Fixes#104484
For >= 3.10, we use `inspect.get_annotations` instead of `getattr(.., "__annotations__")`. [Docs](https://docs.python.org/3/library/inspect.html#inspect.get_annotations) say that get_annotations() "Ignores inherited annotations on classes. If a class doesn’t have its own annotations dict, returns an empty dict.". In practice though, this doesn't seem always true; until you call inspect.getmembers it seems like you still get inherited annotations. In particular, this means that if you script a certain type twice, the first time it may pass scripting but on the second try it may not pass scripting.
This PR adds a more comprehensive handling of get_annotations by recursively reading the annotations of the base types. (TorchScript doesn't officially support this; but since it worked in <3.10, it's now breaking internal stuff as python gets upgraded to 3.10)
Verified in #104486 that the test does actually fail before the changes in this PR were added.
Differential Revision: [D47163891](https://our.internmc.facebook.com/intern/diff/D47163891)
Pull Request resolved: https://github.com/pytorch/pytorch/pull/104485
Approved by: https://github.com/eellison
This PR optimizes the guards overhead introduced by dynamo tracing module forward hooks.
It can and maybe should be followed by a wider change proposed by @voznesenskym to optimize specialized nnmodules by 'observing' any user mutations and directly invalidating the root guard, obviating the need to install other nnmodule guards. (But this observer change seems more involved...)
Idea: maintain a flag, and keep it up to date whenever adding or
removing hooks. Use the flag rather than dict checks to enter the call fast path.
- need to extend RemovableHandle to keep a ref to nnModule so it can update the flag on removal.
- also need to handle the flag in ScriptModule which still uses the python call impl when called from python.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/95931
Approved by: https://github.com/ezyang, https://github.com/voznesenskym
Applies the remaining flake8-comprehension fixes and checks. This changes replace all remaining unnecessary generator expressions with list/dict/set comprehensions which are more succinct, performant, and better supported by our torch.jit compiler. It also removes useless generators such as 'set(a for a in b)`, resolving it into just the set call.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/94676
Approved by: https://github.com/ezyang
closes#35643
This PR is mostly borrowed from #82042. Thanks @Padarn for implementing
the first version and debugging into the errors.
Based on the discussion in #82042 this PR adds a with_kwargs
argument to register_forward_pre_hook and register_forward_hook
methods. When the arg is set to true, the provided hook must accept
kwargs args. Under the hook, this PR adds a
`_forward_pre_hooks_with_kwargs` and a `_forward_hook_with_kwargs`
set to keep track of which hooks accept kwargs.
Differential Revision: [D41431111](https://our.internmc.facebook.com/intern/diff/D41431111)
Pull Request resolved: https://github.com/pytorch/pytorch/pull/89389
Approved by: https://github.com/soulitzer
The way Python-3.10+ deals with annotations has changed, see
https://docs.python.org/3/howto/annotations.html#accessing-the-annotations-dict-of-an-object-in-python-3-10-and-newer
In particular, derived classes would not inherit parent annotations as
following example clearly demonstrates:
```
class Base:
a: int = 3
b: str = 'abc'
class Derived(Base):
pass
print(Derived.__annotations__)
```
if executed by 3.10 runtime, it will return an empty dict, while older
ones return annotations of the base class
Fix that by implementing this difference in `get_annotations` function
I wonder if in general, we should do it even for older versions of Python runtime
Pull Request resolved: https://github.com/pytorch/pytorch/pull/81334
Approved by: https://github.com/suo
Module's overridden `__setattr__` has special handling for parameters, submodules, and buffers, resulting in reduces performance for its default attributes (which are not of those types). Setting them directly results in a significant improvement for module construction speed (~10x).
Pull Request resolved: https://github.com/pytorch/pytorch/pull/77098
Approved by: https://github.com/jbschlosser
Adds support for scripting ParameterDicts and getattr() on them. It does
not support iterating on ParameterDicts because torch/nn/container.py
implementation of ParameterDict.items() uses a generator, which is not
supported by torchscript. torch/nn/container.py would need to be updated
so that iter gets correctly registered in python_sugared_value.cpp
Added a test in test_module_containers.py
Pull Request resolved: https://github.com/pytorch/pytorch/pull/77143
Approved by: https://github.com/eellison
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/71443
cogwheel test inline_cvr_infer_canary_pyper_model_publish is timing out.
The convert_fx call takes > 20 mins for local and local_ro sub modules, which used to take ~ 2 mins.
Test Plan:
Fblearn flow run
* the following cmd took 1113 seconds before the diff and 5002 seconds after.
flow-cli clone-locally 320014219 --run-as-secure-group pytorch_at_scale --operators pyper_model_publish_workflow.pyper_model_publish_workflow.process_torch_package_model_files.process_non_sparse_parameters[0]
Cogwheel test
* Cogwheel test with packages in B3588 (the last good run) took 4694.48s
* Cogwheel test with packages in B3590 (the first timeout) took 13975.83s
* Cogwheel test with the following packages took 4535.04s
* all packages in B3588 except the model publish
* the model publish built with D33469839 (043e84b3d2) reversed (created D33633570)
Reviewed By: albanD, jerryzh168
Differential Revision: D33633570
fbshipit-source-id: dc5e777c48a90c551641a3f79126461f6a60449e
(cherry picked from commit 03ab65023a)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/67254
Fixes https://github.com/pytorch/pytorch/issues/65997
BC breaking:
`output = torch.ops._test.leaky_relu(self=torch.tensor(-1.0))` now fails with the error `TypeError: __call__() got multiple values for argument 'self'` since we call into `OpOverloadBundle`'s `__call__` method that has `self` bound to it as its first argument.
Follow up work:
1. disallow `default` as an overload name for aten operators.
2. Add a method to obtain a list of all overloads (exclude the ones registered by JIT)
3. Add methods/properties to `OpOverload` to access more schema information (types of input and output args etc)
cc ezyang gchanan
Test Plan: Imported from OSS
Reviewed By: pbelevich
Differential Revision: D33469839
Pulled By: anjali411
fbshipit-source-id: c3fc43460f1c7c9651c64b4d46337be21c400621
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/67254
Fixes https://github.com/pytorch/pytorch/issues/65997
TODO: disallow `default` as an overload name for aten operators.
BC breaking:
`output = torch.ops._test.leaky_relu(self=torch.tensor(-1.0))` now fails with the error `TypeError: __call__() got multiple values for argument 'self'` since we call into `OpOverloadBundle`'s `__call__` method that has `self` bound to it as its first argument.
cc ezyang gchanan
Test Plan: Imported from OSS
Reviewed By: albanD
Differential Revision: D33262228
Pulled By: anjali411
fbshipit-source-id: 600dbf511514ea9b41aea3e6b1bc1102dab08909
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/66253
This was initially broken in #65829 and unbroken in #66003, this PR cleans
it up by removing the mypy ignore line.
Test Plan:
```
mypy torch/jit/_recursive.py --no-incremental
```
Imported from OSS
Reviewed By: supriyar
Differential Revision: D31475100
fbshipit-source-id: 46ab2ede72c08b926f4f9a6b03b1a1375b884c8a
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/59956
Issue #50175. Basically two things need to be checked and are lacking currently:
1. Overload declarations should always have a single `pass` statement as the body.
2. There should be always an implementation provided for decls which doesn't
have the torch.jit._overload decorator. So in this case we need to check
whether we are actually compiling a function body with decorator ahead.
Test Plan:
python test/test_jit.py TestScript.test_function_overloads
Imported from OSS
Reviewed By: gmagogsfm
Differential Revision: D29106555
fbshipit-source-id: 2d9d7df2fb51ab6db0e1b726f9644e4cfbf733d6
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/61076
Previously we would always retrieve ignored methods from the
type, which doesn't work when the user has overriden the ignored method
for a specific instance.
This PR changes things up so we retrieve the ignored method as a bound
method from the object being scripted, unwrap it, then re-bind it to the
scriptmodule.
Test Plan: Imported from OSS
Differential Revision: D29504421
Pulled By: suo
fbshipit-source-id: 14649863ea69a8d2180dd2c4341ec9a826039de1
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/44324
**Summary**
This commit adds reference semantics to TorchScript class types;
modifications made to them within TorchScript will be visible in Python.
**Test Plan**
This commit adds a unit test to `TestClassType` that checks that
modifications made to a class type instance passed into TorchScript are
visible in Python after executing the scripted function or module.
**Fixes**
This commit closes#41421.
Test Plan: Imported from OSS
Reviewed By: gmagogsfm
Differential Revision: D24912807
Pulled By: SplitInfinity
fbshipit-source-id: d64ac6211012425b040b987e3358253016e84ca0
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/60003
**Summary**
`infer_concrete_type_builder` in `_recursive.py` assumes `__constants__`
is a `set` if it exists as an attribute on the module being scripted.
Instead, it should create a set out of whatever `__constants__` is.
**Test Plan**
Ran code from the issue.
**Fixes**
This commit fixes#59947.
Test Plan: Imported from OSS
Reviewed By: pbelevich
Differential Revision: D29174243
Pulled By: SplitInfinity
fbshipit-source-id: aeb8bded80038da35478714b6a697a766ac447f5
Summary:
Fixes https://github.com/pytorch/pytorch/issues/54211
This was a little more annoying than expected, because the `exclude = ` key in `mypy.ini` is weird. I'll file an upstream issue about that.
I ignored one file, `torch/distributed/elastic/agent/server/api.py` that had ~8 errors that were hard to figure out. This can be done in a follow-up.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/55712
Reviewed By: walterddr
Differential Revision: D27694976
Pulled By: malfet
fbshipit-source-id: 228d8be6af040343ce46595dabaca212e69ccc68
Summary:
When type inference fails when JITing torchscript module, the error message does not give any implication where the error fails. For example: "Cannot create dict for key type 'int?', only int, float, complex, Tensor and string keys are supported".
This adds the variable name and item to the error message.
Reviewed By: ajaech
Differential Revision: D26327483
fbshipit-source-id: d8c85e7550258d7c56530f5826ff9683ca8b2b94
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/50870
**Summary**
Module attributes whose types cannot be determined based on annotations
or inference based on their values at script time are added to the
concrete type of the corresponding module as "failed attributes". Any
attempt to access them in scripted code produces an error with a message
explaining that the attribute could not be contributed to a
corresponding attribute on the TorchScript module. However, this error
is not more specific than that.
This commit modifies `infer_type` in `_recursive.py` so that it returns
`c10::InferredType` instead, which allows more information about typing
failures to be communicated to the caller through the `reason()` method
on this class. This information is appended to the hint added to the
module concrete type for failed attributes.
**Testing**
This commit adds a unit test to `test_module_containers.py` that checks
that extra information is provided about the reason for the failure
when a module attribute consisting of a list of `torch.nn.Module` fails to convert.
Test Plan: Imported from OSS
Reviewed By: pbelevich
Differential Revision: D26091472
Pulled By: SplitInfinity
fbshipit-source-id: fcad6588b937520f250587f3d9e005662eb9af0d
Summary:
Retake on https://github.com/pytorch/pytorch/issues/40493 after all the feedback from albanD
This PR implements the generic Lazy mechanism and a sample `LazyLinear` layer with the `UninitializedParameter`.
The main differences with the previous PR are two;
Now `torch.nn.Module` remains untouched.
We don't require an explicit initialization or a dummy forward pass before starting the training or inference of the actual module. Making this much simpler to use from the user side.
As we discussed offline, there was the suggestion of not using a mixin, but changing the `__class__` attribute of `LazyLinear` to become `Linear` once it's completely initialized. While this can be useful, by the time being we need `LazyLinear` to be a `torch.nn.Module` subclass since there are many checks that rely on the modules being instances of `torch.nn.Module`.
This can cause problems when we create complex modules such as
```
class MyNetwork(torch.nn.Module):
def __init__(self):
super(MyNetwork, self).__init__()
self.conv = torch.nn.Conv2d(20, 4, 2)
self.linear = torch.nn.LazyLinear(10)
def forward(self, x):
y = self.conv(x).clamp(min=0)
return self.linear(y)
```
Here, when the __setattr__ function is called at the time LazyLinear is registered, it won't be added to the child modules of `MyNetwork`, so we have to manually do it later, but currently there is no way to do such thing as we can't access the parent module from LazyLinear once it becomes the Linear module. (We can add a workaround to this if needed).
TODO:
Add convolutions once the design is OK
Fix docstrings
Pull Request resolved: https://github.com/pytorch/pytorch/pull/44538
Reviewed By: ngimel
Differential Revision: D24162854
Pulled By: albanD
fbshipit-source-id: 6d58dfe5d43bfb05b6ee506e266db3cf4b885f0c
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/45262
**Summary**
This commit adds an API for ignoring arbitrary module attributes during
scripting. A class attribute named `ignored_attributes` containing names
of attributes to ignore can be added to the class of the instance being
scripted. Attributes ignored in this fashion cannot be used in
`forward`, methods used by `forward` or by `exported` methods. They
are, however, copied to the `RecursiveScriptModule` wrapper and can be
used by `ignored` methods and regular Python code.
**Test Plan**
This commit adds unit tests to `TestScriptPy3` to test this new API.
Test Plan: Imported from OSS
Reviewed By: eellison
Differential Revision: D23971882
Pulled By: SplitInfinity
fbshipit-source-id: 8c81fb415fde7b78aa2f87e5d83a477e876a7cc3
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/43872
This PR allows the recursive scripting to have a separate
submodule_stubs_fn to create its submodule with specific user provided
rules.
Fixes https://github.com/pytorch/pytorch/issues/43729
Test Plan: Imported from OSS
Reviewed By: suo
Differential Revision: D23430176
Pulled By: wanchaol
fbshipit-source-id: 20530d7891ac3345b36f1ed813dc9c650b28d27a
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/42390
**Summary**
This commit extends support for properties to include
ScriptModules.
**Test Plan**
This commit adds a unit test that has a ScriptModule with
a user-defined property.
`python test/test_jit_py3.py TestScriptPy3.test_module_properties`
Test Plan: Imported from OSS
Reviewed By: eellison, mannatsingh
Differential Revision: D22880298
Pulled By: SplitInfinity
fbshipit-source-id: 74f6cb80f716084339e2151ca25092b6341a1560
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/44226
**Summary**
At present, the `share_types` argument to `create_script_module` is used
to decide whether to reuse a previously created type for a top-level
module that has not yet been compiled. However, that setting does not apply
to the compilation of submodules of the top-level module; types are
still reused if possible.
This commit modifies `create_script_module` so that the `share_types`
flag is honoured during submodule compilation as well.
**Test Plan**
This commit adds a unit test to `TestTypeSharing` that checks that
submodule types are not shared or reused when `share_types` is set to
`False`.
**Fixes**
This commit fixes#43605.
Test Plan: Imported from OSS
Reviewed By: eellison
Differential Revision: D23602371
Pulled By: SplitInfinity
fbshipit-source-id: b909b8b6abbe3b4cb9be8319ac263ade90e83bd3
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/43093
without this it's hard to tell which module is going wrong
Test Plan:
```
> TypeError:
> 'numpy.int64' object in attribute 'Linear.in_features' is not a valid constant.
> Valid constants are:
> 1. a nn.ModuleList
> 2. a value of type {bool, float, int, str, NoneType, torch.device, torch.layout, torch.dtype}
> 3. a list or tuple of (2)
```
Reviewed By: eellison
Differential Revision: D23148516
fbshipit-source-id: b86296cdeb7b47c9fd69b5cfa479914c58ef02e6
