Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/20787
Set requires_grad=False for bias: this will block the jit tracing.
The as_type fix: The input tensor shape and output tensor shape will be different, which will trigger the assertion failure at https://fburl.com/0m8xy7tc.
Reviewed By: jamesr66a
Differential Revision: D15445092
fbshipit-source-id: 22da41a56ecb9ac092585d0cc1ff0658fb9d631b
Summary:
Stack from [ghstack](https://github.com/ezyang/ghstack):
* **#19578 [jit] Try to script all Python functions**
This adds the `torch.jit._enable_recursive_script` context manager, which will try to compile any Python functions it sees. It's hidden behind an internal context manager for now since it's incomplete (doesn't work for script_methods/Python submodules). If it can't compile the Python function it outputs an error.
](https://our.intern.facebook.com/intern/diff/15386727/)
Pull Request resolved: https://github.com/pytorch/pytorch/pull/19578
Pulled By: driazati
Differential Revision: D15386727
fbshipit-source-id: 4e308f67677b8e9fccfc525a91bb2f4585062048
Summary:
* Constructs a new type at runtime so that `isinstance` checks work for
weak modules assigned to `ScriptModule`s
* Fix some extraneous names in `__constants__`
* Add `in_features` and `out_features` to `nn.Linear` `__constants__`
Fixes#19363
Pull Request resolved: https://github.com/pytorch/pytorch/pull/20190
Pulled By: driazati
Differential Revision: D15302350
fbshipit-source-id: 1d4d21ed44ab9578a4bc2a72396a82e9bbcd387c
Summary:
TensorList, DoubleList, and BoolList were missing from the pickler, so
this adds them.
As a follow up a lot of the code for these could be templated and cut
down
](https://our.intern.facebook.com/intern/diff/15299106/)
Pull Request resolved: https://github.com/pytorch/pytorch/pull/20191
Pulled By: driazati
Differential Revision: D15299106
fbshipit-source-id: f10c0c9af9d60a6b7fb8d93cea9f550b1a7e2415
Summary:
Previously we only had a Python wrapper for `torch.quantized_lstm_cell`. We had the op `torch.quantized_lstm`, but it didn't have a wrapper. This makes the wrapper
Pull Request resolved: https://github.com/pytorch/pytorch/pull/20249
Reviewed By: driazati
Differential Revision: D15250023
Pulled By: jamesr66a
fbshipit-source-id: f05ad784d903e0ef3a62633c8bf80bad79de48ae
Summary:
This PR adds a new trace API `trace_module` that will allow us to trace multiple methods as a part of a single `ScriptModule`
See the example below.
```python
class Net(nn.Module):
def __init__(self):
super(Net, self).__init__()
self.conv = nn.Conv2d(1, 1, 3)
def forward(self, x):
return self.conv(x)
def weighted_kernel_sum(self, weight):
return weight * self.conv.weight
example_weight = torch.rand(1, 1, 3, 3)
example_forward_input = torch.rand(1, 1, 3, 3)
n = Net()
inputs = {'forward' : example_forward_input, 'weighted_kernel_sum' : example_weight}
module = torch.jit.trace_module(n, inputs)
```
Pull Request resolved: https://github.com/pytorch/pytorch/pull/19905
Differential Revision: D15200007
Pulled By: Krovatkin
fbshipit-source-id: 0354d973fe40cb6e58b395bd866df14e0fc29d5b
Summary:
Stack from [ghstack](https://github.com/ezyang/ghstack):
* **#19638 [jit] Serialize attribute module as torch.jit._pickle**
* use `torch.jit._pickle` as the module for globals in the pickle program. Pickle will try to resolve these to the actual functions in `torch.jit._pickle.py` automatically (I believe this can also be overridden to point to whatever functions you want). This means that `pickle.load("my_model/attributes.pkl")` will work instead of having to use a custom `pickle.Unpickler`
* use `REDUCE` opcodes instead of `BUILD` to make use of the last bullet
* use a union in the unpickler to support globals better (+ any future metadata we might need that can't be stored in an `IValue`), this makes some of the code around `IntList`s clearer and lets us get rid of any lookbehind for opcodes
* pickle things as a tuple instead of a list (an immutable result is more semantically correct)](https://our.intern.facebook.com/intern/diff/15111203/)
Pull Request resolved: https://github.com/pytorch/pytorch/pull/19638
Pulled By: driazati
Differential Revision: D15111203
fbshipit-source-id: 526c6c2b63a48eb1cba1c658045a7809730070dd
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/18092
Previously, tracing required all inputs to be either tensors,
or tuples of tensor. Now, we allow users to pass dicts as well.
Differential Revision: D14491795
fbshipit-source-id: 7a2df218e5d00f898d01fa5b9669f9d674280be3
Summary:
This PR propagates where we use first-class modules objects into the compiler. This creates a transitionary state where:
* compiler.cpp creates Graphs where `self` is a Module class and attributes/parameters/buffers/submodules are looked up with `prim::GetAttr`
* GraphExecutor still runs "lowered graphs" where the self object has been removed by a compiler pass `lower_first_class_method`.
* Tracing still creates "lowered graphs", and a pass "lift_lowered_method" creates a first-class method graph for things.
* This PR separates out Method and Function. A script::Function is a pure Graph with no `self` bound. Similar to Python, a script::Method is just a bound `self` and its underlying `script::Function`.
* This PR also separates CompilationUnit from Module. A CompilationUnit is just a list of named script::Functions. Class's have a CompilationUnit holding the class methods, and Modules also have a CompilationUnit holding their Methods. This avoids the weird circular case Module --has a-> Class -> has a -> Module ...
Details:
* In this transitionary state, we maintain two copies of a Graph, first-class module and lowered. Th first-class one has a self argument that is the module's class type. The lowered one is the lowered graph that uses the initial_ivalues inputs.
* When defining lowered methods using `_defined_lowered` we immediately create the first-class equivalent. The reverse is done lazily, creating lowered_methods on demand from the class.
* The two way conversions will be deleted in a future PR when the executor itself runs first-class objects. However this requires more changes to (1) the traces, (2) the python bindings, and (3) the onnx export pass and would make this PR way to large.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/19167
Differential Revision: D14891966
Pulled By: zdevito
fbshipit-source-id: 0b5f03118aa65448a15c7a7818e64089ec93d7ea
Summary:
Since parameter.data will create a new torch.Tensor each time, we get duplicate tensors when call _unique_state_dict now. Try to deduplicate it before creating new tensor.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/18139
Reviewed By: dzhulgakov
Differential Revision: D14511262
Pulled By: houseroad
fbshipit-source-id: cb69795d0b6509721220650bbb19edeb3459a503
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/18598
ghimport-source-id: c74597e5e7437e94a43c163cee0639b20d0d0c6a
Stack from [ghstack](https://github.com/ezyang/ghstack):
* **#18598 Turn on F401: Unused import warning.**
This was requested by someone at Facebook; this lint is turned
on for Facebook by default. "Sure, why not."
I had to noqa a number of imports in __init__. Hypothetically
we're supposed to use __all__ in this case, but I was too lazy
to fix it. Left for future work.
Be careful! flake8-2 and flake8-3 behave differently with
respect to import resolution for # type: comments. flake8-3 will
report an import unused; flake8-2 will not. For now, I just
noqa'd all these sites.
All the changes were done by hand.
Signed-off-by: Edward Z. Yang <ezyang@fb.com>
Differential Revision: D14687478
fbshipit-source-id: 30d532381e914091aadfa0d2a5a89404819663e3
Summary:
* adds attributes to `ScriptModule.__getattr__` so they can be accessed in Python after re-importing
* full support for all the possible values for an `int64_t`
* this necessitated a bunch more `pushWhatever` functions, so re-introduced a templated version to cut down on duplicate code
* tests to validate references / value sharing works
* adds `torch.jit.Unpickler` which people can use to de-serialize the pickle files into Python / have a quick reference on how to do this without PyTorch
Pull Request resolved: https://github.com/pytorch/pytorch/pull/18188
Differential Revision: D14527490
Pulled By: driazati
fbshipit-source-id: efd15579cc04aa2e28c4b2c9490d82d849dee559
Summary:
This defines a generic counters API that users can utilize to provide monitoring functionality in e.g. a production service. We expose both counters for runtime internals as well as a TorchScript API to create user-defined counters. Synopsis of the API:
- `torch/csrc/jit/script/logging.h` specifies the externally-facing API in C++
- `torch/jit/_logging.py` specifies the Python API
We use an interface, `LoggerBase`, to define the interactions between users and a logging backend. Implementing a subclass of `LoggerBase` allows the user to handle these events in a custom way, such as logging into a DB or calling into an infra-specific counters API.
From the frontend perspective, we can create log events in two ways:
1. We provide an `add_stat_value(name, val)` function. This calls into the Logger backend with a key/value pair. For example, we might call `add_stat_value('foo', 1)` to bump an event counter.
2. We provide a `time_point()` function to record a timestamp in nanoseconds. This can be used in conjunction with `add_stat_value` to record runtime wall clock durations.
Examples of frontend usage can be found in `test_jit.py TestLogging`.
We provide a trivial `LockingLogger` implementation as an example and for testing purposes. It is likely not ready for production usage. It demonstrates that a backend implementing the API can do things like specify aggregation types and report these aggregate stats via the `get_counters()` API.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/18235
Differential Revision: D14545060
Pulled By: jamesr66a
fbshipit-source-id: 04099543a1898cfdd411511e46e03d5dce9b4881
Summary:
Deleting batch tensor since we are no longer maintaining the project and keeping it functional is blocking other improvements.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/18575
Differential Revision: D14671126
Pulled By: eellison
fbshipit-source-id: b42d5b699c4d12171ed95e6d3a977532167f0d2c
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/18148
ghimport-source-id: 40a9d745dc9aeba53d098743323fcbd50ca65137
Stack from [ghstack](https://github.com/ezyang/ghstack):
* **#18148 py interop**
Support for converting classes between the Python–TorchScript boundary. Like other TorchScript values, ScriptClasses are native Python values when used in Python and IValues when used in TorchScript.
Notably, there is a copy across this boundary, which will be surprising to users who will expect standard Python reference semantics. I have some ideas for fixing that, but it's a more involved process.
Reviewed By: jamesr66a
Differential Revision: D14526259
fbshipit-source-id: 5916e3032488a42dc7da756c1826d7c040a21ebd
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/18181
ghimport-source-id: 9c23551584a1a1b0b7ac246367f3a7ae1c50b315
Stack from [ghstack](https://github.com/ezyang/ghstack):
* #18184 Fix B903 lint: save memory for data classes with slots/namedtuple
* **#18181 Fix B902 lint error: invalid first argument.**
* #18178 Fix B006 lint errors: using mutable structure in default argument.
* #18177 Fix lstrip bug revealed by B005 lint
A variety of sins were committed:
- Some code was dead
- Some code was actually a staticmethod
- Some code just named it the wrong way
- Some code was purposely testing the omitted case
Signed-off-by: Edward Z. Yang <ezyang@fb.com>
Differential Revision: D14530876
fbshipit-source-id: 292a371d9a76ddc7bfcfd38b6f0da9165290a58e
Summary:
Allows serialization/loading of attributes (`IValue`s of any type).
* metadata (attribute name, type) is stored in the `model.json`
* The binary format is a subset of the `pickle` module that supports the operations necessary for `IValue`s
* Attributes are serialized in the order they are defined on a module to a list in a single `attributes` file, with submodule attributes coming first. This order directly matches the order attributes are listed in `model.json`
* This can be inspected in Python with `pickle.load()` or with `pickletools` (PyTorch need not be installed for this to work)
* A class is used to store a tensor's index into the tensor table of the model, so to unpickle the file you have to use a custom Unpickler:
```python
class TensorID(object):
def __setstate__(self, id):
self.id = id
class JitUnpickler(pickle.Unpickler):
def find_class(self, module, name):
if module == '__main__' and name == 'TensorID':
return TensorID
JitUnpickler(open("my_model/attributes.pkl", "rb")).load()
```
* pickle format: https://svn.python.org/projects/python/trunk/Lib/pickletools.py
* It currently does not support/guarantee that anything saved out with `pickle` (i.e. if you edit `attributes` with `pickle` directly) instead of our tools will be imported correctly
Also will fix#17683 and fix#16367
Followup Work:
* document format / choice of pickle: #17951
* create an example
* list specializations
* int size specializations, large binputs
* do a first pass over attributes to output only necessary `BINPUT` ops
* attribute reassignment (e.g `self.my_attribute = new_value`)
* `tensor.save("some_checkpoint.pkl")` support with tensors embedded in Pickle file
Pull Request resolved: https://github.com/pytorch/pytorch/pull/17423
Differential Revision: D14470965
Pulled By: driazati
fbshipit-source-id: 6a21a9939efdbe59b4bc57fd31d6d630bab5297e
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/17957
So developer knows what action should be taken when model contains nondeterministic node
Reviewed By: dzhulgakov
Differential Revision: D14435923
fbshipit-source-id: 12d930185852f78c54efc8e90c51aa7c7c7faab5
