Summary:
* With the update of split output to dynamic list it breaks the export to onnx.
Now split ir becomes two ops: 1. Dynamic[] <= Split(), and 2. out1, out2, out3
<= Prim::ListUnpack. In this fix these two consecutive ops get fused when being
exported to onnx.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/15092
Reviewed By: dzhulgakov
Differential Revision: D13583832
Pulled By: houseroad
fbshipit-source-id: 3eb18c871e750921ad6d5cc179254bee9bcf4c99
Summary:
[ note: stacked on expect files changes, will unstack once they land ]
This adds DeviceObjType (cannot use DeviceType it is already an enum)
to the type hierarchy and an isDevice/toDevice pair to IValue.
Previous hacks which used an int[] to represent Device are removed
and at::Device is used instead.
Note: the behavior or .to is only a subset of python, we need to
fix the aten op so that it accepts Option[Device] and Optional[ScalarType].
Pull Request resolved: https://github.com/pytorch/pytorch/pull/14666
Reviewed By: suo
Differential Revision: D13290405
Pulled By: zdevito
fbshipit-source-id: 68b4381b292f5418a6a46aaa077f1c902750b134
Summary:
Stacked on https://github.com/pytorch/pytorch/pull/14378, only look at the last commit.
This changes the way methods are defined in TorchScript archives to use
PythonPrint rather than ONNX protobufs.
It also updates torch.proto to directly document the tensor data
structure actually being serialized.
Notes:
* because PythonPrint prints all the methods at once per module, this
removes MethodDef in favor of a single torchscript_area and a separate
caffe2_graphs entry. Note that NetDef's already have method names,
so there is no need or a separate method name entry.
* This switches cpp/pickle area to RecordRef (references to a file in
the container format) since it is possible the data in these arenas
may be large and not suited to json ouput.
* Removes 'annotations' -- annotations should be re-added on the first
commit that actually has a practical use for them. In the current state
it is unlikely they are representing the right information.
* Some expect files have changed because PythonPrint is preserving more
debug name information for parameter names.
* MethodEncoder (the ONNX output format) has been deleted. There is still
some cleanup possible combining EncoderBase and GraphEncode now that there
is only a single pathway using EncoderBase.
* This incorporates the changes from #14397
to define TensorDef
Pull Request resolved: https://github.com/pytorch/pytorch/pull/14400
Reviewed By: suo
Differential Revision: D13231800
Pulled By: zdevito
fbshipit-source-id: af5c1152d0bd6bca8b06c4703f59b161bb19f571
Summary:
This PR adds a `try_outplace` option to the tracer. When `try_outplace` is true, the tracer will attempt to out-of-place ops (similar to how things are done today). When it's false, the correct in-place op is emitted.
I made `try_outplace` false by default, but flipped it to true for ONNX export utils. zdevito jamesr66a, anywhere else I should preserve the existing behavior?
Pull Request resolved: https://github.com/pytorch/pytorch/pull/14254
Reviewed By: eellison
Differential Revision: D13166691
Pulled By: suo
fbshipit-source-id: ce39fdf73ac39811c55100e567466d53108e856b
Summary:
There should really be a single place to erase or do special treatment to the prim::ListConstruct during ONNX export, this will make it consistent across different calls. e.g it will give a correct output graph in the following case:
```python
class Test(torch.nn.Module):
def forward(self, input):
return torch.cat([input, torch.zeros(input.size(0), 1).type_as(input)], dim=1)
```
Before this PR, we have the onnx graph as:
```
graph(%0 : Byte(2, 3)) {
%1 : Long() = onnx::Constant[value={0}](), scope: Test
%2 : Dynamic = onnx::Shape(%0), scope: Test
%3 : Long() = onnx::Gather[axis=0](%2, %1), scope: Test
%4 : Long() = onnx::Constant[value={1}](), scope: Test
%5 : Dynamic = onnx::Unsqueeze[axes=[0]](%3)
%6 : Dynamic = onnx::Unsqueeze[axes=[0]](%4)
%7 : int[] = onnx::Concat[axis=0](%5, %6)
%8 : Float(2, 1) = onnx::ConstantFill[dtype=1, input_as_shape=1, value=0](%7), scope: Test
%9 : Byte(2, 1) = onnx::Cast[to=2](%8), scope: Test
%10 : Byte(2, 4) = onnx::Concat[axis=1](%0, %9), scope: Test
return (%10);
}
```
Which is wrong since onnx does not have a concept of `int[]`, here is the onnx graph after this PR:
```
graph(%0 : Byte(2, 3)) {
%1 : Long() = onnx::Constant[value={0}](), scope: Test
%2 : Dynamic = onnx::Shape(%0), scope: Test
%3 : Long() = onnx::Gather[axis=0](%2, %1), scope: Test
%4 : Long() = onnx::Constant[value={1}](), scope: Test
%5 : Dynamic = onnx::Unsqueeze[axes=[0]](%3)
%6 : Dynamic = onnx::Unsqueeze[axes=[0]](%4)
%7 : Dynamic = onnx::Concat[axis=0](%5, %6)
%8 : Float(2, 1) = onnx::ConstantFill[dtype=1, input_as_shape=1, value=0](%7), scope: Test
%9 : Byte(2, 1) = onnx::Cast[to=2](%8), scope: Test
%10 : Byte(2, 4) = onnx::Concat[axis=1](%0, %9), scope: Test
return (%10);
}
```
Pull Request resolved: https://github.com/pytorch/pytorch/pull/13195
Differential Revision: D12812541
Pulled By: wanchaol
fbshipit-source-id: db6be8bf0cdc85c426d5cbe09a28c5e5d860eb3e
Summary:
This PR changes the compiler to correctly emit in-place operators for augmented assignments (`+=` and friends).
- To better match the Python AST structure, add an `AugAssign` tree view and make `Assign` apply only to `=` assignments.
- Emit those `AugAssign` exprs in the compiler, dispatching to in-place aten ops for tensors and lowering to simple assignments for scalar types.
- In order to preserve (suspect) ONNX export semantics, add a pass to lower the in-place operators to out-of-place operators.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/13364
Differential Revision: D12899734
Pulled By: suo
fbshipit-source-id: bec83be0062cb0235eb129aed78d6110a9e2c146
Summary:
This PR adds optional type to ATen native, autograd, JIT schema and Python Arg parser, closes#9513. It allows us to use optional default values (including None) for function signature and implementations like clamp, etc., and also let us remove the python_default_init hack.
Follow up:
remove python_default_init completely.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/12582
Differential Revision: D10417423
Pulled By: wanchaol
fbshipit-source-id: 1c80f0727bb528188b47c595629e2996be269b89
Summary:
There's some action at a distance issues and not having this is disabling quantization in C2 for prod use cases
ref T34831022
Pull Request resolved: https://github.com/pytorch/pytorch/pull/12538
Differential Revision: D10302931
Pulled By: jamesr66a
fbshipit-source-id: 700dc8c5c4297e942171992266ffb67b815be754
Summary:
- In Python 2, use of `/` (regardless of int/float/Tensor) causes a compiler error if
`from __future__ import division` is not imported in the file.
- The / operator is universally set to do "true" division for integers
- Added a `prim::FloorDiv` operator because it is used in loop unrolling.
The error if users use '/' in python 2 without importing from __future__
occurs when building the JIT AST.
cc apaszke zdevito
Pull Request resolved: https://github.com/pytorch/pytorch/pull/11016
Differential Revision: D9613527
Pulled By: zou3519
fbshipit-source-id: 0cebf44d5b8c92e203167733692ad33c4ec9dac6
Summary:
Operators like aten::chunk used to return a number of tensors, but
now return a list. To make it easier to do shape prop through
aten::chunk and fuse it, I've also introduced prim::ConstantChunk,
which behaves like the previous implementation (has a variable length
output list).
The downside of this PR is that the introduction of more lists to the IR causes the LSTM and MiLSTM graphs to be considered as non-differentiable by the graph executor. I verified that they are still optimize correctly, and my next patch (that changes how the specializations/differentiation works) will restore those.
zdevito
Pull Request resolved: https://github.com/pytorch/pytorch/pull/10949
Reviewed By: zdevito
Differential Revision: D9556823
Pulled By: apaszke
fbshipit-source-id: 33e63b17fc7247cac6cfc05eb7eb9bf069b499ee
Summary:
This was done because it surprising for a decorator to run a function
rather than wrap it, and not simplify the syntax for tracing modules.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/11069
Reviewed By: jamesr66a
Differential Revision: D9583192
Pulled By: zdevito
fbshipit-source-id: b914b7ab4c73c255086465a6576eef3a22de1e13
Summary:
Things like torch.zeros now appear in traces rather than constants.
To continue to support our current level of ONNX export, we run
constant prop to turn these back into constants where possible before
export.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/10935
Differential Revision: D9527427
Pulled By: zdevito
fbshipit-source-id: 552a8bcc01b911251dab7d7026faafdd7a3c758a
Summary:
* Fix the necessary pathways so that tuples and lists can be inputs to the script.
* prevent linear algebra functions from being run in shape prop because
they frequently will error out for nonsense data.
* favor schema-driven python input conversion where possible.
remaining cases where we directly create Stacks without schema are
only for debugging
* Make the error messages when calling script/trace functions more pythonic
* Simplify FlattenTuples -- now that tuples are supported we can choose to only flatten tuples when needed. This may have to be revisited pending onnx test results, but is necessary for making tuple io work.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/10812
Differential Revision: D9477982
Pulled By: zdevito
fbshipit-source-id: ed06fc426e6ef6deb404602a26c435a7fc40ea0c
Summary:
Now, run `python test/onnx/test_operators.py --no-onnx`, we won't introduce any onnx python dependence. (No onnx/protobuf python packages needs to be installed)
The major changes:
- output pbtxt from C++ exporter directly, so the floating format may be slightly different. (This should be fine, since it's just to guard ONNX exporting.)
- ONNX python packages are only imported if we run the ONNX related checks. Those checks are disabled when using `--no-onnx` flag.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/10151
Reviewed By: jamesr66a
Differential Revision: D9130706
Pulled By: houseroad
fbshipit-source-id: ea28cf5db8399929179698ee535137f209e9ce6f
Summary: Minor fix for a bug introduced by D9004285
Reviewed By: anderspapitto
Differential Revision: D9028762
fbshipit-source-id: 9b9c5eef30e61d7ae19784e0418fa29bad2b5564
Summary:
Based on top of #9763 (first 3 commits belong to that PR). The first commits from this PR are "Stop using attributes ..."
I tried to separate the changes into fairly meaningful commits. I can't split them up into smaller PRs, because everything starts working and all tests pass only after the whole sequence, but hopefully this will make reviewing somewhat easier.
Known issues/regressions/future tasks:
- `aten::lerp` and `aten::clamp` are no longer fusable
- `CreateAutodiffSubgraphs` needs a rewrite
- It is much more strict now, and will miss a lot of opportunities, especially when viewing ops are involved. Our previous approach was "ignore the assumption on shape availability in gradient formulas to determine differentiability, and hope that shape prop will be robust enough to actually deliver them before we differentiate", which obviously doesn't scale well to more complex cases. We should either work on reducing the size dependency of grad formulas (feasible e.g. for `view`/`reshape`, unfeasible for `squeeze`/`unsqueeze`), or make `CreateAutodiffSubgraphs` integrate some kind of "I could integrate this node into an AD subgraph, but will I be able to infer the shape of its input" reasoning (kind of like a limited shape prop, that doesn't infer anything, and only tells if it *could* infer something).
- It sometimes creates constant-only (or constants + one node) graphs, which is useless
- Broken `aten::add` in auto-batching, because it gained a non-tensor input. I changed the test for pointwise operations to use `aten::mul` instead, but I needed to disable the LSTM cell test. I'm not sure how scalar constants should be implemented in this case, because I don't fully understand our format. cc: ChunliF
- Graph import does some hacks to recover type of constants. This code should be removed once we'll gain the ability to export the IR along with value types.
- There's still a fair amount of dead code that can be removed. I didn't want to make this diff any bigger, and removing it is an easy task.
- Graph fuser could be improved to use signature matching (possibly using `OperatorSet`) instead of basing on node kinds.
- Manual constant propagation for the `ListConstruct` node in `torch/onnx/utils.py` should be replaced with a proper constant propagation pass (or we should ensure that the one we have handles at least this case before we remove this code).
zdevito
Pull Request resolved: https://github.com/pytorch/pytorch/pull/9807
Reviewed By: ezyang
Differential Revision: D9004285
Pulled By: apaszke
fbshipit-source-id: fe88026a765f6b687354add034c86402362508b7
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/9718
This patch switches the interpreter to use IValue's primitive numbers rather than tensors for computing on integers and floats. In addition to preparing the interpreter for first-class support of other types, this cleans up the handling of primitive numbers, making it possible to just use the normal operator overloading dispatch to find the right implementation for numbers. As a result of this change, a lot of other functionality needed to be updated since it was the first time we use non-tensors in a lot of places in the code base.
Notes:
* Fixes code_template.py so that multi-line strings are indented correctly when used on a standalone line
* Cast operators (`int(x)`) now are functional. Some tests have addition conversions to integers because
we no longer allow implicit tensor -> integer conversions following the same convention as in python
* prim::ListConstruct/createList has been added to the interpreter for creating lists and this has
replaced aten::stack for integers lists
* gen_jit_dispatch.py has been refactored so that non-tensor types use operators on IValues to extract
the primitives
* IValue gains a .to<T> method that is the equivalent of tensor_as but for IValue instead of at::Tensor
* `constant_as<T>` is switched over to using IValues's `.to<T>` method, to make conversion from constant->IValue->C++ type
more consistent. This functionality combined with `toIValue(Value*)` replaces the `tensor_as` and `as_tensor` family of functions.
* conditional expressions (if, loop) and operators related to them are now computed on integers rather than tensors
* IValue gains constructors for constructing from at::Scalar and converting to it. However, IValue itself will always store
the scalars as a double or int64.
* To align with python 3 syntax, TK_INT, TK_FLOAT, and TK_BOOL have been removed from the parser, and int/float/bool are just treated as special identifiers in the compiler,
along with print. These are represented as special sugared values with a `call` method implemented. For int/float/bool this implements casting behavior.
* Dropped shared_from_this from Type/Module. They were not needed and they making debugging harder because they internally throw/catch exceptions.
* Shape propagation has been updated to support running nodes that include floating point primitive types, this required some refactoring of internal functions.
* TensorToNum and NumToTensor have actual implementations as operators now
* regster_prim_ops now contains implementations of math operators for float/int primitive types, and for mixed (prim <+> tensor) versions. This removes the need for special handling in compiler.cpp
* Primitive math is now entirely handled by letting the compiler choose the right overloads. This removes tons of special casing in the compiler.
* incorporates eellison's change to allow casting from return values. Due to the addition of primitive support, the code need slight modifications, so I just pre-merged it here.
* stack.h gains generic vararg versions of push/pop that know how to convert to/from C++ types:
```
at::Tensor a;
at::Scalar b;
pop(stack, a, b);
at::Tensor c = a + b;
push(stack, c);
```
apaszke
Pull Request resolved: https://github.com/pytorch/pytorch/pull/9584
Reviewed By: apaszke
Differential Revision: D8910546
Pulled By: zdevito
fbshipit-source-id: 0f3e60d4d22217f196a8f606549430e43b7e7e30
Addresses #8177
A design doc can be found here: [gist](https://gist.github.com/zou3519/4b7f13f03cc9f3612bd9363e6405fa0a) version or [quip](https://fb.quip.com/azL1AqUckBdo) version
General approach:
- Add NumberType, FloatType, IntType to represent Python numbers, floats and ints.
- Emit these types for python literals
- Change aten_schema such that Scalars are NumberType, int64_t and bool are IntType.
- Emit aten::type_as, prim::NumToTensor, and prim::TensorToNum nodes for tensor-number math. (see examples below)
- Erase NumberType, prim::NumToTensor, and prim::TensorToNum for ONNX export
### Tensor/number math
```
import torch
@torch.jit.script
def fn(x):
return x + 1
```
```
graph(%x : Dynamic) {
%1 : int = prim::Constant[value={1}]()
%2 : Dynamic = prim::NumToTensor(%1)
%3 : Dynamic = aten::type_as(%2, %x)
%4 : Dynamic = aten::add[alpha={1}](%x, %4)
return (%5);
}
```
### Number/Number Math
```
import torch
@torch.jit.script
def fn(zero):
c = 1 + 1
return zero + c
```
```
graph(%zero : Dynamic) {
%1 : int = prim::Constant[value={1}]()
%2 : int = prim::Constant[value={1}]()
%3 : Dynamic = prim::num_to_tensor(%1)
%4 : Dynamic = prim::num_to_tensor(%2)
%5 : Dynamic = aten::add[alpha={1}](%3, %4)
%c : int = prim::TensorToNum(%6) # this is the result of the addition
...
return (%13);
}
```
List of squashed commits:
* Introduce Python Number types
Added: IntType, FloatType, NumberType with
IntType <: NumberType
FloatType <: NumberType
Changed aten_schema so arguments have corresponding types
* Emit a NumberType for python literals.
Also emit a NumberType for Scalar default values.
* Add prim::NumToTensor and prim::TensorToNum
* Add DynamicType -> NumberType implicit cast for bc
* Better ensureTensor error message
* Add ensureTensorOrNumber. Allow passing Number to some functions
Like the range() construct and slices
* Patch IntList to work.
IntList is still a DynamicType in the frontend: a tensor gets built from
a List[int].
Also, IntList[1] is a "union between int and IntList" the way it is
implemented. If the frontend sees an int being passed for an IntList[1]
arg, it converts it to a tensor as well.
* Enforce some order on schemas to avoid overload ambiguity
add(Tensor, Tensor) should appear earlier than add(Tensor, Scalar). This
matches the order in which python_arg_parser parses its arguments.
* Disable std_dim and var_dim tests.
With the new schema information, std(input, keepdim) and std(input, dim)
are ambiguous. This will need to be fixed at a later date.
* Add NumberType erasure pass.
This is used for ONNX export and to ensure that NumberType information
doesn't reach the interpreter
* Add support for mixed tensor/number math ops.
* Tests for new functionality.
Includes:
- Tensor/number math
- number/number math
- EraseNumberTypes pass test
* Patch tests
Update expect tests for:
- decompose_addmm
- loop unrolling tests
Because python numbers are now NumberType, they cannot be returned by
functions anymore. Work around this by using "torch.full", or by adding
a tensor([0]) (taken from FIXME_zerol()). Both approaches are used
because torch.full is more readable, but it is broken in some cases.
* Add erase_number_types to torch/CMakeLists.txt
* Move math back to emitSimpleExpr from emitSugaredExpr
* Remove some dead lines
* Renable some excluded script/trace tests that are fixed.
* Move some tests to expected failure
* Address some comments (more addressing to come)
* Erase relevant aten::type_as nodes in EraseNumberTypes
I also changed it so that EraseNumberTypes is only called for ONNX
export. It is no longer used to prevent
prim::NumToTensor/prim::TensorToNum from reaching shape_analysis or
interpreter.cpp.
shape_analysis infers the type of the output of these nodes to be the
same as their input.
intepreter.cpp treats both of these nodes as no-ops.
* Add reminder to fix std/var
* Call EraseNumberTypes only when exporting a script module
* Update expects after rebase
* [ONNX] Allow specifying only a subset of input/output names
Then we can only specify the "real" names while ignoring the names for all the parameters
* fix
* Update utils.py
* Workaround in onnx to get transposes into init_nets
This adds a pass to ONNX so that it can speculate Transpose
operators so that ONNX's split pass can put them into an init_net
Also fixes a potential bug in onnx peephole where an optimization
across blocks might move a Value and violate scoping.
* Perform shape propagation when embedding a program into a trace.
This ensures the trace still has type information specific to that trace, which will help onnx export succeed in more cases.
* Codemod to update our codebase to 0.4 standard
* Update some of the test scri[ts
* remove Variable in test_clip_grad_value
* fix _symbolic_override_wrapper_maker
* Switch JIT passes to take a graph rather than TracingState
* Add pybind11 binding for ONNX pass from graph
* Fix canonicalize pass
* address comment
* Switch ToONNX to explicitly return new graph
* optimize_graph instead of optimize_trace
Allows you to export an ONNX model as:
Protobuf file (this is what we have now)
Uncompressed zip archive
Compressed zip archive
Directory
* Experimental support for different ONNX export types
* Remove a copy
* Add comment
* Add test cases
* lint
* fix bug
* address comments
* Namespaced symbols
- Our interned strings now have structure, "ns::symname" rather than just
"symname" before. We support efficient namespace testing for uniques
by encoding the namespace in one byte in the Symbol internal representation.
See torch/csrc/jit/interned_strings.h for a more in-depth implementation
discussion.
- All uses of ksymbol are now attr::symbol (or some appropriate namespace).
The valid namespaces are prim, attr, onnx and aten.
- Symbol is bound in Python as a qualified string "attr::symbol", EXCEPT for the
attribute setting/getting API, whose symbols must always be attr
symbols; they get special cased to assume strings are passed.
There's a little bit of naughtiness in the implementation, maybe you know
how to solve it.
- However, the g.op() convenience function assumes that you're generating
ONNX operators, unless you explicitly qualify.
- All ATen operators and nodes have built-in interned strings generated
for them, so you should never have to write a string literal ever again.
The tracing code is adjusted to use it.
- ONNX exporter now properly tests to see that all operators are in
onnx namespace before accepting the export. This is way more
robust than the previous exporter, which would be willing to
export capitalized operators which were not actually ONNX operators.
- A slight organizational change for symbolic.py; this module now ONLY
contains aten operators. In particular, the exporter for Constant
has moved into utils.py (along with Undefined, from the C++ side),
since primitive ops get "special treatment."
- The un-inplacing logic in recording is more robust, so that we don't
delete a trailing underscore from __and__. This never affected us
before because we didn't have any tests for it.
Signed-off-by: Edward Z. Yang <ezyang@fb.com>
