Previously, if someone wrote a python abstract impl but didn't import
the module it is in, then we would raise an error message suggesting
that the user needs to add an abstract impl for the operator.
In addition to this, we suggest that the user try importing the module
associated with the operator in the pystub (it's not guaranteed that
an abstract impl does exist) to avoid confusion.
Test Plan:
- new test
Pull Request resolved: https://github.com/pytorch/pytorch/pull/117770
Approved by: https://github.com/ydwu4, https://github.com/williamwen42
Summary:
We've made the following changes:
- The new way to use the API is `m.impl_abstract_pystub(module, context)`.
Every subsequent m.def of an op inside the TORCH_LIBRARY block gives
the op the `impl_abstract_pystub`.
- Added a mechanism to determine if an operator was defined in Python or C++.
Library.define in Python appends the op to a global set, which is analogous
to what we do for tracking Library.impl.
- If someone does `torch.library.impl_abstract` in Python for an operator, then
we require that it has an `impl_abstract_pystub` specified and we also check
that the module in the `impl_abstract_pystub` is the same as the module where
the call to `torch.library.impl_abstract` exists.
- Unfortunately we can't check the "context" (which is the buck target on
buck-based systems) because buck sits above us.
bypass-github-export-checks
Test Plan: - existing tests
Differential Revision: D51080493
Pull Request resolved: https://github.com/pytorch/pytorch/pull/113182
Approved by: https://github.com/ezyang
Summary:
We've made the following changes:
- The new way to use the API is `m.impl_abstract_pystub(module, context)`.
Every subsequent m.def of an op inside the TORCH_LIBRARY block gives
the op the `impl_abstract_pystub`.
- Added a mechanism to determine if an operator was defined in Python or C++.
Library.define in Python appends the op to a global set, which is analogous
to what we do for tracking Library.impl.
- If someone does `torch.library.impl_abstract` in Python for an operator, then
we require that it has an `impl_abstract_pystub` specified and we also check
that the module in the `impl_abstract_pystub` is the same as the module where
the call to `torch.library.impl_abstract` exists.
- Unfortunately we can't check the "context" (which is the buck target on
buck-based systems) because buck sits above us.
Test Plan: - existing tests
Differential Revision: D50972148
Pull Request resolved: https://github.com/pytorch/pytorch/pull/112851
Approved by: https://github.com/ezyang
We want users to be able to define custom ops in C++ but put the
abstract impl in Python (since it is easier to write them in Python and
the abstract impl better models device semantics and data-dependent
operators).
`m.impl_abstract_pystub(opname, python_module, context)` declares the
abstract_impl of the operator to exist in the given python module.
When the abstract_impl needs to be accessed (either via FakeTensor or
Meta), and it does not exist, the PyTorch Dispatcher will yell
with a descriptive error message.
Some details:
- We construct a new global AbstractImplPyStub mapping in
Dispatcher.cpp. Read/write to this map is protected by the Dispatcher
lock.
- We add a new Meta Tensor fallback kernel. The fallback errors out if there is
no meta kernel, but also offers a nicer error message if we see that there is
a pystub.
- We create a `torch._utils_internal.throw_abstract_impl_not_imported_error`
helper function to throw errors. This way, we can throw different error
messages in OSS PyTorch vs internal PyTorch. To invoke this from C++, we
added a PyInterpreter::throw_abstract_impl_not_imported_error.
Differential Revision: [D49464753](https://our.internmc.facebook.com/intern/diff/D49464753/)
Differential Revision: [D49464753](https://our.internmc.facebook.com/intern/diff/D49464753)
Pull Request resolved: https://github.com/pytorch/pytorch/pull/109529
Approved by: https://github.com/ezyang, https://github.com/bdhirsh
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/66744
Modified loops in files under fbsource/fbcode/caffe2/ from the format
`for(TYPE var=x0;var<x_max;x++)`
to the format
`for(const auto var: irange(xmax))`
This was achieved by running r-barnes's loop upgrader script (D28874212) with some modification to exclude all files under /torch/jit and a number of reversions or unused variable suppression warnings added by hand.
Test Plan: Sandcastle
Reviewed By: ngimel
Differential Revision: D31705358
fbshipit-source-id: d6ea350cbaa8f452fc78f238160e5374be637a48
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/66234
Modified loops in files under fbsource/fbcode/caffe2/ from the format
`for(TYPE var=x0;var<x_max;x++)`
to the format
`for(const auto var: irange(xmax))`
This was achieved by running r-barnes's loop upgrader script (D28874212) with some modification to exclude all files under /torch/jit and a number of reversions or unused variable suppression warnings added by hand.
bypass_size_limit
allow-large-files
Test Plan: Sandcastle
Reviewed By: ngimel
Differential Revision: D30652629
fbshipit-source-id: 0ae6c4bbbb554bad42e372792a6430e1acf15e3e
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/48178
I migrated all call sites that used the legacy dispatcher registration API (RegisterOperators()) to use the new API (TORCH_LIBRARY...). I found all call-sites by running `fbgs RegisterOperators()`. This includes several places, including other OSS code (nestedtensor, torchtext, torchvision). A few things to call out:
For simple ops that only had one registered kernel without a dispatch key, I replaced them with:
```
TORCH_LIBRARY_FRAGMENT(ns, m) {
m.def("opName", fn_name);
}
```
For ops that registered to a specific dispatch key / had multiple kernels registered, I registered the common kernel (math/cpu) directly inside a `TORCH_LIBRARY_FRAGMENT` block, and registered any additional kernels from other files (e.g. cuda) in a separate `TORCH_LIBRARY_IMPL` block.
```
// cpu file
TORCH_LIBRARY_FRAGMENT(ns, m) {
m.def("opName(schema_inputs) -> schema_outputs");
m.impl("opName", torch::dispatch(c10::DispatchKey::CPU, TORCH_FN(cpu_kernel)));
}
// cuda file
TORCH_LIBRARY_IMPL(ns, CUDA, m) {
m.impl("opName", torch::dispatch(c10::DispatchKey::CUDA, TORCH_FN(cuda_kernel)));
}
```
Special cases:
I found a few ops that used a (legacy) `CPUTensorId`/`CUDATensorId` dispatch key. Updated those to use CPU/CUDA- this seems safe because the keys are aliased to one another in `DispatchKey.h`
There were a handful of ops that registered a functor (function class) to the legacy API. As far as I could tell we don't allow this case in the new API, mainly because you can accomplish the same thing more cleanly with lambdas. Rather than delete the class I wrote a wrapper function on top of the class, which I passed to the new API.
There were a handful of ops that were registered only to a CUDA dispatch key. I put them inside a TORCH_LIBRARY_FRAGMENT block, and used a `def()` and `impl()` call like in case two above.
Test Plan: Imported from OSS
Reviewed By: ezyang
Differential Revision: D25056090
Pulled By: bdhirsh
fbshipit-source-id: 8f868b45f545e5da2f21924046e786850eba70d9
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/47322
Updating all call-sites of the legacy dispatcher registration API in fbcode to the new API.
I migrated all call sites that used the legacy dispatcher registration API (RegisterOperators()) to use the new API (TORCH_LIBRARY...). I found all call-sites by running `fbgs RegisterOperators()`. This includes several places, including other OSS code (nestedtensor, torchtext, torchvision). A few things to call out:
For simple ops that only had one registered kernel without a dispatch key, I replaced them with:
```
TORCH_LIBRARY_FRAGMENT(ns, m) {
m.def("opName", fn_name);
}
```
For ops that registered to a specific dispatch key / had multiple kernels registered, I registered the common kernel (math/cpu) directly inside a `TORCH_LIBRARY_FRAGMENT` block, and registered any additional kernels from other files (e.g. cuda) in a separate `TORCH_LIBRARY_IMPL` block.
```
// cpu file
TORCH_LIBRARY_FRAGMENT(ns, m) {
m.def("opName(schema_inputs) -> schema_outputs");
m.impl("opName", torch::dispatch(c10::DispatchKey::CPU, TORCH_FN(cpu_kernel)));
}
// cuda file
TORCH_LIBRARY_IMPL(ns, CUDA, m) {
m.impl("opName", torch::dispatch(c10::DispatchKey::CUDA, TORCH_FN(cuda_kernel)));
}
```
Special cases:
I found a few ops that used a (legacy) `CPUTensorId`/`CUDATensorId` dispatch key. Updated those to use CPU/CUDA- this seems safe because the keys are aliased to one another in `DispatchKey.h`
There were a handful of ops that registered a functor (function class) to the legacy API. As far as I could tell we don't allow this case in the new API, mainly because you can accomplish the same thing more cleanly with lambdas. Rather than delete the class I wrote a wrapper function on top of the class, which I passed to the new API.
There were a handful of ops that were registered only to a CUDA dispatch key. I put them inside a TORCH_LIBRARY_FRAGMENT block, and used a `def()` and `impl()` call like in case two above.
Test Plan: Imported from OSS
Reviewed By: ezyang
Differential Revision: D24714803
Pulled By: bdhirsh
fbshipit-source-id: c809aad8a698db3fd0d832f117f833e997b159e1
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/37700
Certain autograd functions can have optional Tensor arguments. For
this purpose it would be nice to support c10::optional<Tensor> as an argument
for C++ autograd functions.
I've added the appropriate overload to ExtractVariables to ensure this works.
For an example, you can look at D21272807 in terms of how this is used.
ghstack-source-id: 103541789
Test Plan: waitforbuildbot
Differential Revision: D21363491
fbshipit-source-id: 0c8665e9bfe279e6b9ab84a889524fea11fa971c
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/28229
We have `torch::RegisterOperators` for custom ops. `torch::jit::RegisterOperators` had a dual state of being able to register custom ops if called one way and being able to register pure JIT ops if called another way.
This is confusing because you end up in different operator libraries depending on which API exactly you're using.
This PR removes the ability for torch::jit::RegisterOperators to register custom ops and forces people to use the new torch::RegisterOperators.
This was already deprecated before but we now remove it.
ghstack-source-id: 92137305
Test Plan: unit tests
Differential Revision: D17981895
fbshipit-source-id: 0af267dfdc3c6a2736740091cf841bac40deff40
Summary:
In TorchScript and C++ extensions we currently advocate a mix of `torch::` and `at::` namespace usage. In the C++ frontend I had instead exported all symbols from `at::` and some from `c10::` into the `torch::` namespace. This is far, far easier for users to understand, and also avoid bugs around creating tensors vs. variables. The same should from now on be true for the TorchScript C++ API (for running and loading models) and all C++ extensions.
Note that since we're just talking about typedefs, this change does not break any existing code.
Once this lands I will update stuff in `pytorch/tutorials` too.
zdevito ezyang gchanan
Pull Request resolved: https://github.com/pytorch/pytorch/pull/13523
Differential Revision: D12942787
Pulled By: goldsborough
fbshipit-source-id: 76058936bd8707b33d9e5bbc2d0705fc3d820763
Summary:
This is likely currently broken due to symbol visibility issues, but we will investigate it using this PR.
CC orionr yf225
Pull Request resolved: https://github.com/pytorch/pytorch/pull/11527
Differential Revision: D10444104
Pulled By: goldsborough
fbshipit-source-id: 4c447beeb9671598ecfc846cb5c507ef143459fe
Summary:
Since we're making parts of the JIT public as part of loading script modules, they should be on the cppdocs website.
Orthogonal: We decided not to export things like `IValue` into the `torch` namespace, so `RegisterOperators` shouldn't be there either.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/11712
Differential Revision: D9837578
Pulled By: goldsborough
fbshipit-source-id: 4c06d2fa9dd4b4216951f27424c2ce795febab9c
Summary:
A couple fixes I deem necessary to the TorchScript C++ API after writing the tutorial:
1. When I was creating the custom op API, I created `torch/op.h` as the one-stop header for creating custom ops. I now notice that there is no good header for the TorchScript C++ story altogether, i.e. when you just want to load a script module in C++ without any custom ops necessarily. The `torch/op.h` header suits that purpose just as well of course, but I think we should rename it to `torch/script.h`, which seems like a great name for this feature.
2. The current API for the CMake we provided was that we defined a bunch of variables like `TORCH_LIBRARY_DIRS` and `TORCH_INCLUDES` and then expected users to add those variables to their targets. We also had a CMake function that did that for you automatically. I now realized a much smarter way of doing this is to create an `IMPORTED` target for the libtorch library in CMake, and then add all this stuff to the link interface of that target. Then all downstream users have to do is `target_link_libraries(my_target torch)` and they get all the proper includes, libraries and compiler flags added to their target. This means we can get rid of the CMake function and all that stuff. orionr AFAIK this is a much, much better way of doing all of this, no?
3. Since we distribute libtorch with `D_GLIBCXX_USE_CXX11_ABI=0`, dependent libraries must set this flag too. I now add this to the interface compile options of this imported target.
4. Fixes to JIT docs.
These could likely be 4 different PRs but given the release I wouldn't mind landing them all asap.
zdevito dzhulgakov soumith
Pull Request resolved: https://github.com/pytorch/pytorch/pull/11682
Differential Revision: D9839431
Pulled By: goldsborough
fbshipit-source-id: fdc47b95f83f22d53e1995aa683e09613b4bfe65
Summary:
This PR adds support for using custom ops in ScriptModules, the last step for our custom op strategy. You can now write
```
import torch
torch.ops.load_library('libcustom_ops.so')
class Model(torch.jit.ScriptModule):
def __init__(self):
super(Model, self).__init__()
torch.jit.script_method
def forward(self, input):
return torch.ops.custom.op(input) + 1
model = Model()
model.forward(torch.ones(5)) # Works
model.save("model.pt") # Works
model = torch.jit.load("model.pt") # Works
```
You can then load the `model.pt` in C++ and execute its `forward` method!
Missing for this was the fact that the script compiler didn't know to convert `ops.custom.op` into a `BuiltinFunction` which then emits a function call. For this I came up with the following strategy inside `torch/csrc/jit/scrip/init.cpp`:
1. When we access `torch.ops`, we return a `CustomOpValue` (subclass of `PythonValue`), whose purpose is only to return a `CustomOpNamespaceValue` (subclass of `PythonValue`) whenever something under it is accessed.
2. `CustomOpNamespaceValue` will then for each field accessed on it return a `BuiltinFunction`.
This doesn't reduce performance for any calls that are not to `torch.ops` (as opposed to inspecting every function call's name the call site, for example).
I also had to fix `BuiltinFunction` to not assume the namespace is always `aten::`.
A lot of other changes are just tidying up the Python and C++ test harness before I integrate it in CI.
zdevito dzhulgakov
Pull Request resolved: https://github.com/pytorch/pytorch/pull/10610
Differential Revision: D9387832
Pulled By: goldsborough
fbshipit-source-id: c00f431db56c7502a66fe1f813fe78067f428ecb
Summary:
This is the last step in the custom operator implementation: providing a way to build from C++ and Python. For this I:
1. Created a `FindTorch.cmake` taken largely from ebetica with a CMake function to easily create simple custom op libraries
2. Created a ` torch/op.h` header for easy inclusion of necessary headers,
3. Created a test directory `pytorch/test/custom_operator` which includes the basic setup for a custom op.
1. It defines an op in `op.{h,cpp}`
2. Registers it with the JIT using `RegisterOperators`
3. Builds it into a shared library via a `CMakeLists.txt`
4. Binds it into Python using a `setup.py`. This step makes use of our C++ extension setup that we already have. No work, yey!
The pure C++ and the Python builds are separate and not coupled in any way.
zdevito soumith dzhulgakov
Pull Request resolved: https://github.com/pytorch/pytorch/pull/10226
Differential Revision: D9296839
Pulled By: goldsborough
fbshipit-source-id: 32f74cafb6e3d86cada8dfca8136d0dfb1f197a0
