FutureWarning,
)
def custom_bwd(bwd):
"""
``torch.cuda.amp.custom_bwd(args...)`` is deprecated. Please use
``torch.amp.custom_bwd(args..., device_type='cuda')`` instead.
"""
warnings.warn(
"torch.cuda.amp.custom_bwd(args...) is deprecated. Please use torch.amp.custom_bwd(args..., device_type='cuda') instead."
)
return functools.partial(torch.amp.custom_bwd, device_type="cuda")(bwd)

10
torch/cuda/amp/grad_scaler.py
View File

@ -1,4 +1,4 @@
from typing_extensions import deprecated
import warnings
import torch
@ -11,11 +11,6 @@ class GradScaler(torch.amp.GradScaler):
``torch.cuda.amp.GradScaler(args...)`` is deprecated. Please use ``torch.amp.GradScaler("cuda", args...)`` instead.
"""
@deprecated(
"`torch.cuda.amp.GradScaler(args...)` is deprecated. "
"Please use `torch.amp.GradScaler('cuda', args...)` instead.",
category=FutureWarning,
)
def __init__(
self,
init_scale: float = 2.0**16,
@ -24,6 +19,9 @@ class GradScaler(torch.amp.GradScaler):
growth_interval: int = 2000,
enabled: bool = True,
) -> None:
warnings.warn(
"torch.cuda.amp.GradScaler(args...) is deprecated. Please use torch.amp.GradScaler('cuda', args...) instead."
)
super().__init__(
"cuda",
init_scale=init_scale,

17
torch/cuda/memory.py
View File

@ -9,7 +9,6 @@ import warnings
from inspect import signature
from typing import Any, Dict, Optional, Tuple, Union
from typing_extensions import deprecated
import torch
from torch import _C
@ -447,21 +446,21 @@ def max_memory_reserved(device: Union[Device, int] = None) -> int:
return memory_stats(device=device).get("reserved_bytes.all.peak", 0)
@deprecated(
"`torch.cuda.memory_cached` has been renamed to `torch.cuda.memory_reserved`",
category=FutureWarning,
)
def memory_cached(device: Union[Device, int] = None) -> int:
r"""Deprecated; see :func:`~torch.cuda.memory_reserved`."""
warnings.warn(
"torch.cuda.memory_cached has been renamed to torch.cuda.memory_reserved",
FutureWarning,
)
return memory_reserved(device=device)
@deprecated(
"`torch.cuda.max_memory_cached` has been renamed to `torch.cuda.max_memory_reserved`",
category=FutureWarning,
)
def max_memory_cached(device: Union[Device, int] = None) -> int:
r"""Deprecated; see :func:`~torch.cuda.max_memory_reserved`."""
warnings.warn(
"torch.cuda.max_memory_cached has been renamed to torch.cuda.max_memory_reserved",
FutureWarning,
)
return max_memory_reserved(device=device)

8
torch/cuda/nccl.py
View File

@ -89,9 +89,8 @@ def reduce(
)
else:
warnings.warn(
"`nccl.reduce` with an output tensor list is deprecated. "
"Please specify a single output tensor with argument 'output' instead instead.",
FutureWarning,
"nccl.reduce with an output tensor list is deprecated. "
"Please specify a single output tensor with argument 'output' instead instead."
)
_output = outputs[root]
elif not isinstance(output, torch.Tensor) and isinstance(
@ -100,8 +99,7 @@ def reduce(
# User called old API with positional arguments of list of output tensors.
warnings.warn(
"nccl.reduce with an output tensor list is deprecated. "
"Please specify a single output tensor.",
FutureWarning,
"Please specify a single output tensor."
)
_output = output[root]
else:

20
torch/distributed/_composable/fully_shard.py
View File

@ -1,5 +1,5 @@
import warnings
from typing import Callable, Iterable, Optional, Union
from typing_extensions import deprecated
import torch
import torch.distributed as dist
@ -38,13 +38,6 @@ from torch.distributed.fsdp.wrap import _Policy
@contract(state_cls=_FSDPState)
@deprecated(
"`torch.distributed._composable.fully_shard` is being deprecated. "
"You can continue to use the wrapper based FSDP. "
"See usage in: https://github.com/pytorch/pytorch/blob/main/torch/distributed/fsdp/fully_sharded_data_parallel.py. "
"`torch.distributed._composable.fully_shard` will be removed after PyTorch 2.5.",
category=FutureWarning,
)
def fully_shard(
module: nn.Module,
*,
@ -62,7 +55,16 @@ def fully_shard(
Optional[Iterable[torch.nn.Parameter]], Optional[Iterable[torch.nn.Module]]
] = None,
) -> nn.Module:
"""Applies ``FullyShardedDataParallel`` (FSDP) semantics to ``module``."""
"""
Applies ``FullyShardedDataParallel` (FSDP) semantics to ``module``.
"""
warnings.warn(
"``torch.distributed._composable.fully_shard`` is being deprecated."
"You can contintue to use the wrapper based FSDP."
"See usage in: https://github.com/pytorch/pytorch/blob/main/torch/distributed/fsdp/fully_sharded_data_parallel.py."
"``torch.distributed._composable.fully_shard`` will be removed after PyTorch 2.5."
)
torch._C._log_api_usage_once("torch.distributed.fully_shard")
# Enforce the new auto wrap policy
if policy is not None and not isinstance(policy, _Policy):

3
torch/distributed/_functional_collectives.py
View File

@ -766,8 +766,7 @@ def _resolve_group_name(group: RANK_TYPES, tag: str = "") -> str:
warnings.warn(
"The combination of ranks + tag as process group "
"identifier has been deprecated. Please switch to "
"using ProcessGroup, DeviceMesh, or group name instead.",
FutureWarning,
"using ProcessGroup, DeviceMesh, or group name instead."
)
return c10d._resolve_group_name_by_ranks_and_tag(cast(List[int], group), tag)
else:

15
torch/distributed/_shard/checkpoint/__init__.py
View File

@ -5,15 +5,8 @@ import torch
import warnings
from torch.distributed.checkpoint import * # noqa: F403
with warnings.catch_warnings():
warnings.simplefilter("always")
warnings.warn(
"`torch.distributed._shard.checkpoint` will be deprecated, "
"use `torch.distributed.checkpoint` instead",
DeprecationWarning,
stacklevel=2,
)
warnings.warn(
"torch.distributed._shard.checkpoint will be deprecated, use torch.distributed.checkpoint instead",
DeprecationWarning
)
sys.modules['torch.distributed._shard.checkpoint'] = torch.distributed.checkpoint

11
torch/distributed/_shard/sharded_tensor/api.py
View File

@ -10,7 +10,6 @@ from typing import (
cast,
TYPE_CHECKING,
)
from typing_extensions import deprecated
import copy
import warnings
from functools import reduce
@ -397,7 +396,7 @@ class ShardedTensor(ShardedTensorBase):
return reduce(operator.mul, shard_md.shard_sizes) # type: ignore[attr-defined]
if enforce_dtype:
warnings.warn("`enforce_dtype` is deprecated. Please use `dtype` instead.", FutureWarning)
warnings.warn("enforce_dtype is deprecated. Please use dtype instead.")
rank = dist.get_rank(self._process_group)
full_size = self.metadata().size
@ -738,7 +737,6 @@ class ShardedTensor(ShardedTensorBase):
return sharded_tensor
@classmethod
@deprecated(DEPRECATE_MSG, category=FutureWarning)
def _init_from_local_tensor(
cls,
local_tensor: torch.Tensor,
@ -803,6 +801,8 @@ class ShardedTensor(ShardedTensorBase):
We fully rely on the user to ensure local tensor is sharded based on the
sharding spec.
"""
warnings.warn(DEPRECATE_MSG)
if not local_tensor.is_contiguous():
raise ValueError('local_tensor is not a contiguous Tensor.')
@ -980,7 +980,6 @@ class ShardedTensor(ShardedTensorBase):
"""
return self._sharding_spec
@deprecated(DEPRECATE_MSG, category=FutureWarning)
def reshard(self, resharding_spec: shard_spec.ShardingSpec) -> ShardedTensor:
"""
Reshard a sharded tensor given the ``resharding_spec``. For now, we only support
@ -1051,6 +1050,8 @@ class ShardedTensor(ShardedTensorBase):
tensor([[3], [3], [5], [5], [7], [7], [9], [9]]) # Rank 2
tensor([[4], [4], [6], [6], [8], [8], [10], [10]]) # Rank 3
"""
warnings.warn(DEPRECATE_MSG)
if (
not isinstance(resharding_spec, shard_spec.ChunkShardingSpec) or
not isinstance(self._sharding_spec, shard_spec.ChunkShardingSpec)
@ -1095,7 +1096,6 @@ class ShardedTensor(ShardedTensorBase):
return self.local_shards()[0].tensor
@classmethod
@deprecated(DEPRECATE_MSG, category=FutureWarning)
def __torch_function__(cls, func, types, args=(), kwargs=None):
def dispatch(st: ShardedTensor, func: Callable):
# Dispatch to custom user provided op first if it exists.
@ -1120,6 +1120,7 @@ class ShardedTensor(ShardedTensorBase):
f"torch function '{func.__name__}', with args: {args} and "
f"kwargs: {kwargs} not supported for ShardedTensor!")
warnings.warn(DEPRECATE_MSG)
# Find ShardedTensor instance to get process_group and sharding_spec.
st_instance = None

14
torch/distributed/_sharded_tensor/__init__.py
View File

@ -5,14 +5,8 @@ import torch
import warnings
from torch.distributed._shard.sharded_tensor import * # noqa: F403
with warnings.catch_warnings():
warnings.simplefilter("always")
warnings.warn(
"`torch.distributed._sharded_tensor` will be deprecated, "
"use `torch.distributed._shard.sharded_tensor` instead",
DeprecationWarning,
stacklevel=2,
)
warnings.warn(
"torch.distributed._sharded_tensor will be deprecated, use torch.distributed._shard.sharded_tensor instead",
DeprecationWarning
)
sys.modules['torch.distributed._sharded_tensor'] = torch.distributed._shard.sharded_tensor

13
torch/distributed/_sharding_spec/__init__.py
View File

@ -5,15 +5,10 @@ import torch
import warnings
from torch.distributed._shard.sharding_spec import * # noqa: F403
with warnings.catch_warnings():
warnings.simplefilter("always")
warnings.warn(
"`torch.distributed._sharding_spec` will be deprecated, "
"use `torch.distributed._shard.sharding_spec` instead",
DeprecationWarning,
stacklevel=2,
)
warnings.warn(
"torch.distributed._sharding_spec will be deprecated, use torch.distributed._shard.sharding_spec instead",
DeprecationWarning
)
import torch.distributed._shard.sharding_spec as _sharding_spec
sys.modules['torch.distributed._sharding_spec'] = _sharding_spec

2
torch/distributed/_tensor/api.py
View File

@ -746,7 +746,6 @@ def distribute_module(
warnings.warn(
"Deprecating input_fn that takes two arguments (inputs, device_mesh), "
"please use input_fn that takes in (module, inputs, device_mesh) instead!",
FutureWarning,
)
module.register_forward_pre_hook(lambda _, inputs: input_fn(inputs, device_mesh)) # type: ignore[call-arg]
elif num_args == 3:
@ -766,7 +765,6 @@ def distribute_module(
warnings.warn(
"Deprecating output_fn that takes two arguments (inputs, device_mesh), "
"please use output_fn that takes in (module, inputs, device_mesh) instead!",
FutureWarning,
)
module.register_forward_hook(
lambda mod, inputs, outputs: output_fn(outputs, device_mesh) # type: ignore[call-arg]

1
torch/distributed/algorithms/_checkpoint/checkpoint_wrapper.py
View File

@ -233,7 +233,6 @@ def checkpoint_wrapper(
f"Please specify {CheckpointImpl.NO_REENTRANT} as "
f"{CheckpointImpl.REENTRANT} will soon be removed as "
"the default and eventually deprecated.",
FutureWarning,
stacklevel=1,
)
return CheckpointWrapper(

4
torch/distributed/algorithms/ddp_comm_hooks/default_hooks.py
View File

@ -85,7 +85,7 @@ def fp16_compress_hook(
decompressed_tensor.copy_(value)
return decompressed_tensor
if torch.compiler.is_compiling():
if torch._utils.is_compiling():
grad = dist._functional_collectives.all_reduce(
compressed_tensor, "sum", group_to_use
)
@ -134,7 +134,7 @@ def bf16_compress_hook(
decompressed_tensor.copy_(value)
return decompressed_tensor
if torch.compiler.is_compiling():
if torch._utils.is_compiling():
grad = dist._functional_collectives.all_reduce(
compressed_tensor, "sum", group_to_use
)

10
torch/distributed/checkpoint/state_dict_loader.py
View File

@ -1,7 +1,6 @@
import os
import warnings
from typing import Any, cast, Dict, Optional, Set, Union
from typing_extensions import deprecated
import torch
import torch.distributed as dist
@ -18,11 +17,6 @@ from .utils import _all_gather_keys, _api_bc_check, _DistWrapper, _profile
__all__ = ["load_state_dict", "load"]
@deprecated(
"`load_state_dict` is deprecated and will be removed in future versions. "
"Please use `load` instead.",
category=FutureWarning,
)
def load_state_dict(
state_dict: Dict[str, Any],
storage_reader: StorageReader,
@ -32,6 +26,10 @@ def load_state_dict(
planner: Optional[LoadPlanner] = None,
) -> None:
"""This method is deprecated. Please switch to 'load'."""
warnings.warn(
"'load_state_dict' is deprecated and will be removed in future versions. "
"Please use 'load' instead."
)
storage_reader.reset()
with _profile():
# TODO: test returning `load` here instead.

11
torch/distributed/checkpoint/state_dict_saver.py
View File

@ -3,7 +3,6 @@ import os
import warnings
from concurrent.futures import Future, ThreadPoolExecutor
from typing import cast, Optional, Union
from typing_extensions import deprecated
import torch
import torch.distributed as dist
@ -25,11 +24,6 @@ from .utils import _api_bc_check, _DistWrapper, _profile
__all__ = ["save_state_dict", "save", "async_save"]
@deprecated(
"`save_state_dict` is deprecated and will be removed in future versions."
"Please use `save` instead.",
category=FutureWarning,
)
def save_state_dict(
state_dict: STATE_DICT_TYPE,
storage_writer: StorageWriter,
@ -39,6 +33,11 @@ def save_state_dict(
planner: Optional[SavePlanner] = None,
) -> Metadata:
"""This method is deprecated. Please switch to 'save'."""
warnings.warn(
"'save_state_dict' is deprecated and will be removed in future versions."
"Please use 'save' instead."
)
storage_writer.reset()
# TODO: test returning `save` here instead.

65
torch/distributed/distributed_c10d.py
View File

@ -14,7 +14,6 @@ import warnings
from collections import namedtuple
from datetime import timedelta
from typing import Any, Callable, Dict, Optional, Tuple, Union, List, TYPE_CHECKING
from typing_extensions import deprecated
import torch
from torch._C._distributed_c10d import (
@ -365,12 +364,11 @@ class _reduce_op:
setattr(self, k, v)
self.__members__ = ReduceOp.RedOpType.__members__
@deprecated(
"`torch.distributed.reduce_op` is deprecated, "
"please use `torch.distributed.ReduceOp` instead",
category=FutureWarning,
)
def __getattribute__(self, key):
warnings.warn(
"torch.distributed.reduce_op is deprecated, please use "
"torch.distributed.ReduceOp instead"
)
return object.__getattribute__(self, key)
@ -677,8 +675,7 @@ def _get_pg_default_device(group: Optional[ProcessGroup] = None) -> torch.device
warnings.warn(
f"You are using a Backend {type(group)} as a ProcessGroup. "
"This usage is deprecated since PyTorch 2.0. Please use a public API "
"of PyTorch Distributed instead.",
FutureWarning,
"of PyTorch Distributed instead."
)
# Most users create Gloo with private API for object collectives
_world.pg_default_device[group] = torch.device("cpu")
@ -832,15 +829,13 @@ def get_global_rank(group: ProcessGroup, group_rank: int) -> int:
return rank
raise ValueError(f"Group rank {group_rank} is not part of group {group}")
# TODO: remove this once the ecosystem moves away from it.
@deprecated(
"`torch.distributed.distributed_c10d._get_global_rank` is deprecated, "
"please use `torch.distributed.distributed_c10d.get_global_rank` instead",
category=FutureWarning,
)
def _get_global_rank(group, rank) -> int:
"""Use get_global_rank as this method is deprecated."""
warnings.warn(
"torch.distributed.distributed_c10d._get_global_rank is deprecated "
"please use torch.distributed.distributed_c10d.get_global_rank instead"
)
return get_global_rank(group, rank)
@ -2291,12 +2286,6 @@ def all_reduce(tensor, op=ReduceOp.SUM, group=None, async_op=False):
work.wait()
@_exception_logger
@deprecated(
"`torch.distributed.all_reduce_coalesced` will be deprecated. If you must "
"use it, please revisit our documentation later at "
"https://pytorch.org/docs/main/distributed.html#collective-functions",
category=FutureWarning,
)
def all_reduce_coalesced(tensors, op=ReduceOp.SUM, group=None, async_op=False):
"""
WARNING: at this time individual shape checking is not implemented across nodes.
@ -2331,6 +2320,11 @@ def all_reduce_coalesced(tensors, op=ReduceOp.SUM, group=None, async_op=False):
None, if not async_op or if not part of the group.
"""
warnings.warn(
"torch.distributed.all_reduce_coalesced will be deprecated. If you must "
"use it, please revisit our documentation later at "
"https://pytorch.org/docs/main/distributed.html#collective-functions"
)
if isinstance(tensors, torch.Tensor):
tensors = [tensors]
_check_tensor_list(tensors, "tensor")
@ -3204,11 +3198,6 @@ def all_gather_into_tensor(output_tensor, input_tensor, group=None, async_op=Fal
@_exception_logger
@deprecated(
"`torch.distributed._all_gather_base` is a private function and will be deprecated. "
"Please use `torch.distributed.all_gather_into_tensor` instead.",
category=FutureWarning,
)
def _all_gather_base(output_tensor, input_tensor, group=None, async_op=False):
"""
Single tensor all gather. Gathers a single tensor from all ranks, and puts them in a single output tensor.
@ -3230,16 +3219,15 @@ def _all_gather_base(output_tensor, input_tensor, group=None, async_op=False):
`all_gather_into_tensor` instead.
"""
warnings.warn(
"torch.distributed._all_gather_base is a private function and will be "
"deprecated. Please use torch.distributed.all_gather_into_tensor "
"instead."
)
return all_gather_into_tensor(output_tensor, input_tensor, group, async_op)
@_exception_logger
@deprecated(
"`torch.distributed.all_gather_coalesced` will be deprecated. If you must use it, "
"please revisit our documentation later at "
"https://pytorch.org/docs/main/distributed.html#collective-functions",
category=FutureWarning,
)
def all_gather_coalesced(
output_tensor_lists, input_tensor_list, group=None, async_op=False
):
@ -3286,6 +3274,11 @@ def all_gather_coalesced(
performance improvements but users of this function should take extra care
to ensure that each node passes in tensors whose shapes match across nodes.
"""
warnings.warn(
"torch.distributed.all_gather_coalesced will be deprecated. If you must "
"use it, please revisit our documentation later at "
"https://pytorch.org/docs/main/distributed.html#collective-functions"
)
# We only check basic compatibility with C++ params here, C++ code will
# do shape and type checking.
if _rank_not_in_group(group):
@ -3615,11 +3608,6 @@ def reduce_scatter_tensor(output, input, op=ReduceOp.SUM, group=None, async_op=F
work.wait()
@deprecated(
"`torch.distributed._reduce_scatter_base` is a private function and will be deprecated. "
"Please use `torch.distributed.reduce_scatter_tensor` instead.",
category=FutureWarning,
)
def _reduce_scatter_base(output, input, op=ReduceOp.SUM, group=None, async_op=False):
"""
Reduces, then scatters a flattened tensor to all processes in a group.
@ -3640,6 +3628,11 @@ def _reduce_scatter_base(output, input, op=ReduceOp.SUM, group=None, async_op=Fa
`reduce_scatter_tensor` instead.
"""
warnings.warn(
"torch.distributed._reduce_scatter_base is a private function and will "
"be deprecated. Please use torch.distributed.reduce_scatter_tensor "
"instead."
)
return reduce_scatter_tensor(output, input, op, group, async_op)

12
torch/distributed/elastic/metrics/api.py
View File

@ -8,10 +8,10 @@
import abc
import time
import warnings
from collections import namedtuple
from functools import wraps
from typing import Dict, Optional
from typing_extensions import deprecated
__all__ = ['MetricsConfig', 'MetricHandler', 'ConsoleMetricHandler', 'NullMetricHandler', 'MetricStream',
'configure', 'getStream', 'prof', 'profile', 'put_metric', 'publish_metric', 'get_elapsed_time_ms',
@ -137,7 +137,6 @@ def prof(fn=None, group: str = "torchelastic"):
return wrap
@deprecated("Deprecated, use `@prof` instead", category=FutureWarning)
def profile(group=None):
"""
@profile decorator adds latency and success/failure metrics to any given function.
@ -149,6 +148,8 @@ def profile(group=None):
@metrics.profile("my_metric_group")
def some_function(<arguments>):
"""
warnings.warn("Deprecated, use @prof instead", DeprecationWarning)
def wrap(func):
@wraps(func)
def wrapper(*args, **kwargs):
@ -186,11 +187,10 @@ def put_metric(metric_name: str, metric_value: int, metric_group: str = "torchel
getStream(metric_group).add_value(metric_name, metric_value)
@deprecated(
"Deprecated, use `put_metric(metric_group)(metric_name, metric_value)` instead",
category=FutureWarning,
)
def publish_metric(metric_group: str, metric_name: str, metric_value: int):
warnings.warn(
"Deprecated, use put_metric(metric_group)(metric_name, metric_value) instead"
)
metric_stream = getStream(metric_group)
metric_stream.add_value(metric_name, metric_value)

3
torch/distributed/fsdp/_init_utils.py
View File

@ -446,8 +446,7 @@ def _init_core_state(
elif sharding_strategy == ShardingStrategy.NO_SHARD:
warnings.warn(
"The `NO_SHARD` sharding strategy is deprecated. If having issues, "
"please use `DistributedDataParallel` instead.",
FutureWarning,
"please use DistributedDataParallel instead.",
# Level 1 is here, level 2 is from `FullyShardedDataParallel`, and
# level 3 is from the true caller
stacklevel=3,

8
torch/distributed/fsdp/fully_sharded_data_parallel.py
View File

@ -1201,9 +1201,8 @@ class FullyShardedDataParallel(nn.Module, _FSDPState):
def _warn_optim_input(optim_input):
if optim_input is not None:
warnings.warn(
"The `optim_input` argument is deprecated and will be removed after PyTorch 1.13. "
"You may remove it from your code without changing its functionality.",
FutureWarning,
"The `optim_input` argument is deprecated and will be removed after PyTorch 1.13. You may remove it "
"from your code without changing its functionality."
)
@staticmethod
@ -1222,8 +1221,7 @@ class FullyShardedDataParallel(nn.Module, _FSDPState):
warnings.warn(
f"``FullyShardedDataParallel.{curr}``is being deprecated and is "
f"replaced by ``FullyShardedDataParallel.{new}``. "
f"``FullyShardedDataParallel.{curr}`` may be removed after PyTorch 2.2.",
FutureWarning,
f"``FullyShardedDataParallel.{curr}`` may be removed after PyTorch 2.2."
)
@staticmethod

22
torch/distributed/launch.py
View File

@ -159,7 +159,7 @@ will not pass ``--local-rank`` when you specify this flag.
"""
from typing_extensions import deprecated as _deprecated
import warnings
from torch.distributed.run import get_args_parser, run
@ -188,17 +188,17 @@ def launch(args):
run(args)
@_deprecated(
"The module torch.distributed.launch is deprecated\n"
"and will be removed in future. Use torchrun.\n"
"Note that --use-env is set by default in torchrun.\n"
"If your script expects `--local-rank` argument to be set, please\n"
"change it to read from `os.environ['LOCAL_RANK']` instead. See \n"
"https://pytorch.org/docs/stable/distributed.html#launch-utility for \n"
"further instructions\n",
category=FutureWarning,
)
def main(args=None):
warnings.warn(
"The module torch.distributed.launch is deprecated\n"
"and will be removed in future. Use torchrun.\n"
"Note that --use-env is set by default in torchrun.\n"
"If your script expects `--local-rank` argument to be set, please\n"
"change it to read from `os.environ['LOCAL_RANK']` instead. See \n"
"https://pytorch.org/docs/stable/distributed.html#launch-utility for \n"
"further instructions\n",
FutureWarning,
)
args = parse_args(args)
launch(args)

14
torch/distributed/optim/__init__.py
View File

@ -5,8 +5,6 @@ optimizer locally on the workers where the parameters live. The distributed
optimizer can use any of the local optimizer :ref:`optimizer-algorithms` to
apply the gradients on each worker.
"""
import warnings
import torch
from torch import optim
@ -26,15 +24,9 @@ from .functional_sgd import _FunctionalSGD
from .named_optimizer import _NamedOptimizer
from .utils import as_functional_optim
with warnings.catch_warnings():
warnings.simplefilter("always")
warnings.warn(
"`TorchScript` support for functional optimizers is deprecated "
"and will be removed in a future PyTorch release. "
"Consider using the `torch.compile` optimizer instead.",
DeprecationWarning,
stacklevel=2,
)
from warnings import warn
warn("TorchScript support for functional optimizers is"
"deprecated and will be removed in a future PyTorch release. Consider using the torch.compile optimizer instead.")
# DistributedOptimizer imports torch.distributed.rpc names, so gate availability
# based on RPC being available.

18
torch/distributed/pipeline/__init__.py
View File

@ -1,13 +1,7 @@
import warnings
with warnings.catch_warnings():
warnings.simplefilter("always")
warnings.warn(
"`torch.distributed.pipeline` is deprecated. For up-to-date pipeline parallel "
"implementation, please refer to the PiPPy library under the PyTorch "
"organization (Pipeline Parallelism for PyTorch): "
"https://github.com/pytorch/PiPPy",
DeprecationWarning,
stacklevel=2,
)
warnings.warn(
"torch.distributed.pipeline is deprecated. For up-to-date pipeline parallel "
"implementation, please refer to the PiPPy library under the PyTorch "
"organization (Pipeline Parallelism for PyTorch): "
"https://github.com/pytorch/PiPPy"
)

5
torch/distributed/tensor/parallel/_utils.py
View File

@ -22,10 +22,7 @@ def _deprecate_warnings(func_name: str, extra_msg: str) -> None:
"""
# TODO: Will follow up with dynamo POC to make warnings.warn working with dynamo.
if not is_torchdynamo_compiling():
warnings.warn(
f"{func_name} is deprecated and will be removed soon. {extra_msg}",
FutureWarning,
)
warnings.warn(f"{func_name} is deprecated and will be removed soon. {extra_msg}")
def _validate_tp_mesh_dim(

8
torch/distributions/distribution.py
View File

@ -1,6 +1,5 @@
import warnings
from typing import Any, Dict, Optional, Tuple
from typing_extensions import deprecated
import torch
from torch.distributions import constraints
@ -172,15 +171,14 @@ class Distribution:
"""
raise NotImplementedError
@deprecated(
"`sample_n(n)` will be deprecated. Use `sample((n,))` instead.",
category=FutureWarning,
)
def sample_n(self, n: int) -> torch.Tensor:
"""
Generates n samples or n batches of samples if the distribution
parameters are batched.
"""
warnings.warn(
"sample_n will be deprecated. Use .sample((n,)) instead", UserWarning
)
return self.sample(torch.Size((n,)))
def log_prob(self, value: torch.Tensor) -> torch.Tensor:

30
torch/fx/experimental/unification/multipledispatch/dispatcher.py
View File

@ -1,6 +1,5 @@
from warnings import warn
import inspect
from typing_extensions import deprecated
from .conflict import ordering, ambiguities, super_signature, AmbiguityWarning
from .utils import expand_tuples
from .variadic import Variadic, isvariadic
@ -28,21 +27,24 @@ def ambiguity_warn(dispatcher, ambiguities):
warn(warning_text(dispatcher.name, ambiguities), AmbiguityWarning)
@deprecated(
"`halt_ordering` is deprecated, you can safely remove this call.",
category=FutureWarning,
)
def halt_ordering():
"""Deprecated interface to temporarily disable ordering."""
"""Deprecated interface to temporarily disable ordering.
"""
warn(
'halt_ordering is deprecated, you can safely remove this call.',
DeprecationWarning,
)
@deprecated(
"`restart_ordering` is deprecated, if you would like to eagerly order the dispatchers, "
"you should call the `reorder()` method on each dispatcher.",
category=FutureWarning,
)
def restart_ordering(on_ambiguity=ambiguity_warn):
"""Deprecated interface to temporarily resume ordering."""
"""Deprecated interface to temporarily resume ordering.
"""
warn(
'restart_ordering is deprecated, if you would like to eagerly order'
'the dispatchers, you should call the ``reorder()`` method on each'
' dispatcher.',
DeprecationWarning,
)
def variadic_signature_matches_iter(types, full_signature):
@ -314,12 +316,14 @@ class Dispatcher:
result = self.funcs[signature]
yield result
@deprecated("`resolve()` is deprecated, use `dispatch(*types)`", category=FutureWarning)
def resolve(self, types):
""" Determine appropriate implementation for this type signature
.. deprecated:: 0.4.4
Use ``dispatch(*types)`` instead
"""
warn("resolve() is deprecated, use dispatch(*types)",
DeprecationWarning)
return self.dispatch(*types)
def __getstate__(self):

10
torch/hub.py
View File

@ -13,7 +13,6 @@ import warnings
import zipfile
from pathlib import Path
from typing import Dict, Optional, Any
from typing_extensions import deprecated
from urllib.error import HTTPError, URLError
from urllib.request import urlopen, Request
from urllib.parse import urlparse # noqa: F401
@ -681,13 +680,10 @@ def _is_legacy_zip_format(filename: str) -> bool:
return False
@deprecated(
'Falling back to the old format < 1.6. This support will be '
'deprecated in favor of default zipfile format introduced in 1.6. '
'Please redo torch.save() to save it in the new zipfile format.',
category=FutureWarning,
)
def _legacy_zip_load(filename: str, model_dir: str, map_location: MAP_LOCATION, weights_only: bool) -> Dict[str, Any]:
warnings.warn('Falling back to the old format < 1.6. This support will be '
'deprecated in favor of default zipfile format introduced in 1.6. '
'Please redo torch.save() to save it in the new zipfile format.')
# Note: extractall() defaults to overwrite file if exists. No need to clean up beforehand.
# We deliberately don't handle tarfile here since our legacy serialization format was in tar.
# E.g. resnet18-5c106cde.pth which is widely used.

4
torch/jit/_script.py
View File

@ -1094,9 +1094,7 @@ def _script_impl(
if optimize is not None:
warnings.warn(
"`optimize` is deprecated and has no effect. "
"Use `with torch.jit.optimized_execution()` instead",
FutureWarning,
"`optimize` is deprecated and has no effect. Use `with torch.jit.optimized_execution() instead"
)
# No-op for modules, functions, class instances that are already scripted

8
torch/jit/_trace.py
View File

@ -978,9 +978,7 @@ def trace(
return func
if optimize is not None:
warnings.warn(
"`optimize` is deprecated and has no effect. "
"Use `with torch.jit.optimized_execution()` instead",
FutureWarning,
"`optimize` is deprecated and has no effect. Use `with torch.jit.optimized_execution() instead"
)
from torch._utils_internal import (
@ -1187,9 +1185,7 @@ def trace_module(
return mod
if optimize is not None:
warnings.warn(
"`optimize` is deprecated and has no effect. "
"Use `with torch.jit.optimized_execution()` instead",
FutureWarning,
"`optimize` is deprecated and has no effect. Use `with torch.jit.optimized_execution() instead"
)
var_lookup_fn = _create_interpreter_name_lookup_fn(0)

12
torch/library.py
View File

@ -1,6 +1,5 @@
from ._ops import OpOverload
from typing import Any, Optional, Set, List, Union, Callable, Tuple, Dict, Sequence
from typing_extensions import deprecated
import traceback
import torch
import weakref
@ -9,6 +8,7 @@ import inspect
import re
import contextlib
import sys
import warnings
from torch._library.custom_ops import custom_op, _maybe_get_opdef, device_types_t, CustomOpDef
import torch._library as _library
@ -451,15 +451,15 @@ def _(lib: Library, name, dispatch_key=""):
return wrap
@deprecated(
"`torch.library.impl_abstract` was renamed to `torch.library.register_fake`. Please use that "
"instead; we will remove `torch.library.impl_abstract` in a future version of PyTorch.",
category=FutureWarning,
)
def impl_abstract(qualname, func=None, *, lib=None, _stacklevel=1):
r"""This API was renamed to :func:`torch.library.register_fake` in PyTorch 2.4.
Please use that instead.
"""
warnings.warn("torch.library.impl_abstract was renamed to "
"torch.library.register_fake. Please use that instead; "
"we will remove torch.library.impl_abstract in a future "
"version of PyTorch.",
DeprecationWarning, stacklevel=2)
if func is not None:
_stacklevel = _stacklevel + 1
return register_fake(qualname, func, lib=lib, _stacklevel=_stacklevel)

2
torch/multiprocessing/spawn.py
View File

@ -277,5 +277,5 @@ def spawn(fn, args=(), nprocs=1, join=True, daemon=False, start_method="spawn"):
"To use a different start_method use:\n\t\t"
" torch.multiprocessing.start_processes(...)"
)
warnings.warn(msg, FutureWarning)
warnings.warn(msg)
return start_processes(fn, args, nprocs, join, daemon, start_method="spawn")

18
torch/nn/functional.py
View File

@ -1818,8 +1818,7 @@ See :class:`~torch.nn.Softplus` for more details.
def _get_softmax_dim(name: str, ndim: int, stacklevel: int) -> int:
warnings.warn(
f"Implicit dimension choice for {name} has been deprecated. "
"Change the call to include dim=X as an argument.",
f"Implicit dimension choice for {name} has been deprecated. Change the call to include dim=X as an argument.",
stacklevel=stacklevel,
)
if ndim == 0 or ndim == 1 or ndim == 3:
@ -3824,10 +3823,7 @@ def upsample(input, size=None, scale_factor=None, mode="nearest", align_corners=
affects the outputs.
"""
warnings.warn(
"`nn.functional.upsample` is deprecated. "
"Use `nn.functional.interpolate` instead.",
)
warnings.warn("nn.functional.upsample is deprecated. Use nn.functional.interpolate instead.")
return interpolate(input, size, scale_factor, mode, align_corners)
@ -4147,10 +4143,7 @@ def upsample_nearest(input, size=None, scale_factor=None): # noqa: F811
{backward_reproducibility_note}
"""
# DeprecationWarning is ignored by default
warnings.warn(
"`nn.functional.upsample_nearest` is deprecated. "
"Use `nn.functional.interpolate` instead.",
)
warnings.warn("nn.functional.upsample_nearest is deprecated. Use nn.functional.interpolate instead.")
return interpolate(input, size, scale_factor, mode="nearest")
@ -4206,10 +4199,7 @@ def upsample_bilinear(input, size=None, scale_factor=None): # noqa: F811
{backward_reproducibility_note}
"""
# DeprecationWarning is ignored by default
warnings.warn(
"`nn.functional.upsample_bilinear` is deprecated. "
"Use `nn.functional.interpolate` instead.",
)
warnings.warn("nn.functional.upsample_bilinear is deprecated. Use nn.functional.interpolate instead.")
return interpolate(input, size, scale_factor, mode="bilinear", align_corners=True)

6
torch/nn/init.py
View File

@ -599,11 +599,7 @@ def _make_deprecate(meth):
old_name = new_name[:-1]
def deprecated_init(*args, **kwargs):
warnings.warn(
f"`nn.init.{old_name}` is now deprecated in favor of `nn.init.{new_name}`.",
FutureWarning,
stacklevel=2,
)
warnings.warn(f"nn.init.{old_name} is now deprecated in favor of nn.init.{new_name}.", stacklevel=2)
return meth(*args, **kwargs)
deprecated_init.__doc__ = fr"""

10
torch/nn/modules/activation.py
View File

@ -219,16 +219,10 @@ class Hardtanh(Module):
) -> None:
super().__init__()
if min_value is not None:
warnings.warn(
"keyword argument `min_value` is deprecated and rename to `min_val`",
FutureWarning,
)
warnings.warn("keyword argument min_value is deprecated and rename to min_val")
min_val = min_value
if max_value is not None:
warnings.warn(
"keyword argument `max_value` is deprecated and rename to `max_val`",
FutureWarning,
)
warnings.warn("keyword argument max_value is deprecated and rename to max_val")
max_val = max_value
self.min_val = min_val

11
torch/nn/modules/container.py
View File

@ -1,3 +1,4 @@
import warnings
from collections import OrderedDict, abc as container_abcs
from itertools import chain, islice
import operator
@ -9,7 +10,6 @@ from torch._jit_internal import _copy_to_script_wrapper
from typing import Any, Dict, Iterable, Iterator, Mapping, Optional, overload, Tuple, TypeVar, Union
from typing_extensions import Self
from typing_extensions import deprecated
__all__ = ['Container', 'Sequential', 'ModuleList', 'ModuleDict', 'ParameterList', 'ParameterDict']
@ -29,14 +29,13 @@ def _addindent(s_, numSpaces):
return s
@deprecated(
"`nn.Container` is deprecated. "
"All of it's functionality is now implemented in `nn.Module`. Subclass that instead.",
category=FutureWarning,
)
class Container(Module):
def __init__(self, **kwargs: Any) -> None:
super().__init__()
# DeprecationWarning is ignored by default <sigh>
warnings.warn("nn.Container is deprecated. All of it's functionality "
"is now implemented in nn.Module. Subclass that instead.")
for key, value in kwargs.items():
self.add_module(key, value)

15
torch/nn/modules/conv.py
View File

@ -1,4 +1,5 @@
import math
import warnings
import torch
from torch import Tensor
@ -12,7 +13,6 @@ from torch._torch_docs import reproducibility_notes
from ..common_types import _size_1_t, _size_2_t, _size_3_t
from typing import Optional, List, Tuple, Union
from typing_extensions import deprecated
__all__ = ['Conv1d', 'Conv2d', 'Conv3d', 'ConvTranspose1d', 'ConvTranspose2d', 'ConvTranspose3d',
'LazyConv1d', 'LazyConv2d', 'LazyConv3d', 'LazyConvTranspose1d', 'LazyConvTranspose2d',
@ -40,6 +40,9 @@ convolution_notes = \
:math:`(C_\text{in}=C_\text{in}, C_\text{out}=C_\text{in} \times \text{K}, ..., \text{groups}=C_\text{in})`."""} # noqa: B950
class _ConvNd(Module):
__constants__ = ['stride', 'padding', 'dilation', 'groups',
@ -607,6 +610,7 @@ class Conv3d(_ConvNd):
return self._conv_forward(input, self.weight, self.bias)
class _ConvTransposeNd(_ConvNd):
def __init__(self, in_channels, out_channels, kernel_size, stride,
padding, dilation, transposed, output_padding,
@ -1117,13 +1121,10 @@ class ConvTranspose3d(_ConvTransposeNd):
# `_ConvTransposeNd` is really not a mixin anymore (but multiple inheritance as
# above would still work).
class _ConvTransposeMixin(_ConvTransposeNd):
@deprecated(
"`_ConvTransposeMixin` is a deprecated internal class. "
"Please consider using public APIs.",
category=FutureWarning,
)
def __init__(self, *args, **kwargs):
warnings.warn(
"_ConvTransposeMixin is a deprecated internal class. "
"Please consider using public APIs.")
super().__init__(*args, **kwargs)

12
torch/nn/modules/loss.py
View File

@ -1,3 +1,5 @@
import warnings
from .distance import PairwiseDistance
from .module import Module
from .. import functional as F
@ -5,7 +7,6 @@ from .. import _reduction as _Reduction
from torch import Tensor
from typing import Callable, Optional
from typing_extensions import deprecated
__all__ = ['L1Loss', 'NLLLoss', 'NLLLoss2d', 'PoissonNLLLoss', 'GaussianNLLLoss', 'KLDivLoss',
'MSELoss', 'BCELoss', 'BCEWithLogitsLoss', 'HingeEmbeddingLoss', 'MultiLabelMarginLoss',
@ -217,15 +218,12 @@ class NLLLoss(_WeightedLoss):
return F.nll_loss(input, target, weight=self.weight, ignore_index=self.ignore_index, reduction=self.reduction)
@deprecated(
"`NLLLoss2d` has been deprecated. "
"Please use `NLLLoss` instead as a drop-in replacement and see "
"https://pytorch.org/docs/main/nn.html#torch.nn.NLLLoss for more details.",
category=FutureWarning,
)
class NLLLoss2d(NLLLoss):
def __init__(self, weight: Optional[Tensor] = None, size_average=None, ignore_index: int = -100,
reduce=None, reduction: str = 'mean') -> None:
warnings.warn("NLLLoss2d has been deprecated. "
"Please use NLLLoss instead as a drop-in replacement and see "
"https://pytorch.org/docs/main/nn.html#torch.nn.NLLLoss for more details.")
super().__init__(weight, size_average, ignore_index, reduce, reduction)

19
torch/nn/modules/module.py
View File

@ -1346,8 +1346,7 @@ class Module:
warnings.warn("Using non-full backward hooks on a Module that does not return a "
"single Tensor or a tuple of Tensors is deprecated and will be removed "
"in future versions. This hook will be missing some of the grad_output. "
"Please use register_full_backward_hook to get the documented behavior.",
FutureWarning)
"Please use register_full_backward_hook to get the documented behavior.")
return
else:
result = (result,)
@ -1357,8 +1356,7 @@ class Module:
warnings.warn("Using non-full backward hooks on a Module that does not take as input a "
"single Tensor or a tuple of Tensors is deprecated and will be removed "
"in future versions. This hook will be missing some of the grad_input. "
"Please use register_full_backward_hook to get the documented behavior.",
FutureWarning)
"Please use register_full_backward_hook to get the documented behavior.")
return
else:
inputs = (inputs,)
@ -1368,13 +1366,11 @@ class Module:
if len(out_grad_fn) == 0 or (len(out_grad_fn) == 1 and grad_fn not in out_grad_fn):
warnings.warn("Using a non-full backward hook when outputs are nested in python data structure "
"is deprecated and will be removed in future versions. This hook will be missing "
"some grad_output.",
FutureWarning)
"some grad_output.")
elif len(out_grad_fn) > 1:
warnings.warn("Using a non-full backward hook when outputs are generated by different autograd Nodes "
"is deprecated and will be removed in future versions. This hook will be missing "
"some grad_output. Please use register_full_backward_hook to get the documented behavior.",
FutureWarning)
"some grad_output. Please use register_full_backward_hook to get the documented behavior.")
else:
# At this point the grad_output part of the hook will most likely be correct
inputs_grad_fn = {i.grad_fn for i in inputs if i.grad_fn is not None}
@ -1385,8 +1381,7 @@ class Module:
warnings.warn("Using a non-full backward hook when the forward contains multiple autograd Nodes "
"is deprecated and will be removed in future versions. This hook will be missing "
"some grad_input. Please use register_full_backward_hook to get the documented "
"behavior.",
FutureWarning)
"behavior.")
def register_forward_pre_hook(
self,
@ -1910,9 +1905,7 @@ class Module:
warnings.warn(
"Positional args are being deprecated, use kwargs instead. Refer to "
"https://pytorch.org/docs/main/generated/torch.nn.Module.html#torch.nn.Module.state_dict"
" for details.",
FutureWarning,
)
" for details.")
if destination is None:
destination = OrderedDict()

6
torch/nn/modules/rnn.py
View File

@ -3,7 +3,6 @@ import warnings
import numbers
import weakref
from typing import List, Tuple, Optional, overload
from typing_extensions import deprecated
import torch
from torch import Tensor
@ -25,11 +24,8 @@ def _apply_permutation(tensor: Tensor, permutation: Tensor, dim: int = 1) -> Ten
return tensor.index_select(dim, permutation)
@deprecated(
"`apply_permutation` is deprecated, please use `tensor.index_select(dim, permutation)` instead",
category=FutureWarning,
)
def apply_permutation(tensor: Tensor, permutation: Tensor, dim: int = 1) -> Tensor:
warnings.warn("apply_permutation is deprecated, please use tensor.index_select(dim, permutation) instead")
return _apply_permutation(tensor, permutation, dim)

11
torch/nn/parallel/__init__.py
View File

@ -1,5 +1,3 @@
from typing_extensions import deprecated
from .parallel_apply import parallel_apply
from .replicate import replicate
from .data_parallel import DataParallel, data_parallel
@ -9,11 +7,8 @@ from .distributed import DistributedDataParallel
__all__ = ['replicate', 'scatter', 'parallel_apply', 'gather', 'data_parallel',
'DataParallel', 'DistributedDataParallel']
@deprecated(
"`torch.nn.parallel.DistributedDataParallelCPU` is deprecated, "
"please use `torch.nn.parallel.DistributedDataParallel` instead.",
category=FutureWarning,
)
def DistributedDataParallelCPU(*args, **kwargs):
import warnings
warnings.warn("torch.nn.parallel.DistributedDataParallelCPU is deprecated, "
"please use torch.nn.parallel.DistributedDataParallel instead.")
return DistributedDataParallel(*args, **kwargs)

4
torch/nn/parallel/comm.py
View File

@ -226,9 +226,7 @@ def gather(tensors, dim=0, destination=None, *, out=None):
if destination == -1:
warnings.warn(
'Using -1 to represent CPU tensor is deprecated. Please use a '
'device object or string instead, e.g., "cpu".',
FutureWarning,
)
'device object or string instead, e.g., "cpu".')
destination = _get_device_index(destination, allow_cpu=True, optional=True)
return torch._C._gather(tensors, dim, destination)
else:

7
torch/nn/parallel/distributed.py
View File

@ -771,8 +771,7 @@ class DistributedDataParallel(Module, Joinable):
# do not receive gradients.
warnings.warn(
"The `check_reduction` argument in `DistributedDataParallel` "
"module is deprecated. Please avoid using it.",
FutureWarning,
"module is deprecated. Please avoid using it."
)
# Check that a module does not have Uninitialized parameters
@ -1467,7 +1466,7 @@ class DistributedDataParallel(Module, Joinable):
def _should_disable_cpp_reducer(self) -> bool:
return self._use_python_reducer and (
torch.compiler.is_compiling() or self._force_to_disable_cpp_reducer
torch._utils.is_compiling() or self._force_to_disable_cpp_reducer
)
def _pre_forward(self, *inputs, **kwargs):
@ -1480,7 +1479,7 @@ class DistributedDataParallel(Module, Joinable):
h.remove()
self._accum_grad_hooks.clear()
if not self._lazy_init_ran and not torch.compiler.is_compiling():
if not self._lazy_init_ran and not torch._utils.is_compiling():
self._lazy_init()
if self._delay_all_reduce_all_params:

8
torch/nn/parallel/scatter_gather.py
View File

@ -1,17 +1,13 @@
import torch
from typing import Any, Dict, List, Optional, Sequence, Tuple, TypeVar, Union, overload
from typing_extensions import deprecated
from ._functions import Scatter, Gather
import warnings
__all__ = ['scatter', 'scatter_kwargs', 'gather']
@deprecated(
"`is_namedtuple` is deprecated, please use the python checks instead",
category=FutureWarning,
)
def is_namedtuple(obj: Any) -> bool:
# Check if type was created from collections.namedtuple or a typing.NamedTuple.
warnings.warn("is_namedtuple is deprecated, please use the python checks instead")
return _is_namedtuple(obj)
def _is_namedtuple(obj: Any) -> bool:

9
torch/nn/utils/clip_grad.py
View File

@ -1,6 +1,6 @@
import warnings
import functools
from typing import Union, Iterable, List, Dict, Tuple, Optional, cast
from typing_extensions import deprecated
import torch
from torch import Tensor
@ -99,11 +99,6 @@ def clip_grad_norm_(
return total_norm
@deprecated(
"`torch.nn.utils.clip_grad_norm` is now deprecated "
"in favor of `torch.nn.utils.clip_grad_norm_`.",
category=FutureWarning,
)
def clip_grad_norm(
parameters: _tensor_or_tensors, max_norm: float, norm_type: float = 2.,
error_if_nonfinite: bool = False, foreach: Optional[bool] = None) -> torch.Tensor:
@ -113,6 +108,8 @@ def clip_grad_norm(
This method is now deprecated in favor of
:func:`torch.nn.utils.clip_grad_norm_`.
"""
warnings.warn("torch.nn.utils.clip_grad_norm is now deprecated in favor "
"of torch.nn.utils.clip_grad_norm_.", stacklevel=2)
return clip_grad_norm_(parameters, max_norm, norm_type, error_if_nonfinite, foreach)

14
torch/nn/utils/stateless.py
View File

@ -1,7 +1,7 @@
import contextlib
import warnings
from collections import defaultdict
from typing import Any, Dict, Iterator, Optional, Set, Tuple, Union
from typing_extensions import deprecated
import torch
from torch import Tensor
@ -148,12 +148,6 @@ def _reparametrize_module(
)
@deprecated(
"`torch.nn.utils.stateless.functional_call` is deprecated as of PyTorch 2.0 "
"and will be removed in a future version of PyTorch. "
"Please use `torch.func.functional_call` instead which is a drop-in replacement.",
category=FutureWarning,
)
def functional_call(
module: "torch.nn.Module",
parameters_and_buffers: Dict[str, Tensor],
@ -222,6 +216,12 @@ def functional_call(
Returns:
Any: the result of calling ``module``.
"""
warnings.warn(
"This API is deprecated as of PyTorch 2.0 and will be removed in a future "
"version of PyTorch. Please use torch.func.functional_call instead "
"which is a drop-in replacement for this API."
)
return _functional_call(
module,
parameters_and_buffers,

9
torch/nn/utils/weight_norm.py
View File

@ -2,7 +2,7 @@ r"""Weight Normalization from https://arxiv.org/abs/1602.07868."""
from torch.nn.parameter import Parameter, UninitializedParameter
from torch import _weight_norm, norm_except_dim
from typing import Any, TypeVar
from typing_extensions import deprecated
import warnings
from ..modules import Module
__all__ = ['WeightNorm', 'weight_norm', 'remove_weight_norm']
@ -24,12 +24,9 @@ class WeightNorm:
return _weight_norm(v, g, self.dim)
@staticmethod
@deprecated(
"`torch.nn.utils.weight_norm` is deprecated "
"in favor of `torch.nn.utils.parametrizations.weight_norm`.",
category=FutureWarning,
)
def apply(module, name: str, dim: int) -> 'WeightNorm':
warnings.warn("torch.nn.utils.weight_norm is deprecated in favor of torch.nn.utils.parametrizations.weight_norm.")
for hook in module._forward_pre_hooks.values():
if isinstance(hook, WeightNorm) and hook.name == name:
raise RuntimeError(f"Cannot register two weight_norm hooks on the same parameter {name}")

6
torch/optim/adadelta.py
View File

@ -254,7 +254,7 @@ def _single_tensor_adadelta(
has_complex: bool,
):
# If compiling, the compiler will handle cudagraph checks, see note [torch.compile x capturable]
if not torch.compiler.is_compiling() and capturable:
if not torch._utils.is_compiling() and capturable:
capturable_supported_devices = _get_capturable_supported_devices(
supports_xla=False
)
@ -310,7 +310,7 @@ def _multi_tensor_adadelta(
assert not differentiable, "_foreach ops don't support autograd"
# If compiling, the compiler will handle cudagraph checks, see note [torch.compile x capturable]
if not torch.compiler.is_compiling() and capturable:
if not torch._utils.is_compiling() and capturable:
capturable_supported_devices = _get_capturable_supported_devices(
supports_xla=False
)
@ -413,7 +413,7 @@ def adadelta(
# this check is slow during compilation, so we skip it
# if it's strictly needed we can add this check back in dynamo
if not torch.compiler.is_compiling() and not all(
if not torch._utils.is_compiling() and not all(
isinstance(t, torch.Tensor) for t in state_steps
):
raise RuntimeError(

6
torch/optim/adam.py
View File

@ -353,7 +353,7 @@ def _single_tensor_adam(
step_t = state_steps[i]
# If compiling, the compiler will handle cudagraph checks, see note [torch.compile x capturable]
if not torch.compiler.is_compiling() and capturable:
if not torch._utils.is_compiling() and capturable:
capturable_supported_devices = _get_capturable_supported_devices()
assert (
param.device.type == step_t.device.type
@ -466,7 +466,7 @@ def _multi_tensor_adam(
)
# If compiling, the compiler will handle cudagraph checks, see note [torch.compile x capturable]
if not torch.compiler.is_compiling() and capturable:
if not torch._utils.is_compiling() and capturable:
capturable_supported_devices = _get_capturable_supported_devices(
supports_xla=False
)
@ -743,7 +743,7 @@ def adam(
# this check is slow during compilation, so we skip it
# if it's strictly needed we can add this check back in dynamo
if not torch.compiler.is_compiling() and not all(
if not torch._utils.is_compiling() and not all(
isinstance(t, torch.Tensor) for t in state_steps
):
raise RuntimeError(

6
torch/optim/adamax.py
View File

@ -243,7 +243,7 @@ def _single_tensor_adamax(
step_t = state_steps[i]
# If compiling, the compiler will handle cudagraph checks, see note [torch.compile x capturable]
if not torch.compiler.is_compiling() and capturable:
if not torch._utils.is_compiling() and capturable:
capturable_supported_devices = _get_capturable_supported_devices()
assert (
param.device.type == step_t.device.type
@ -315,7 +315,7 @@ def _multi_tensor_adamax(
return
# If compiling, the compiler will handle cudagraph checks, see note [torch.compile x capturable]
if not torch.compiler.is_compiling() and capturable:
if not torch._utils.is_compiling() and capturable:
capturable_supported_devices = _get_capturable_supported_devices(
supports_xla=False
)
@ -424,7 +424,7 @@ def adamax(
See :class:`~torch.optim.Adamax` for details.
"""
if not torch.compiler.is_compiling() and not all(
if not torch._utils.is_compiling() and not all(
isinstance(t, torch.Tensor) for t in state_steps
):
raise RuntimeError(

6
torch/optim/adamw.py
View File

@ -354,7 +354,7 @@ def _single_tensor_adamw(
step_t = state_steps[i]
# If compiling, the compiler will handle cudagraph checks, see note [torch.compile x capturable]
if not torch.compiler.is_compiling() and capturable:
if not torch._utils.is_compiling() and capturable:
capturable_supported_devices = _get_capturable_supported_devices()
assert (
param.device.type == step_t.device.type
@ -467,7 +467,7 @@ def _multi_tensor_adamw(
)
# If compiling, the compiler will handle cudagraph checks, see note [torch.compile x capturable]
if not torch.compiler.is_compiling() and capturable:
if not torch._utils.is_compiling() and capturable:
capturable_supported_devices = _get_capturable_supported_devices(
supports_xla=False
)
@ -728,7 +728,7 @@ def adamw(
See :class:`~torch.optim.AdamW` for details.
"""
if not torch.compiler.is_compiling() and not all(
if not torch._utils.is_compiling() and not all(
isinstance(t, torch.Tensor) for t in state_steps
):
raise RuntimeError(

4
torch/optim/asgd.py
View File

@ -214,7 +214,7 @@ def _single_tensor_asgd(
step_t = state_steps[i]
# If compiling, the compiler will handle cudagraph checks, see note [torch.compile x capturable]
if not torch.compiler.is_compiling() and capturable:
if not torch._utils.is_compiling() and capturable:
capturable_supported_devices = _get_capturable_supported_devices()
assert (
param.device.type
@ -287,7 +287,7 @@ def _multi_tensor_asgd(
assert not differentiable, "_foreach ops don't support autograd"
# If compiling, the compiler will handle cudagraph checks, see note [torch.compile x capturable]
if not torch.compiler.is_compiling() and capturable:
if not torch._utils.is_compiling() and capturable:
capturable_supported_devices = _get_capturable_supported_devices(
supports_xla=False
)

4
torch/optim/nadam.py
View File

@ -304,7 +304,7 @@ def _single_tensor_nadam(
exp_avg_sq = torch.view_as_real(exp_avg_sq)
# If compiling, the compiler will handle cudagraph checks, see note [torch.compile x capturable]
if not torch.compiler.is_compiling() and capturable:
if not torch._utils.is_compiling() and capturable:
capturable_supported_devices = _get_capturable_supported_devices()
assert (
param.device.type == mu_product.device.type == step_t.device.type
@ -390,7 +390,7 @@ def _multi_tensor_nadam(
assert not differentiable, "_foreach ops don't support autograd"
# If compiling, the compiler will handle cudagraph checks, see note [torch.compile x capturable]
if not torch.compiler.is_compiling() and capturable:
if not torch._utils.is_compiling() and capturable:
capturable_supported_devices = _get_capturable_supported_devices(
supports_xla=False
)

Some files were not shown because too many files have changed in this diff Show More