Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/69550
Fix the wiki URL.
Also minor reorganization in onnx.rst.
Test Plan: Imported from OSS
Reviewed By: msaroufim
Differential Revision: D32994269
Pulled By: malfet
fbshipit-source-id: 112acfe8b7c778d7e3c2cef684023fdaf2c6ec9c
(cherry picked from commit f0787fabde)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/69547
ScriptModule export introduces duplicated ONNX initializers for shared weights, unnecessarily increases ONNX model size. This PR de-duplicates ONNX initializers for model exported in eval mode, by checking if the underlying tensors share the same `data_ptr`, `strides` and `sizes`.
Test Plan: Imported from OSS
Reviewed By: msaroufim
Differential Revision: D32994271
Pulled By: malfet
fbshipit-source-id: 10ac66638b6255890875272472aa9ed07a5b1d9a
Co-authored-by: BowenBao <bowbao@microsoft.com>
(cherry picked from commit d7cbde940c)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/68491
* Allows implementing symbolic functions for domains other than `aten`, for example `prim`, in symbolic_opset#.py.
* Allows symbolic function to access extra context if needed, through `SymbolicFunctionState`.
* Particularly, the `prim::PythonOp` special case can access node without the need of passing node through inputs. Updates will be made downstreams, and in a follow-up PR we will remove the previous workaround in exporter.
* `prim::Loop`, `prim::If`, etc are now moved outside of `_run_symbolic_function` from utils.py, and to symbolic_opset9.py.
Motivation for this change:
- Better maintainability and reducing complexity. Easier to add symbolic for operators, both simple and complex ones (that need additional context), without the former needing to know the existence of the latter.
- The design idea was long outdated. prim ops are no longer rare special cases, and they shouldn't all be handled inside `_run_symbolic_function`. As a result this function becomes too clumsy. There were also prim ops symbolic added in symbolic_opset#.py with signature `prim_[opname]`, creating separation and confusion.
Test Plan: Imported from OSS
Reviewed By: jansel
Differential Revision: D32483782
Pulled By: malfet
fbshipit-source-id: f9affc31b1570af30ffa6668da9375da111fd54a
Co-authored-by: BowenBao <bowbao@microsoft.com>
(cherry picked from commit 1e04ffd2fd)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/68490
The use of ATEN as a fallback operator during ONNX conversion is important for increasing operator coverage or even provide more efficient implementations over some ONNX ops.
Currently this feature is available through `OperatorExportTypes.ONNX_ATEN_FALLBACK`,
but it also performs changes to the graph that are runnable by Caffe2, only.
This PR introduces restricts caffe2-specific graph transformations for `ONNX_ATEN_FALLBACK`
operator export type for when pytorch is built with caffe2 support (aka BUILD_CAFFE2=1 during build)
The first version of this PR introduced a new operator export type `ONNX_ATEN__STRICT_FALLBACK`,
which essentially is the same as `ONNX_ATEN_FALLBACK` but without caffe2 transformations.
It was preferred to not introduce a new operator export type, but to refine the existing aten fallback one
## BC-breaking note
### The global constant `torch.onnx.PYTORCH_ONNX_CAFFE2_BUNDLE` is removed in favor of
a less visible `torch.onnx._CAFFE2_ATEN_FALLBACK`.
`PYTORCH_ONNX_CAFFE2_BUNDLE` is really a dead code flag always set to False.
One alternative would be fixing it, but #66658 disables Caffe2 build by default.
Making a Caffe2 feature a private one seems to make more sense for future deprecation.
### The method `torch.onnx.export` now defaults to ONNX when `operator_export_type` is not specified.
Previously `torch.onnx.export's operator_export_type` intended to default to `ONNX_ATEN_FALLBACK` when `PYTORCH_ONNX_CAFFE2_BUNDLE` was set, but it would never happen as `PYTORCH_ONNX_CAFFE2_BUNDLE` is always undefined
Co-authored-by: Nikita Shulga <nshulga@fb.com>
Test Plan: Imported from OSS
Reviewed By: jansel
Differential Revision: D32483781
Pulled By: malfet
fbshipit-source-id: e9b447db9466b369e77d747188685495aec3f124
(cherry picked from commit 5fb1eb1b19)
Cover more cases of scope inferencing where consecutive nodes don't have valid scope information. Usually these nodes are created in some pass where authors forgot to assign meaningful scope to them.
* One rule of `InferScope` is to check if the current node's outputs' users share the same scope. Recursively run `InferScope` on the user nodes if they are missing scope as well. Since the graph is SSA, the depth is finite.
* Fix one pass that missed scope information for a new node.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/71897
Export should fail if export_modules_as_functions is set and opset_version<15.
This is because opeset_version < 15 implies IR version < 8, which means no local function support.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/71619
Summary:
For some ONNX exported models, the inputs/outputs names have sometimes a numeric value and this makes pretty hard to inspect the generated graphs in the case of large models.
The solution in this PR was initially submitted to our internal utilities library by take-cheeze https://github.com/pfnet/pytorch-pfn-extras/pull/102
Now we would like to upstream this change by adding an extra kwarg when exporting the model to allow replacing these numeric names with actual debuggable ones.
As an example, the following code shows that the module output is `3`
```python
g, p, o = _model_to_graph(module, torch.ones(1, 10))
for n in g.nodes():
for v in n.outputs():
print(v.debugName())
```
output
```
3
```
With this PR
```
v3_Gemm
```
This allows identifying this out as a value from the associated Gemm layer.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/68976
Reviewed By: jansel
Differential Revision: D33662246
Pulled By: msaroufim
fbshipit-source-id: 45f56eef2a84d9a318db20c6a6de6c2743b9cd99
(cherry picked from commit 513c1d28f1)
Fix the wiki URL.
Also minor reorganization in onnx.rst.
[ONNX] restore documentation of public functions (#69623)
The build-docs check requires all public functions to be documented.
These should really not be public, but we'll fix that later.'
Pull Request resolved: https://github.com/pytorch/pytorch/pull/71609
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/69549
[ONNX] minor clarifications of docstrings
1. Make description of ONNX_ATEN_FALLBACK more accurate (after #67460).
2. Specify minimum and maximum values for opset_version. This is pretty
important information and we should make users dig through source
code to find it.
Test Plan: Imported from OSS
Reviewed By: malfet
Differential Revision: D32994267
Pulled By: msaroufim
fbshipit-source-id: ba641404107baa23506d337eca742fc1fe9f0772
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/69546
The arg is not used and was previously deprecated.
Also remove torch.onnx._export_to_pretty_string. It's redundant with the
public version.
Test Plan: Imported from OSS
Reviewed By: malfet
Differential Revision: D32994270
Pulled By: msaroufim
fbshipit-source-id: f8f3933b371a0d868d9247510bcd73c31a9d6fcc
Summary:
- PyTorch and ONNX has supported BFloat16, add this to unblock some mixed-precision training model.
- Support PyTorch TNLG model to use BFloat16 tensors for the inputs/outputs of the layers that run on the NPU.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/66788
Reviewed By: jansel
Differential Revision: D32283510
Pulled By: malfet
fbshipit-source-id: 150d69b1465b2b917dd6554505eca58042c1262a
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/65993
This PR attempts to port `index_add` to structured kernels, but does more than that:
* Adds an `out=` variant to `index_add`
* Revises `native_functions.yaml` registrations, to not have multiple entries and instead pass default value to `alpha`.
* Changes in `derivatives.yaml` file for autograd functioning
* Revises error messages, please see: https://github.com/pytorch/pytorch/pull/65993#issuecomment-945441615
Follow-up PRs in near future will attempt to refactor the OpInfo test, and will give another look at tests in `test/test_torch.py` for this function. (hence the use of ghstack for this)
~This is WIP because there are tests failing for `Dimname` variant on mobile/android builds, and I'm working on fixing them.~
Issue tracker: https://github.com/pytorch/pytorch/issues/55070
Test Plan: Imported from OSS
Reviewed By: ejguan
Differential Revision: D32646426
fbshipit-source-id: b035ecf843a9a27d4d1e18b202b035adc2a49ab5
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/67808
torch.reciprocal implicitly casts the inputs to float, and ONNX
Reciprocal requires floating point inputs.
Also separate the reciprocal test from other tests, and test different
input types.
Test Plan: Imported from OSS
Reviewed By: msaroufim
Differential Revision: D32181307
Pulled By: malfet
fbshipit-source-id: 3e1109b3c85a49c51dc713656a900b4ee78c8340
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/67806
Previously new_full would fail with errors like:
`TypeError: only integer tensors of a single element can be converted to an index`
And full_like would trigger warnings like:
`DeprecationWarning: an integer is required (got type float). Implicit conversion to integers using __int__ is deprecated, and may be removed in a future version of Python.`
Test Plan: Imported from OSS
Reviewed By: msaroufim
Differential Revision: D32181301
Pulled By: malfet
fbshipit-source-id: 2cf262cfef36c18e7b2423efe1e1d4fa3438f0ba
Co-authored-by: Bowen Bao <bowbao@microsoft.com>
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/67805
Also fix Reduce ops on binary_cross_entropy_with_logits
The graph says the output is a scalar but with `keepdims=1`
(the default), the output should be a tensor of rank 1. We set keep
`keepdims=0` to make it clear that we want a scalar output.
This previously went unnoticed because ONNX Runtime does not strictly
enforce shape inference mismatches if the model is not using the latest
opset version.
Test Plan: Imported from OSS
Reviewed By: msaroufim
Differential Revision: D32181304
Pulled By: malfet
fbshipit-source-id: 1462d8a313daae782013097ebf6341a4d1632e2c
Co-authored-by: Bowen Bao <bowbao@microsoft.com>
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/67803
* Addresses comments from #63589
[ONNX] remove torch::onnx::PRODUCER_VERSION (#67107)
Use constants from version.h instead.
This simplifies things since we no longer have to update
PRODUCER_VERSION for each release.
Also add TORCH_VERSION to version.h so that a string is available for
this purpose.
[ONNX] Set `ir_version` based on opset_version. (#67128)
This increases the odds that the exported ONNX model will be usable.
Before this change, we were setting the IR version to a value which may
be higher than what the model consumer supports.
Also some minor clean-up in the test code:
* Fix string replacement.
* Use a temporary file so as to not leave files around in the test
current working directory.
Test Plan: Imported from OSS
Reviewed By: msaroufim
Differential Revision: D32181306
Pulled By: malfet
fbshipit-source-id: 02f136d34ef8f664ade0bc1985a584f0e8c2b663
Co-authored-by: BowenBao <bowbao@microsoft.com>
Co-authored-by: Gary Miguel <garymiguel@microsoft.com>
Co-authored-by: Nikita Shulga <nshulga@fb.com>
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/67271
* [ONNX] Use Reciprocal operator instead of Div(1, x).
This is a more readable and perhaps more performant way to export
torch.reciprocal.
* Use Reciprocal in caffe to operator to import onnx
Test Plan: Imported from OSS
Reviewed By: msaroufim
Differential Revision: D31962519
Pulled By: malfet
fbshipit-source-id: d926e75b1c8312b9a980c9a1207a1a93ba0c71e0
Co-authored-by: take-cheeze <takechi101010@gmail.com>
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/67270
* Add dim argument to all symbolic
* All symbolic depends on any symbolic
Test Plan: Imported from OSS
Reviewed By: msaroufim
Differential Revision: D31962518
Pulled By: malfet
fbshipit-source-id: f7ee05cf4eff5880fc508154267e060952b5b42d
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/66143
Delete test_list_remove. There's no point in testing conversion of
this model since TorchScript doesn't support it.
Add a link to an issue tracking test_embedding_bag_dynamic_input.
[ONNX] fix docs (#65379)
Mainly fix the sphinx build by inserting empty before
bulleted lists.
Also some minor improvements:
Remove superfluous descriptions of deprecated and ignored args.
The user doesn't need to know anything other than that they are
deprecated and ignored.
Fix custom_opsets description.
Make indentation of Raises section consistent with Args section.
[ONNX] publicize func for discovering unconvertible ops (#65285)
* [ONNX] Provide public function to discover all unconvertible ATen ops
This can be more productive than finding and fixing a single issue at a
time.
* [ONNX] Reorganize test_utility_funs
Move common functionality into a base class that doesn't define any
tests.
Add a new test for opset-independent tests. This lets us avoid running
the tests repeatedly for each opset.
Use simple inheritance rather than the `type()` built-in. It's more
readable.
* [ONNX] Use TestCase assertions rather than `assert`
This provides better error messages.
* [ONNX] Use double quotes consistently.
[ONNX] Fix code block formatting in doc (#65421)
Test Plan: Imported from OSS
Reviewed By: jansel
Differential Revision: D31424093
fbshipit-source-id: 4ced841cc546db8548dede60b54b07df9bb4e36e
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/66140
* Add new argument to export api to enable users specifying `nn.Module` classes that they wish to be exported as local function in ONNX model.
* Refactor `torch/csrc/jit/serialization/export.cpp`, and remove redundant `EncoderBase` class.
* ~~Contains changes from #63268~~
* Depends on #63716 to update onnx submodule.
Test Plan: Imported from OSS
Reviewed By: jansel
Differential Revision: D31424098
fbshipit-source-id: c949d0b01c206c30b4182c2dd1a5b90e32b7a0d3
Co-authored-by: BowenBao <bowbao@microsoft.com>
Summary:
Apex O2 hook state_dict to return fp16 weights as fp32. Exporter cannot identify them as same tensors.
Since this hook is only used by optimizer, it is safe to remove this hook while exporting.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/66700
Reviewed By: zou3519
Differential Revision: D31695132
Pulled By: malfet
fbshipit-source-id: 977bdf57240002498f3ad0f1a8046c352e9860e6
Summary:
This moves it to where the user would expect it to be based on the
documentation and all the other public classes in the torch.onnx module.
Also rename it from ONNXCheckerError, since the qualified name
torch.onnx.ONNXCheckerError is otherwise redundant.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/66644
Reviewed By: malfet
Differential Revision: D31662559
Pulled By: msaroufim
fbshipit-source-id: bc8a57b99c2980490ede3974279d1124228a7406
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/64578
* Fix remainder export for edge case when input is negative. New export relies on true_divide export.
* Simplified true_divide export. Cleaned up redundant code which is handled by scalar type analysis pass. Removed dependency on `onnx::Where`, thus supports opset 7 & 8.
Fixes#60179
Test Plan: Imported from OSS
Reviewed By: jansel
Differential Revision: D30919601
Pulled By: malfet
fbshipit-source-id: 0f78621c0ac3bdb6bf4225e049ba5f470dc8ab12
Co-authored-by: BowenBao <bowbao@microsoft.com>
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/64382
* This `use_external_data_format` parameter is used for large models cannot be exported because of the 2GB protobuf limit.
* When `use_external_data_format` set to True, the model is exported in ONNX external data format, in which case some of the model parameters are stored in external binary files and not in the ONNX model file itself.
* This PR will set this paramter to DEPRECATED and check the model proto sizes by code instead of by user, if the sizes lager than 2GB, then `use_external_data_format = True` automatically.
Test Plan: Imported from OSS
Reviewed By: ezyang
Differential Revision: D30905265
Pulled By: malfet
fbshipit-source-id: 82b4e17bfa6a8de2bfd700a5282c12f6835603cb
Co-authored-by: hwangdeyu <dejack953@outlook.com>
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/64380
* `example_outputs` used to determine the type and shape of the outputs without tracing the execution of the model. And it must be provided when exporting a ScriptModule or ScriptFunction when using export() function.
* Since we can work out `example_outputs` in internal function instead of being provided by user, so we deprecated this argument in the export() function to increase user experience of calling this function.
Test Plan: Imported from OSS
Reviewed By: ezyang
Differential Revision: D30905266
Pulled By: malfet
fbshipit-source-id: d00b00d7d02b365d165028288ad915678caa51f2
Co-authored-by: hwangdeyu <dejack953@outlook.com>
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/64373
* Fix some bad formatting and clarify things in onnx.rst.
* In `export_to_pretty_string`:
* Add documentation for previously undocumented args.
* Document that `f` arg is ignored and mark it deprecated.
* Update tests to stop setting `f`.
* Warn if `_retain_param_name` is set.
* Use double quotes for string literals in test_operators.py.
Test Plan: Imported from OSS
Reviewed By: ezyang
Differential Revision: D30905271
Pulled By: malfet
fbshipit-source-id: 3627eeabf40b9516c4a83cfab424ce537b36e4b3
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/64372
custom_opsets arg from torch.onnx.export() is no needed to be removed.
Add some supplementary description and tests for easier understanding.
Test Plan: Imported from OSS
Reviewed By: ezyang
Differential Revision: D30905269
Pulled By: malfet
fbshipit-source-id: 489fbee0e2c1d6c5405c9bf7dfd85223ed981a44
Co-authored-by: hwangdeyu <dejack953@outlook.com>
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/64371
As of now, the "strip_doc_string" parameter was described as below:
strip_doc_string (bool, default True): do not include the field
doc_string``` from the exported model. Otherwise the field will mention the source code locations for model``.
This is usually useless to users who want to transform a PyTorch model to ONNX one. Only when the user wants to debug the export process, these source code locations could provide benefits.
To make the export() function more friendly by providing less parameters, we combined "strip_doc_string" into "verbose" parameter. If a user set verbose to True, it means the users need some log information for debugging the export process and this is similar with the purpose of strip_doc_string parameter.
But the usage of these 2 arguments are opposite: setting verbose to True means we want to print log information to help debug, which means strip_doc_string should be False. And this is how we replace strip_doc_string with verbose argument in this PR.
This PR will still keep it in torch.onnx.export() function for backward support while the usage of it has been combined with verbose argument.
Test Plan: Imported from OSS
Reviewed By: ezyang
Differential Revision: D30905268
Pulled By: malfet
fbshipit-source-id: 2f06eb805c01fe15ff7a1b4f6595c937ba716d60
Co-authored-by: fatcat-z <zhang-ji@outlook.com>
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/64370
As of now, the "_retain_param_name" parameter has no description in PyTorch docs website. According to code, this argument determines if we keep the original parameter names of PyTorch model in the final ONNX graph. If this is False, those original parameter names will be replaced with a series of integers starting from 1.
Since setting numbers as parameter names make no sense to users, we remove this argument from the torch.onnx.export() function to increase user experience of calling this function.
This PR will still keep it in torch.onnx.export() function for backward support while all backend logic has been changed to work as _retain_param_name is set to True.
Test Plan: Imported from OSS
Reviewed By: ezyang
Differential Revision: D30905270
Pulled By: malfet
fbshipit-source-id: ca60757ca17daaff937e9f08da42596086795f4a
Co-authored-by: fatcat-z <zhang-ji@outlook.com>
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/64369
As of now, the "enable_onnx_checker" parameter was described as below:
enable_onnx_checker (bool, default True): If True the ONNX model checker will be run to ensure the exported model is a valid ONNX model.
An invalid ONNX graph is useless to users so such checker should be done for each call.
In this PR, we will still write the model to an ONNX file even it is invalid. And the exception will be thrown after the ONNX file has been created. This enables user output an invalid ONNX graph for debug.
This PR will still keep it in torch.onnx.export() function for backward support while all backend logic has been changed to work as enable_onnx_checker is set to True.
Test Plan: Imported from OSS
Reviewed By: ezyang
Differential Revision: D30905267
Pulled By: malfet
fbshipit-source-id: 3ad3f68e77fcec012cc7ef674cc9a61755eebc9e
Co-authored-by: fatcat-z <zhang-ji@outlook.com>
Summary:
Replace `torch.tensor([numpy.arange(a, b, c)])` with `torch.arange(a, b, c).unsqueeze(0)`
Replace `tuple(numpy.add(a, b))` with `tuple( x + y for (x, y) in zip(a, b)`
As `numpy` is an optional dependency, it shouldn't be used in PyTorch core by default
Pull Request resolved: https://github.com/pytorch/pytorch/pull/65188
Reviewed By: mruberry
Differential Revision: D31009490
Pulled By: malfet
fbshipit-source-id: 528e48f055bf9ac1de1fd7e94c0be41915df9a0b
Summary:
This is step 3/7 of https://github.com/pytorch/pytorch/issues/50276. It only adds support for the argument but doesn't implement new indexing modes yet.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/62722
Test Plan:
Verified this is not FC breaking by adding logging to both meshgrid
overloads and then called meshgrid twice:
`meshgrid(*tensors)`
and
`meshgrid(*tensors, indexing='ij')`
This confirmed that the former signature triggered the original native
function and the latter signature triggered the new native function.
Reviewed By: H-Huang
Differential Revision: D30394313
Pulled By: dagitses
fbshipit-source-id: e265cb114d8caae414ee2305dc463b34fdb57fa6
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/62764Fixes#58733
- Support dynamic interleave for cases with dynamic repeat values
- Moved repeat_interleave symbolic from opset 11 to opset 13, as sequence as output types for loop outputs is needed for this change
Test Plan: Imported from OSS
Reviewed By: SplitInfinity
Differential Revision: D30375179
Pulled By: msaroufim
fbshipit-source-id: 787f96bf91d124fd0483761088c5f4ae930d96a9
Co-authored-by: Shubham Bhokare <shubhambhokare@gmail.com>
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/62763
This PR is to fix the issue that the graph inputs might be updated when we export the model in inference mode.
When a model is export in inference mode, some optimizations will be made. One side effect of these optimizations is: the inputs of graph might be adjusted. Such optimizatiosn include:
1. Conv and BatchNorm op fusion.
2. Do constant folding.
If the user sets export_params=False, or set keep_initializers_as_inputs=True, it's highly possible that the user wants to provide the corresponding parameters or initiliazers as the inputs of the graph.
In such situation, no matter the model is export in inference mode or training mode, exporter needs to prevent above optimizations from adjusting the graph inputs. By this, the inputs of graph could match inputs that users provided.
The changes in this PR, add an additional common judgement to see if the above optimizations needs to be done or not. From the value of export_params and keep_initializers_as_inputs arguments, infer if the graph inputs are allowed to be adjusted.
If no, these optimizations will be ignored, even other requirements are matched.
Besides these code changes, the comments of some parameters below have been updated so that users have more thoughts when they consider how to leverage these parameters for different purposes:
1. export_params
2. training
3. do_constant_folding
4. keep_initializers_as_inputs
Test Plan: Imported from OSS
Reviewed By: SplitInfinity
Differential Revision: D30375183
Pulled By: msaroufim
fbshipit-source-id: 4db8b9695649eb32a3a0fefa950ee2e5651bdba0
Co-authored-by: fatcat-z <jiz@microsoft.com>
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/61560
1. Add a new symbolic function broadcast_tensors() to support exporting torch.broadcast_tensors() function. This is required by exporting torch.distribution.normal() function.
2. Add a new symbolic function normal() to support exporting torch.distribution.normal() function.
3. Add relative tests for normal and uniform ops as well.
Test Plan: Imported from OSS
Reviewed By: nikithamalgifb
Differential Revision: D29767995
Pulled By: SplitInfinity
fbshipit-source-id: acfe5e7801d00c0df8ca46966bbd6015fed0045e
Co-authored-by: Jay Zhang <jiz@microsoft.com>
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/61558
When we construct an empty list by python list comprehension, we need to avoid converting the node without inputs to onnx::Concat in shape_type_inference.cpp and peephole.cpp because it will create an invalid Concat node which doesn't have inputs.
In addition, update the code to avoid passing a Sequence input to an onnx::Cast node which doesn't accept Sequence data type as an input.
Add tests for the validation as well.
Test Plan: Imported from OSS
Reviewed By: nikithamalgifb
Differential Revision: D29767989
Pulled By: SplitInfinity
fbshipit-source-id: f97f172ff20eebda4c3744c7a934df36716f12a2
Co-authored-by: fatcat-z <jiz@microsoft.com>
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/60249
* Add introductory paragraph explaining what ONNX is and what the
torch.onnx module does.
* In "Tracing vs Scripting" and doc-string for torch.onnx.export(),
clarify that exporting always happens on ScriptModules and that
tracing and scripting are the two ways to produce a ScriptModule.
* Remove examples of using Caffe2 to run exported models.
Caffe2's website says it's deprecated, so it's probably best not to
encourage people to use it by including it in examples.
* Remove a lot of content that's redundant:
* The example of how to mix tracing and scripting, and instead
link to Introduction to TorchScript, which includes very similar
content.
* "Type annotations" section. Link to TorchScript docs which explain
that in more detail.
* "Using dictionaries to handle Named Arguments as model inputs"
section. It's redundant with the description of the `args` argument
to `export()`, which appears on the same page once the HTML
is generated.
* Remove the list of supported Tensor indexing patterns. If it's not
in the list of unsupported patterns, users can assume it's
supported, so having both is redundant.
* Remove the list of supported operators and models.
I think the list of supported operators is not very useful.
A list of supported model architectures may be useful, but in
reality it's already very out of date. We should add it back if
/ when we have a system for keeping it up to date.
* "Operator Export Type" section. It's redundant with the description
of the `operator_export_type` arg to to `export()`, which appears on
the same page once the HTML is generated.
* "Use external data format" section. It's redundant with the
description of the `use_external_data_format` arg to `export()`.
* "Training" section. It's redundant with the
description of the `training` arg to `export()`.
* Move the content about different operator implementations producing
different results from the "Limitations" section into the doc for the
`operator_export_type` arg.
* Document "quantized" -> "caffe2" behavior of
OperatorExportTypes.ONNX_ATEN_FALLBACK.
* Combing the text about using torch.Tensor.item() and the text about
using NumPy types into a section titled
"Avoid NumPy and built-in Python types", since they're both
fundamentally about the same issue.
* Rename "Write PyTorch model in Torch way" to "Avoiding Pitfalls".
* Lots of minor fixes: spelling, grammar, brevity, fixing links, adding
links.
* Clarify limitation on input and output types. Phrasing it in terms of
PyTorch types is much more accessible than in terms of TorchScript
types. Also clarify what actually happens when dict and str are used
as inputs and outputs.
* In Supported operators, use torch function and class names and link
to them. This is more user friendly than using the internal aten
op names.
* Remove references to VariableType.h, which doesn't appear to contain
the information that it once did. Instead refer to the generated
.pyi files.
* Remove the text in the FAQ about appending to lists within loops.
I think this limitation is no longer present
(perhaps since https://github.com/pytorch/pytorch/pull/51577).
* Minor fixes to some code I read along the way.
* Explain the current rationale for the weird ::prim_PythonOp op name.
Test Plan: Imported from OSS
Reviewed By: zou3519, ZolotukhinM
Differential Revision: D29494912
Pulled By: SplitInfinity
fbshipit-source-id: 7756c010b2320de0692369289604403d28877719
Co-authored-by: Gary Miguel <garymiguel@microsoft.com>
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/60246
* Adds support for linspace op
* Modifies arange symbolic in opset 9 to replicate the same behavior in which dtype is determined (similar to opset 11) as in https://pytorch.org/docs/stable/generated/torch.arange.html
* Enabled some arange unit tests which were disabled for opset 9
Test Plan: Imported from OSS
Reviewed By: zou3519, ZolotukhinM
Differential Revision: D29494911
Pulled By: SplitInfinity
fbshipit-source-id: bddff18a90f8a78121c8ecdd1dafc15c69962d66
Co-authored-by: Shubham Bhokare <shubhambhokare@gmail.com>
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/60245
Fix after b9bdb07a0261ab5a0b1038f290fa03af6ce0415f. Improving previous fix on two aspects
* Not only checks 0 on first dimension for empty tensor.
* Do not assume empty tensor when shape is not accessible.
Test Plan: Imported from OSS
Reviewed By: zou3519, ZolotukhinM
Differential Revision: D29494917
Pulled By: SplitInfinity
fbshipit-source-id: 02587c3c3be0510312c1a1959f28cab12d81812d
Co-authored-by: BowenBao <bowbao@microsoft.com>
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/59538
Four mealv2 models can export in torch 1.8.1, but fails when torch master introduces relu6 a few months back.
Test Plan: Imported from OSS
Reviewed By: nikithamalgifb, ansley
Differential Revision: D29046607
Pulled By: SplitInfinity
fbshipit-source-id: d9cf7050e4ac0dad892441305ffebc19ba84e2be
Co-authored-by: David <jiafa@microsoft.com>
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/59537
PyTorch sum over empty tensor gives 0, while ONNX produces an error.
torch.sum will be translated into onnx::ReduceSum op. Per the definition of ReduceSum, update the keepdims attribute for this scenario.
Test Plan: Imported from OSS
Reviewed By: nikithamalgifb, ansley
Differential Revision: D29046604
Pulled By: SplitInfinity
fbshipit-source-id: 6f5f3a66cb8eda8b5114b8474dda6fcdbae73469
Co-authored-by: fatcat-z <jiz@microsoft.com>
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/58697
1. Add a symbolic function for aten::roll() op in symbolic_opset9.py.
2. Add a test with multiple scenarios as well.
Test Plan: Imported from OSS
Reviewed By: driazati, bhosmer
Differential Revision: D28714807
Pulled By: SplitInfinity
fbshipit-source-id: eae85f2dcf02737c9256a180f6905a935ca3f57e
Co-authored-by: fatcat-z <jiz@microsoft.com>
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/58695
As PEP8 says: "Pick a rule and stick to it." [1]
[1] https://www.python.org/dev/peps/pep-0008/#string-quotes
Test Plan: Imported from OSS
Reviewed By: driazati
Differential Revision: D28714811
Pulled By: SplitInfinity
fbshipit-source-id: c95103aceb1725c17c034dc6fc8216627f189548
Co-authored-by: Gary Miguel <garymiguel@microsoft.com>
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/58692
This is a fix for exporting fairseq models, see:
```python
model = torch.hub.load(github, 'conv.wmt14.en-fr', tokenizer='moses', bpe='subword_nmt')
model = torch.hub.load(github, 'conv.wmt17.en-de', tokenizer='moses', bpe='subword_nmt')
```
With this fix, and comment out model script one line `GradMultiply`, these two models can be exported successfully with perf met.
The original PR https://github.com/pytorch/pytorch/pull/57708 has merging issue, use this one instead.
Test Plan: Imported from OSS
Reviewed By: driazati
Differential Revision: D28714809
Pulled By: SplitInfinity
fbshipit-source-id: 71c2de6cec7ee05af68560996acf47d97af46fb2
Co-authored-by: David <jiafa@microsoft.com>
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/58691
Note the first commit in this PR has its own pull request here since it seemed self-contained:
https://github.com/pytorch/pytorch/pull/57082
* [ONNX] simplify batch_first logic in RNN tests
* [ONNX] support GRU with packed input in scripting mode
This required two changes:
* Add as_tensor to symbolic_opset9.py
* Change torch::jit::pushPackingPastRnn to recognize and properly
replace another use of the batch_sizes output of prim::PackPadded.
Previously the code assumed that the first use was as input to the
RNN operator. However in some cases, it is also used to compute
max_batch_size. For example in this code:
https://github.com/pytorch/pytorch/blob/febff45/torch/nn/modules/rnn.py#L815-L815
With these changes the GRU tests now pass in scripting mode for opset
version >= 11.
Test Plan: Imported from OSS
Reviewed By: driazati
Differential Revision: D28714805
Pulled By: SplitInfinity
fbshipit-source-id: f19647a04533d9ec76399a8793b3f712ea0337d2
Co-authored-by: Gary Miguel <garymiguel@microsoft.com>
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/58417
Same as title.
Test Plan:
Rely on CI signal.
Update unit test to exercise new code path as well.
Reviewed By: ngimel
Differential Revision: D28482927
fbshipit-source-id: 3ec8682810ed5c8547b1e8d3869924480ce63dcd
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/57830
This is PR is aiming to support tensor.index_add_() method in symbolic function. We leverage scatter_add() to implement this function while ONNX doesn't have a corresponding operator.
Notes:
1. 4 tests have been added for some scenarios.
2. If there are duplicated value in 'index' parameter, the export will still execute successfully but the results are wrong. Add a warning for every call to this symbolic function. And if we detect that the rank of 'index' is greater than the size of the 'dim' dimension, will raise an exception to stop exporting an incorrect ONNX file.
Test Plan: Imported from OSS
Reviewed By: malfet
Differential Revision: D28393518
Pulled By: SplitInfinity
fbshipit-source-id: f487ca2c63fec47c6ab74f1a7783dae7f3b8d1ef
Co-authored-by: Jay Zhang <jiz@microsoft.com>
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/57599
Currently, if we call tensor.to() method and pass a device as the parameter. It will fail, because in symbolic function of to() we didn't handle such case.
So add a check in the beginning of this symbolic function, if this is a device cast, we return self directly. A test has also been added.
Test Plan: Imported from OSS
Reviewed By: malfet
Differential Revision: D28393523
Pulled By: SplitInfinity
fbshipit-source-id: c41e3c0293932fc90dedb544daadd9c5d4b48792
Co-authored-by: Jay Zhang <jiz@microsoft.com>
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/57598
Add a doc string to explain what it does and how to use it.
Remove hack around a bug in Python 2's functools.wrap().
Python 2 is no longer supported.
Test Plan: Imported from OSS
Reviewed By: malfet
Differential Revision: D28393519
Pulled By: SplitInfinity
fbshipit-source-id: aae8c5e7b49e2ad2d24a0e86f8ba47f1cd080e46
Co-authored-by: Gary Miguel <garymiguel@microsoft.com>
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/57596
Add the corresponding symbolic function and test for fill_() function.
Test Plan: Imported from OSS
Reviewed By: malfet
Differential Revision: D28393520
Pulled By: SplitInfinity
fbshipit-source-id: 3e177f88d3776d0d4a9d5e7ec7df4e6629738799
Co-authored-by: Jay Zhang <jiz@microsoft.com>
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/49352
In this PR, we replace all definitions of slice to take None parameters for the start, end, and step. This will simplify the compiler logic
Test Plan:
test_jit test cases
Imported from OSS
Reviewed By: jamesr66a, nikithamalgifb
Differential Revision: D25929903
fbshipit-source-id: 5bfc6bad514a8aafbef2dacc706f95f867fe85f1
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/57079
Testing onnx 1.9 release, we see that the old bug is triggered for the caffe2 test:
`pytest test/onnx/test_pytorch_onnx_caffe2_quantized.py::TestQuantizedOps::test_small_model`
This is because the graph inputs
```python
graph(%x.1 : Tensor,
%conv1._packed_params : __torch__.torch.classes.quantized.Conv2dPackedParamsBase,
%conv2._packed_params : __torch__.torch.classes.quantized.Conv2dPackedParamsBase,
%fc.bias : Float(10, strides=[1], requires_grad=0, device=cpu),
%fc.weight : Float(10, 72, strides=[72, 1], requires_grad=0, device=cpu)):
```
contains `Conv2dPackedParamsBase` which is a PackedParams.
When we do flatten, we will flatten to several tensors, then the shape inference for input misaligned.
This PR record how may tensors got flattened in PackeParams, and skip by these number rather than 1, then the UT passed.
Note that tuple case should still follow the original logic.
Test Plan: Imported from OSS
Reviewed By: SplitInfinity
Differential Revision: D28393949
Pulled By: malfet
fbshipit-source-id: 98d48aad27e5ca03fb10d260f8e625478d996ee2
Co-authored-by: David <jiafa@microsoft.com>
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/50903
First part of #50010. Also fixes#51127.
Test Plan: Imported from OSS
Reviewed By: ngimel
Differential Revision: D27911345
Pulled By: mruberry
fbshipit-source-id: 7138fddc935802918ab9ff19f4bc1b9f4d745d41
Summary:
Adjust how MutationRemover is used to avoid creating aliasDb multiple times for the same graph.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/56675
Reviewed By: pbelevich
Differential Revision: D27945692
Pulled By: SplitInfinity
fbshipit-source-id: a6c548438e88ddee18ef03a6f0461ab9eaaaa829
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/56173
* Create `InplaceConverter` and `ValueTracker` to keep track of aliases of values throughout the graph. For a given value, a new alias is created every time when there is an inplace operation, SetAttr, or through nested blocks owned by If/Loop nodes.
* Fix bug where controlflow node output types are not set, when the complete node is unable to run ONNX shape inference due to containing non-onnx node.
* Add symbolic for `__not__` ~~and `prim_min`~~(update: moved to a separate PR), and update `index_put` opset9 to support case of assignment without providing indices.
* Bump ORT version in CI test.
Test Plan: Imported from OSS
Reviewed By: pbelevich
Differential Revision: D27866138
Pulled By: SplitInfinity
fbshipit-source-id: ab5c9188740c50f783ceba4d54fda43c26e2fde7
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/56172
Enable the standardOps include **Add\Sub\Mul\Div\Gemm\Pow\Mod** with low precision input in ORT
Test Plan: Imported from OSS
Reviewed By: pbelevich
Differential Revision: D27866136
Pulled By: SplitInfinity
fbshipit-source-id: f2cf5649fffefd68c0cc7b6dce94198751636727
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/56166
Support tensordot in symbolic function of opset 12, and add tests accordingly.
Test Plan: Imported from OSS
Reviewed By: pbelevich
Differential Revision: D27866140
Pulled By: SplitInfinity
fbshipit-source-id: 68e218cfbd630900fb92871fc7c0de3e7e8c8c3d
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/56165
Add implementation for cases when
- interleaving happens along dim which consist of dynamic axes
Test Plan: Imported from OSS
Reviewed By: pbelevich
Differential Revision: D27866137
Pulled By: SplitInfinity
fbshipit-source-id: 7fef1b2c614f2e24a677b7ca0886bb37bd0ab479
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/56163
* [ONNX] Improve index_put symbolic to handle singular Bool updates (#53690)
Adds support for cases where the updates to the index_put node is a single Bool value, such as the case shown below
```
mask[indices] = True
```
Fixes#53507
* [ONNX] Support primitive type input/outputs and attributes (#53550)
Support primitive type attributes. Needed for Silero model.
* [ONNX] Fix if output shape mismatch error & Fix graph input directly used as output (#53219)
Fix if output shape mismatch error & Fix graph input directly used as output
* Add support for hann_window operator.
* [ONNX] Replace decomposeLinear pre process pass with a symbolic (#53077)
Replace decomposeLinear pre process pass with a symbolic
* Add a test case for dtype is None.
* Resolve flake8 issue.
* Remove one unused test case.
* Add support for hann_window operator.
* Add a test case for dtype is None.
* Remove one unused test case.
Test Plan: Imported from OSS
Reviewed By: pbelevich
Differential Revision: D27866145
Pulled By: SplitInfinity
fbshipit-source-id: e0b43df9ecd1a95cd7ac297213aba453bbaf2913
Co-authored-by: Shubham Bhokare <32080845+shubhambhokare1@users.noreply.github.com>
Co-authored-by: Negin Raoof <neginmr@utexas.edu>
Co-authored-by: Bowen Bao <bowbao@microsoft.com>
Co-authored-by: Ksenija Stanojevic <KsenijaS@users.noreply.github.com>
Summary:
This PR adds a `padding_idx` parameter to `nn.EmbeddingBag` and `nn.functional.embedding_bag`. As with `nn.Embedding`'s `padding_idx` argument, if an embedding's index is equal to `padding_idx` it is ignored, so it is not included in the reduction.
This PR does not add support for `padding_idx` for quantized or ONNX `EmbeddingBag` for opset10/11 (opset9 is supported). In these cases, an error is thrown if `padding_idx` is provided.
Fixes https://github.com/pytorch/pytorch/issues/3194
Pull Request resolved: https://github.com/pytorch/pytorch/pull/49237
Reviewed By: walterddr, VitalyFedyunin
Differential Revision: D26948258
Pulled By: jbschlosser
fbshipit-source-id: 3ca672f7e768941f3261ab405fc7597c97ce3dfc
Summary:
Fixes https://github.com/pytorch/pytorch/issues/51652.
In particular:
- the main implementation is in `torch.linalg.det` now. `torch.det` is just a deprecated alias to it
- add a new `OpInfo` for `torch.linalg.det`
- remove the old-style tests for `torch.det` (this is similar to what we did for `torch.linalg.slogdet`, see https://github.com/pytorch/pytorch/issues/49194)
- added a `out=` argument to `torch.linalg.det`, but **not** to `torch.det`.
It is worth noting that I had to skip few tests:
- `TestGradientsCuda::test_fn_gradgrad_linalg_det_cuda_float64`. This is not a regression: the functionality is broken also on master, but the test is not executed properly due to https://github.com/pytorch/pytorch/issues/53361.
And the following tests which fails only on ROCm:
- `test_variant_consistency_jit_cuda_{float64,float32}`
- `test_fn_grad_cuda_float64`
I think that the ROCm tests fail because the current linalg.det backward is unstable if the matrix has repeated singular values, see https://github.com/pytorch/pytorch/issues/53364 .
(At the moment of writing some CI jobs are still running but I believe the build will be green, since the only difference wrt the last push is the skip of the ROCm tests)
Pull Request resolved: https://github.com/pytorch/pytorch/pull/53119
Reviewed By: H-Huang
Differential Revision: D27441999
Pulled By: mruberry
fbshipit-source-id: 5eab14c4f0a165e0cf9ec626c3f4bb23359f2a9e
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/54870
copy_operator before going into onnx exporter is being decomposed into aten::expand_as and aten::index_put.
There is a scenario where inputs to copy are not of the same type, but copy op in torch does implicit casting that is not currently reflected inside onnx exporter. This PR is adding casting inside index_put symbolic in case when tensor self is not of the same type as values.
Test Plan: Imported from OSS
Reviewed By: nikithamalgifb
Differential Revision: D27408975
Pulled By: SplitInfinity
fbshipit-source-id: 15022703e76b9c98b02285c06b13d44f3c4a3f00
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/54869
Add symbolic fuction to support torch.outer export to onnx.
Support for transfo-xl-wt103 model.
Test Plan: Imported from OSS
Reviewed By: nikithamalgifb
Differential Revision: D27408978
Pulled By: SplitInfinity
fbshipit-source-id: 70c89a9fc1a5e4a4ddcf674afb1e82e492a7d3b9
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/54866
Replace decomposeLinear pre process pass with a symbolic
Test Plan: Imported from OSS
Reviewed By: nikithamalgifb
Differential Revision: D27408981
Pulled By: SplitInfinity
fbshipit-source-id: d2d76cab3383122a60df1f356742a33db56adc71
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/54863
Adds support for cases where the updates to the index_put node is a single Bool value, such as the case shown below
```
mask[indices] = True
```
Fixes#53507
Test Plan: Imported from OSS
Reviewed By: nikithamalgifb
Differential Revision: D27408977
Pulled By: SplitInfinity
fbshipit-source-id: bcfb55b50ce76b3d4913ffbc16cdef1f98cb7a84
Summary:
* Lowering NLLLoss/CrossEntropyLoss to ATen dispatch
* This allows the MLC device to override these ops
* Reduce code duplication between the Python and C++ APIs.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/53789
Reviewed By: ailzhang
Differential Revision: D27345793
Pulled By: albanD
fbshipit-source-id: 99c0d617ed5e7ee8f27f7a495a25ab4158d9aad6
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/53529
Supported for ONNX export after opset 10.
This is not exportable to opsets < 10 due to
1. onnx::IsInf is introduced in opset 10
2. onnx::Equal does not accept float tensor prior to opset 11
Test Plan: Imported from OSS
Reviewed By: pbelevich, malfet
Differential Revision: D26922418
Pulled By: SplitInfinity
fbshipit-source-id: 69bcba50520fa3d69db4bd4c2b9f88c00146fca7
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/53312
- Add support for aten::repeat_interleave
- NOTE: Also adds fix for cases with split op where input tensor sizes are not known but _outputs is provided
Test Plan: Imported from OSS
Reviewed By: pbelevich, malfet
Differential Revision: D26922422
Pulled By: SplitInfinity
fbshipit-source-id: 5362d0d8ccfdc14c15e1ae73fd70c4c113f823e6
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/53307
This PR did symbolic shape inference, in the onnx pass _jit_pass_onnx_graph_shape_type_inference.
It creates a singleton ConstantValueMap.
It leverages constant folding technique and did a per-op based handling for ConstantValueMap.
As a byproduct, it enables fold_if pass for dynamic axes cases, typically for faster-rcnn etc.
The core change is in `torch/csrc/jit/passes/onnx/shape_type_inference.cpp` and `torch/csrc/jit/passes/onnx/constant_map.cpp`.
We usually need copy tensor to store in the ConstantValueMap, otherwise the underlying value may change. I see this issue in (1) from_blob (2) get value from Constant node.
Test Plan: Imported from OSS
Reviewed By: pbelevich, malfet
Differential Revision: D26922414
Pulled By: SplitInfinity
fbshipit-source-id: 7654dc13d1de8d9496ad4be89f1454260d7bdeb0
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/53306
* [ONNX] Fix for sequence of mutations in blocks (#51577)
Fixes consecutive mutations in a tensor inside blocks.
Also, support append and pop in blocks.
* Support inplace operations + indexing
* Clean up old pass for remove mutations
* Add loop test
* Fixes for set attr in loops
* Removing the new jit API flag
* [ONNX] Redesign onnx pass to enable shape type dependent pattern conversion - cont (#51795)
With the introduction of ONNX shape inference, shape and type are inferred on the fly as operators get converted from ATen to ONNX when running symbolic function. This resolves the shape/type requirement for the symbolic functions. The pre-onnx passes however, can not be supported by shape inference, since at that stage the operators in the graph are still ATen operators.
This PR is to update the design of ONNX pass, to enable a mechanism of capturing subgraphs of ATen operators of certain patterns, and convert them later, when shape/type information of upstream operators are available.
The new design will require pre-onnx passes that need shape/type to be written in two parts, encapsulation and conversion.
The encapsulation part will find the nodes of patterns, like how pre-onnx passes were written previously. But instead of converting the nodes, it will encapsulate them into a sub-block of a new placeholder node. This part is called before onnx pass, so it runs before calling symbolic functions.
The conversion part will be called inside the onnx pass. In onnx pass, run_symbolic_func will be called for each node in topological order. When it reaches the placeholder node, the conversion part will be invoked. It will convert the nodes inside the sub-block based on pattern. By that time, it will have shape/type of upstream operators available. After the conversion is complete, the placeholder node will be removed, and nodes inside its sub-block converted. Run_symbolic_func will be called for these nodes, and they will be converted from ATen operator to ONNX operator.
This PR includes several other fixes, listed below.
* ~~replace helper.cpp with onnx_utils.cpp for holding utility functions.~~
* fix EraseNumberTypes on Bool type, the code was outdated that back then Bool type doesn't exist.
* ~~enable onnx shape inference in export with parameter/initializer data.~~
* other code clean ups.
* fix insertion of identity nodes for loop opset 13 sequence output.
~~PR depends on #51603~~
* Fix after merge
* clang
* Fix clang
* Fix clang
* Fix warning message.
* Fixes for non-model param attributes
* Fix for caffe2
* Additional test
* clang
* Skip test for lower opsets
* fix clang-tidy
* Update init.cpp
* Update remove_inplace_ops_for_onnx.cpp
* Update remove_inplace_ops_for_onnx.cpp
* Update remove_inplace_ops_for_onnx.cpp
* Fix for clang formatting
Test Plan: Imported from OSS
Reviewed By: pbelevich, malfet
Differential Revision: D26922416
Pulled By: SplitInfinity
fbshipit-source-id: e7108620b39b6404c594910786c4d275fee59d84
Co-authored-by: Bowen Bao <bowbao@microsoft.com>
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/53305Fixes#52436
For opset 9 of onnx Pow, if X is int32, Y is float, we will cast back to int32 which is consistent with X type.
However, pytorch is still float. The aten graph sometimes does not bind with the type for operators,
we are fine with the float type and don't want to cast back.
Even if X, Y are int32, the resulting float32 and int32 makes no difference.
Test Plan: Imported from OSS
Reviewed By: pbelevich, malfet
Differential Revision: D26922425
Pulled By: SplitInfinity
fbshipit-source-id: f8c09524acee0de615df10a14310ca1dd583831e
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/53304
With the introduction of ONNX shape inference, shape and type are inferred on the fly as operators get converted from ATen to ONNX when running symbolic function. This resolves the shape/type requirement for the symbolic functions. The pre-onnx passes however, can not be supported by shape inference, since at that stage the operators in the graph are still ATen operators.
This PR is to update the design of ONNX pass, to enable a mechanism of capturing subgraphs of ATen operators of certain patterns, and convert them later, when shape/type information of upstream operators are available.
The new design will require pre-onnx passes that need shape/type to be written in two parts, encapsulation and conversion.
The encapsulation part will find the nodes of patterns, like how pre-onnx passes were written previously. But instead of converting the nodes, it will encapsulate them into a sub-block of a new placeholder node. This part is called before onnx pass, so it runs before calling symbolic functions.
The conversion part will be called inside the onnx pass. In onnx pass, run_symbolic_func will be called for each node in topological order. When it reaches the placeholder node, the conversion part will be invoked. It will convert the nodes inside the sub-block based on pattern. By that time, it will have shape/type of upstream operators available. After the conversion is complete, the placeholder node will be removed, and nodes inside its sub-block converted. Run_symbolic_func will be called for these nodes, and they will be converted from ATen operator to ONNX operator.
This PR includes several other fixes, listed below.
* ~~replace helper.cpp with onnx_utils.cpp for holding utility functions.~~
* fix EraseNumberTypes on Bool type, the code was outdated that back then Bool type doesn't exist.
* ~~enable onnx shape inference in export with parameter/initializer data.~~
* other code clean ups.
* fix insertion of identity nodes for loop opset 13 sequence output.
~~PR depends on #51603~~
Test Plan: Imported from OSS
Reviewed By: SplitInfinity
Differential Revision: D26922417
Pulled By: malfet
fbshipit-source-id: 14ed06158d539e2451c2e5e63ba1b32fb0f75095
Summary:
Context: https://github.com/pytorch/pytorch/pull/53299#discussion_r587882857
These are the only hand-written parts of this diff:
- the addition to `.github/workflows/lint.yml`
- the file endings changed in these four files (to appease FB-internal land-blocking lints):
- `GLOSSARY.md`
- `aten/src/ATen/core/op_registration/README.md`
- `scripts/README.md`
- `torch/csrc/jit/codegen/fuser/README.md`
The rest was generated by running this command (on macOS):
```
git grep -I -l ' $' -- . ':(exclude)**/contrib/**' ':(exclude)third_party' | xargs gsed -i 's/ *$//'
```
I looked over the auto-generated changes and didn't see anything that looked problematic.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/53406
Test Plan:
This run (after adding the lint but before removing existing trailing spaces) failed:
- https://github.com/pytorch/pytorch/runs/2043032377
This run (on the tip of this PR) succeeded:
- https://github.com/pytorch/pytorch/runs/2043296348
Reviewed By: walterddr, seemethere
Differential Revision: D26856620
Pulled By: samestep
fbshipit-source-id: 3f0de7f7c2e4b0f1c089eac9b5085a58dd7e0d97
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/52350
When onnx export creates a 0-dim tensor of constant type, this action overrides the type promotion logic as quoted in #9515. In order to prevent this from happening this PR adds the following functionality.
If the data type is a floating point type, it is converted to a 0-dim double tensor, else it is converted to a 0-dim tensor of its original type
Test Plan: Imported from OSS
Reviewed By: malfet
Differential Revision: D26490325
Pulled By: SplitInfinity
fbshipit-source-id: 4c47c69c9b6523d2e45b74c2541d6d8ca7e28fc9
Summary:
Fixes https://github.com/pytorch/pytorch/issues/39502
This PR adds support for exporting **fake_quantize_per_channel_affine** to a pair of QuantizeLinear and DequantizeLinear. Per tensor support was added by PR https://github.com/pytorch/pytorch/pull/39738.
`axis` attribute of QuantizeLinear and DequantizeLinear, which is required for per channel support, is added in opset13 added by https://github.com/onnx/onnx/pull/2772.
[update 1/20/2021]: opset13 is being supported on master, the added function is now properly tested. Code also rebased to new master.
The function is also tested offline with the following code
```python
import torch
from torch import quantization
from torchvision import models
qat_resnet18 = models.resnet18(pretrained=True).eval().cuda()
qat_resnet18.qconfig = quantization.QConfig(
activation=quantization.default_fake_quant, weight=quantization.default_per_channel_weight_fake_quant)
quantization.prepare_qat(qat_resnet18, inplace=True)
qat_resnet18.apply(quantization.enable_observer)
qat_resnet18.apply(quantization.enable_fake_quant)
dummy_input = torch.randn(16, 3, 224, 224).cuda()
_ = qat_resnet18(dummy_input)
for module in qat_resnet18.modules():
if isinstance(module, quantization.FakeQuantize):
module.calculate_qparams()
qat_resnet18.apply(quantization.disable_observer)
qat_resnet18.cuda()
input_names = [ "actual_input_1" ]
output_names = [ "output1" ]
torch.onnx.export(qat_resnet18, dummy_input, "quant_model.onnx", verbose=True, opset_version=13)
```
It can generate the desired graph.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/42835
Reviewed By: houseroad
Differential Revision: D26293823
Pulled By: SplitInfinity
fbshipit-source-id: 300498a2e24b7731b12fa2fbdea4e73dde80e7ea
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/51524
* def unsafe_chunk() support and test in ops13.
* Use _unsqueeze_helper insteadof Unsqueeze operator
* Cast the splits into long.
* Change the test to a fixed dimension.
* Update test_pytorch_onnx_onnxruntime.py
* Disable test_loop_with_list for opset 13.
Test Plan: Imported from OSS
Reviewed By: pbelevich
Differential Revision: D26203123
Pulled By: SplitInfinity
fbshipit-source-id: b273aeff8339faa0e8e9f1fcfbf877d1b703209f
Co-authored-by: Negin Raoof <neginmr@utexas.edu>
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/51518
* enable remaining test in opset13
* add comments for error version test info
* fix comments:opset12 unbind problem
* add ignore[no-redef]
* fix format
Test Plan: Imported from OSS
Reviewed By: pbelevich
Differential Revision: D26203122
Pulled By: SplitInfinity
fbshipit-source-id: e7d95bd2ce13f79f11965be82f640379cd55ff0f
Co-authored-by: hwangdeyu <deyhuang@qq.com>
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/51517
Fix get/set attributes when getting/setting a model parameter.
This PR also fixes inplace ops in If blocks.
Test Plan: Imported from OSS
Reviewed By: pbelevich
Differential Revision: D26203116
Pulled By: SplitInfinity
fbshipit-source-id: bed6ee6dd92b5b43febc8c584a6872290f8fe33f
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/50954
* Replace optional parameters of Resize with placeholder for ops13.
* Use common methods to handle different versions.
* Correct flake8 issue.
* Update per comments.
* Add something to trigger CI again.
* Trigger another round of CI.
Test Plan: Imported from OSS
Reviewed By: pbelevich
Differential Revision: D26050882
Pulled By: SplitInfinity
fbshipit-source-id: aea6205a1ba4a0621fe1ac9e0c7d94b92b6d8f21
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/50911
Need to replace dtype of export created scalars from float to double. (In torch implicit conversion logic, python numbers are double)
Test case skipped in CI due to that current CI job env does not have CUDA support.
Test Plan: Imported from OSS
Reviewed By: pbelevich
Differential Revision: D26050889
Pulled By: SplitInfinity
fbshipit-source-id: 1fdde23a68d4793e6b9a82840acc213e5c3aa760
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/50910
Fixing pytorch/vision#3251 (PR #49410 triggers the torch vision test build failure, on three tests test_faster_rcnn, test_mask_rcnn, test_keypoint_rcnn. )
The offending PR is fine on pytorch UT, because the torchvision and pytorch test has a gap when we merge them - we are using different test API on two sides, therefore causing some discrepancy.
This PR bridge the gap for the above three tests, and disable _jit_pass_onnx_fold_if pass until it gets fixed.
Allow _jit_pass_onnx_fold_if only when dynamic_axes is None.
Test Plan: Imported from OSS
Reviewed By: pbelevich
Differential Revision: D26050886
Pulled By: SplitInfinity
fbshipit-source-id: b765ffe30914261866dcc761f0d0999fd16169e3
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/50906
In opset 13, squeeze/unsqueeze is updated to take axes as input, instead of attribute.
Test Plan: Imported from OSS
Reviewed By: pbelevich
Differential Revision: D26050883
Pulled By: SplitInfinity
fbshipit-source-id: 7b5faf0e016d476bc75cbf2bfee6918d77e8aecd
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/50905
Adds an additional run of onnx shape inference after constant folding, since initializer may have changed and affected shape inference.
Test Plan: Imported from OSS
Reviewed By: pbelevich
Differential Revision: D26050881
Pulled By: SplitInfinity
fbshipit-source-id: 9e5d69c52b647133cd3a0781988e2ad1d1a9c09d
Summary:
Handle sequence output shape and type inference.
This PR fixes value type of sequence outputs. Prior to this, all model sequence type outputs were unfolded for ONNX models.
This PR also enable shape inference for sequence outputs to represent the dynamic shape of these values.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/46542
Reviewed By: ezyang
Differential Revision: D24924236
Pulled By: bzinodev
fbshipit-source-id: 506e70a38cfe31069191d7f40fc6375239c6aafe
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/48668
Combine tests for `fmod` and `remainder`.
## BC-breaking Note:
In order to make `remainder` operator have type promotion, we have to introduce BC breaking.
### 1.7.1:
In the case where the second argument is a python number, the result is casted to the dtype of the first argument.
```python
>>> torch.remainder(x, 1.2)
tensor([0, 0, 0, 0, 0], dtype=torch.int32)
```
### This PR:
In the case where the second argument is a python number, the dtype of result is determined by type promotion of both inputs.
```python
>>> torch.remainder(x, 1.2)
tensor([1.0000, 0.8000, 0.6000, 0.4000, 0.2000])
```
Test Plan: Imported from OSS
Reviewed By: mruberry
Differential Revision: D25869136
Pulled By: ejguan
fbshipit-source-id: 8e5e87eec605a15060f715952de140f25644008c
Summary:
Based on existing implementation of var_mean, values of dim have to be sequential and start with zero. The formats listed below are cause scenarios with incompatible dimension for the Sub node.
-> dim[1, 2]
-> dim[0, 2]
-> dim[2, 0]
The changes in this PR allow such formats to be supported in var_mean
Pull Request resolved: https://github.com/pytorch/pytorch/pull/48949
Reviewed By: houseroad
Differential Revision: D25540272
Pulled By: SplitInfinity
fbshipit-source-id: 59813a77ff076d138655cc8c17953358f62cf137
Summary:
For the added UT and existing UTs, this code is independent and ready for review.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/48734
Reviewed By: izdeby
Differential Revision: D25502677
Pulled By: bzinodev
fbshipit-source-id: 788b4eaa5e5e8b5df1fb4956fbd25928127bb199
Summary:
Fixes https://github.com/pytorch/pytorch/issues/46213
I didn't yet update the documentation, will add those change soon. A few other things that I didn't do, but want to clarify if I maybe should.
1. I didn't expose projections in c++ API: torch/csrc/api/src/nn/modules/rnn.cpp. Let me know if this is desirable and I will add those changes.
2. I didn't expose projections in "lstm_cell" function and "_thnn_differentiable_lstm_cell_backward" functions from aten/src/ATen/native/RNN.cpp. As far as I understand, they are not needed for nn.LSTM CPU execution. For lstm_cell, projections don't bring any real benefit, since if cell is used separately, it can be easily added in Python. For "_thnn_differentiable_lstm_cell_backward", I'm actually not sure where exactly that function is used, so I also disabled projections there for now. Please let me know if I should change that.
3. I added check that projections are not supported for quantized LSTMs to quantized_lstm_<data/input> functions. But I didn't add any checks to LSTMCell code. It seems that since I disabled projections in "lstm_cell" function, they should also not be available for quantized models through any other API than quantized_lstm_<data/input>. Please let me know if I'm not correct and I will add checks to other places.
4. Projections are not supported for CuDNN versions < 7.1.2. Should I add the check for CuDNN version and disable projections in that case? If so, what will be the best way to do that?
5. Currently I added projection weight as the last weight, so the layout is "w_ih, w_hh, b_ih, b_hh, w_hr". This breaks the assumption that biases come after weights and thus I had to add additional if-s in various places. Alternative way would be to have "w_ih, w_hh, w_hr, b_ih, b_hh" layout, in which case the assumption will be true. But in that case I will need to split the loop in get_parameters function from aten/src/ATen/native/cudnn/RNN.cpp. And in some cases, I will still need to add an "undefined" tensor in the 3rd position, because we get all 5 weights from CuDNN most of the time. So I'm not sure which way is better. Let me know if you think I should change to the weights-then-biases layout.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/47725
Reviewed By: zou3519
Differential Revision: D25449794
Pulled By: ngimel
fbshipit-source-id: fe6ce59e481d1f5fd861a8ff7fa13d1affcedb0c
Summary:
The args parameter of ONNX export is changed to better support optional arguments such that args is represented as:
args (tuple of arguments or torch.Tensor, a dictionary consisting of named arguments (optional)):
a dictionary to specify the input to the corresponding named parameter:
- KEY: str, named parameter
- VALUE: corresponding input
Pull Request resolved: https://github.com/pytorch/pytorch/pull/47367
Reviewed By: H-Huang
Differential Revision: D25432691
Pulled By: bzinodev
fbshipit-source-id: 9d4cba73cbf7bef256351f181f9ac5434b77eee8
Summary:
`isCompleteTensor()` only returns true when both scalar type and shape is present. All dimensions in the shape must be static. This high requirement is unnecessary for many use cases such as when only rank or scalar type needs to be known.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/48162
Reviewed By: malfet
Differential Revision: D25340823
Pulled By: bzinodev
fbshipit-source-id: 1fef61f44918f4339dd6654fb725b18cd58d99cf
Summary:
When converting a model that uses `torch.chunk`, it does not work when we have a dynamic input axes, because `Split` split attr is static for opset 11. Therefore, we convert it using `Slice` (support opset 11+). This PR also handles the cases that the input axes cannot be divided by the number of outputs. Pytorch works a way that fit the first (n-1) outputs for the same dim, and remaining for the last one. Added UT for it.
The existing code on `sequence` `split` cannot be leveraged here, because `start`, `end` of `Slice` are static there, but dynamic here.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/48176
Reviewed By: bdhirsh
Differential Revision: D25274862
Pulled By: bzinodev
fbshipit-source-id: 7d213a7605ad128aca133c057d6dd86c65cc6de9
Summary:
Onnx op Gather index need be int32 or int64. However, we don't have this Cast in our converter.
Therefore, it fails the following UT (for opset 11+)
`seq_length.type().scalarType()` is None, so `_arange_cast_helper()` cannot treat it as all integral, then it will cast all to float. Then this float value will be used as Gather index, hence it throws error in ORT about float type index.
The fix is that we need cast Gather index type to Long if it is not int/long.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/47653
Reviewed By: heitorschueroff
Differential Revision: D25298056
Pulled By: mruberry
fbshipit-source-id: 05e3a70ccfd74612233c63ec5bb78e060b211909
Summary:
Fixes https://github.com/pytorch/pytorch/issues/45215
Still need to resolve a few mypy issues before a review. In special, there is an error which I don't know how to solve, see:
```python
torch/onnx/utils.py:437: error: Name 'is_originally_training' is not defined [name-defined]
if training is None or training == TrainingMode.EVAL or (training == TrainingMode.PRESERVE and not is_originally_training):
```
`is_originally_training` is used but never defined/imported on [`torch/onnx/utils.py`](ab5cc97fb0/torch/onnx/utils.py (L437)),
Pull Request resolved: https://github.com/pytorch/pytorch/pull/45258
Reviewed By: zhangguanheng66
Differential Revision: D25254920
Pulled By: ezyang
fbshipit-source-id: dc9dc036da43dd56b23bd6141e3ab92e1a16e3b8
Summary:
Onnx op Gather index need be int32 or int64. However, we don't have this Cast in our converter.
Therefore, it fails the following UT (for opset 11+)
`seq_length.type().scalarType()` is None, so `_arange_cast_helper()` cannot treat it as all integral, then it will cast all to float. Then this float value will be used as Gather index, hence it throws error in ORT about float type index.
The fix is that we need cast Gather index type to Long if it is not int/long.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/47653
Reviewed By: ejguan
Differential Revision: D25097324
Pulled By: bzinodev
fbshipit-source-id: 42da1412d1b972d4d82c17fb525879c2575820c9
Summary:
Support exporting with `as_tuple = true`
Example:
`torch.nonzero(x, as_tuple=True)`
This is the same as
`torch.unbind(torch.nonzero(x), 1)`
Pull Request resolved: https://github.com/pytorch/pytorch/pull/47421
Reviewed By: malfet
Differential Revision: D24870760
Pulled By: bzinodev
fbshipit-source-id: 06ca1e7ecf95fbf7c28eebce800df958c83264c8
Summary:
* Enable ONNX shape inference by default.
* ONNX could potentially set inferred shape in output instead of value_infos, checking both to be sure.
* Small fix in symbol_map to avoid overlooking dup symbols.
* Fix scalar_type_analysis to be consistent with PyTorch scalar type promotion logic.
* Correctly handle None dim_param from ONNX inferred shape.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/46629
Reviewed By: ailzhang
Differential Revision: D24900171
Pulled By: bzinodev
fbshipit-source-id: 83d37fb9daf83a2c5969d8383e4c8aac986c35fb
Summary:
This PR is a bug fix.
As UT shows, for multiple-dimensional tensors, the current conversion for _len returns the total number of the tensors. But it should return the first dimension length, as pytorch _len defines.
Need `Squeeze` op at the end to ensure it outputs a scalar value.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/47538
Reviewed By: malfet
Differential Revision: D24870717
Pulled By: bzinodev
fbshipit-source-id: c53c745baa6d2fb7cc1de55a19bd2eedb2ad5272
Summary:
Cases with bool inputs to index_put nodes were handled for tracing purposes. This PR adds support for similar situations in scripting
Pull Request resolved: https://github.com/pytorch/pytorch/pull/46866
Reviewed By: malfet
Differential Revision: D24870818
Pulled By: bzinodev
fbshipit-source-id: 2d75ca6f5f4b79d8c5ace337633c5aed3bdc4be7
Summary:
Export as `onnx::MatMul` if possible since it has less constraint. Resolves issue with exporting `weight_norm` in scripting that fails onnx shape inference with `onnx::Gemm` in unreachable `if` subgraph.
Updates skipped tests list.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/46570
Reviewed By: ngimel
Differential Revision: D24657480
Pulled By: bzinodev
fbshipit-source-id: 08d47cc9fc01c4a73a9d78c964fef102d12cc21c
Summary:
Previously did not considered the case were optional `m` is not provided in `torch.eye(n, m)`.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/47016
Reviewed By: ejguan
Differential Revision: D24735916
Pulled By: bzinodev
fbshipit-source-id: ec9b410fc59f27d77d4ae40cb38a67537abb3cd8
Summary:
- rand/randn: the type signature of int[] is different in scripting, thus failing the check.
- where: scripting produces dynamic cases which are supported by `unbind` export of higher opsets.
- test_list_pass: this test fails when using new scripting api, should be fixed by https://github.com/pytorch/pytorch/issues/45369
Pull Request resolved: https://github.com/pytorch/pytorch/pull/45793
Reviewed By: mrshenli
Differential Revision: D24566096
Pulled By: bzinodev
fbshipit-source-id: 6fe0925c66dee342106d71c9cbc3c95cabe639f7
Summary:
The current `_var_mean` implementation cannot ensure non-negative for variance, because it is actually `E(X^2)-(E(X))^2`: numerically when the dimension number is large and X is close to 0, it can have negative numbers (like our UT shows). The new implementation is `(E(X-E(X))^2)`, it ensures non-negative because the expectation of square is non-negative for sure.
The UT passes for the new implementation (but fails for the existing one). So it is good to go.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/47240
Reviewed By: ejguan
Differential Revision: D24735729
Pulled By: bzinodev
fbshipit-source-id: 136f448dd16622b2b46f40cdf6cb2fccf357c48d
Summary:
Enables the use of NoneType arguments to inputs tuple in the export API
Pull Request resolved: https://github.com/pytorch/pytorch/pull/45792
Reviewed By: heitorschueroff
Differential Revision: D24312784
Pulled By: bzinodev
fbshipit-source-id: 1717e856b56062add371af7dc09cdd9c7b5646da
Summary:
Only under static axes does opset 9 supports no-op squeeze when dim is not 1.
Updating the test case where it was setting dynamic axes.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/45369
Reviewed By: anjali411
Differential Revision: D24280180
Pulled By: bzinodev
fbshipit-source-id: d7cda88ab338a1c41a68052831dcebe739a3843c
Summary:
Even when dim is None, there are cases when flatten can be exported.
Also enable test_densenet in scripting mode
Pull Request resolved: https://github.com/pytorch/pytorch/pull/45632
Reviewed By: VitalyFedyunin
Differential Revision: D24116994
Pulled By: bzinodev
fbshipit-source-id: 76da6c073ddf79bba64397fd56b592de850034c4
Summary:
* Support propagating `dim_param` in ONNX by encoding as `ShapeSymbol` in `SymbolicShape` of outputs. If export is called with `dynamic_axes` provided, shape inference will start with these axes set as dynamic.
* Add new test file `test_pytorch_onnx_shape_inference.py`, reusing all test cases from `test_pytorch_onnx_onnxruntime.py`, but focus on validating shape for all nodes in graph. Currently this is not enabled in the CI, since there are still quite some existing issues and corner cases to fix. The test is default to run only at opset 12.
* Bug fixes, such as div, _len, and peephole.cpp passes for PackPadded, and LogSoftmaxCrossEntropy.
* This PR depends on existing PR such as 44332.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/44920
Reviewed By: eellison
Differential Revision: D23958398
Pulled By: bzinodev
fbshipit-source-id: 00479d9bd19c867d526769a15ba97ec16d56e51d
Summary:
Export of embedding bag with dynamic list of offsets.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/44693
Reviewed By: malfet
Differential Revision: D23831980
Pulled By: bzinodev
fbshipit-source-id: 3eaff1a0f20d1bcfb8039e518d78c491be381e1a
Summary:
Export of view op with dynamic input shape is broken when using tensors with a 0-dim.
This fix removes symbolic use of static input size to fix this issue.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/43558
Reviewed By: ailzhang
Differential Revision: D23965090
Pulled By: bzinodev
fbshipit-source-id: 628e9d7ee5d53375f25052340ca6feabf7ba7c53
Summary:
Fix a couple of issues with scripting inplace indexing in prepare_inplace_ops_for_onnx pass.
1- Tracing index copy (such as cases lik x[1:3] = data) already applies broadcasting on rhs if needed. The broadcasting node (aten::expand) is missing in scripting cases.
2- Inplace indexing with ellipsis (aten::copy_) is replaced with aten::index_put and then handled with slice+select in this pass.
Support for negative indices for this op added.
Shape inference is also enabled for scripting tests using new JIT API.
A few more tests are enabled for scripting.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/44351
Reviewed By: ezyang
Differential Revision: D23880267
Pulled By: bzinodev
fbshipit-source-id: 78b33444633eb7ae0fbabc7415e3b16001f5207f
Summary:
in ONNX NegativeLogLikelihoodLoss specification, ignore_index is optional without default value.
therefore, when convert nll op to ONNX, we need to set ignore_index attribute even if it is not specified (e.g. ignore_index=-100).
Pull Request resolved: https://github.com/pytorch/pytorch/pull/44816
Reviewed By: ezyang
Differential Revision: D23880354
Pulled By: bzinodev
fbshipit-source-id: d0bdd58d0a4507ed9ce37133e68533fe6d1bdf2b
Summary:
Moved description of tool and changes in function name
Pull Request resolved: https://github.com/pytorch/pytorch/pull/44124
Reviewed By: albanD
Differential Revision: D23674618
Pulled By: bzinodev
fbshipit-source-id: 5db0bb14fc106fc96358b1e0590f08e975388c6d
Summary:
* Support sequence type (de)serialization, enables onnx shape inference on sequence nodes.
* Fix shape inference with block input/output: e.g. Loop and If nodes.
* Fix bugs in symbolic discovered by coverage of onnx shape inference.
* Improve debuggability: added more jit logs. For simplicity, the default log level, when jit log is enabled, will not dump ir graphs.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/43929
Reviewed By: albanD
Differential Revision: D23674604
Pulled By: bzinodev
fbshipit-source-id: ab6aacb16d0e3b9a4708845bce27c6d65e567ba7
Summary:
Follow up to https://github.com/pytorch/pytorch/pull/36404
Adding prim::device and prim::dtype to list of skipped peepholes when we run inlining. In the long term another fix may not be to encode shape / dtype info on the traced graph, because it is not guaranteed to be correct. This is blocked by ONNX currently.
Partial fix for https://github.com/pytorch/pytorch/issues/43134
Pull Request resolved: https://github.com/pytorch/pytorch/pull/43363
Reviewed By: glaringlee
Differential Revision: D23383987
Pulled By: eellison
fbshipit-source-id: 2e9c5160d39d690046bd9904be979d58af8d3a20
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/43025
- Use new overloads that better reflect the arguments to interpolate.
- More uniform interface for upsample ops allows simplifying the Python code.
- Also reorder overloads in native_functions.yaml to give them priority.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/37177
ghstack-source-id: 106938111
Test Plan:
test_nn has pretty good coverage.
Relying on CI for ONNX, etc.
Didn't test FC because this change is *not* forward compatible.
To ensure backwards compatibility, I ran this code before this change
```python
def test_func(arg):
interp = torch.nn.functional.interpolate
with_size = interp(arg, size=(16,16))
with_scale = interp(arg, scale_factor=[2.1, 2.2], recompute_scale_factor=False)
with_compute = interp(arg, scale_factor=[2.1, 2.2])
return (with_size, with_scale, with_compute)
traced_func = torch.jit.trace(test_func, torch.randn(1,1,1,1))
sample = torch.randn(1, 3, 7, 7)
output = traced_func(sample)
assert not torch.allclose(output[1], output[2])
torch.jit.save(traced_func, "model.pt")
torch.save((sample, output), "data.pt")
```
then this code after this change
```python
model = torch.jit.load("model.pt")
sample, golden = torch.load("data.pt")
result = model(sample)
for r, g in zip(result, golden):
assert torch.allclose(r, g)
```
Reviewed By: AshkanAliabadi
Differential Revision: D21209991
fbshipit-source-id: 5b2ebb7c3ed76947361fe532d1dbdd6faa3544c8
Summary:
Update repeat op so that the inputs to sizes argument can a mixture of dynamic and constant inputs
Pull Request resolved: https://github.com/pytorch/pytorch/pull/43430
Reviewed By: houseroad
Differential Revision: D23494257
Pulled By: bzinodev
fbshipit-source-id: 90c5e90e4f73e98f3a9d5c8772850e72cecdf0d4
Summary:
Duplicate of https://github.com/pytorch/pytorch/issues/41413
This PR initiates the process of updating the torchsciprt backend interface used by ONNX exporter.
Replace jit lower graph pass by freeze module pass
Enable ScriptModule tests for ONNX operator tests (ORT backend) and model tests by default.
Replace jit remove_inplace_ops pass with remove_mutation and consolidation all passes for handling inplace ops.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/43791
Reviewed By: houseroad
Differential Revision: D23421872
Pulled By: bzinodev
fbshipit-source-id: a98710c45ee905748ec58385e2a232de2486331b
Summary:
During scripting, combination of shape (or size()) and slice (e.g x.shape[2:]) produces following error:
slice() missing 1 required positional argument: 'step'
This happens because aten::slice has 2 signatures:
- aten::slice(Tensor self, int dim, int start, int end, int step) -> Tensor
- aten::slice(t[] l, int start, int end, int step) -> t[]
and when a list is passed instead of tensor the 2nd of the two slice signatures is called, and since it has 4 instead of 5 arguments it produces the above exception.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/42935
Reviewed By: houseroad
Differential Revision: D23398435
Pulled By: bzinodev
fbshipit-source-id: 4151a8f878c520cea199b265973fb476b17801fe
