# Summary
This PR creates _flash_attention_backward and _scaled_dot_product_flash_attention_backward native functions and registers them to the respective derivatives.yaml.
The goal is to replicate the torch.autograd.Function defined in the FlashAttention repo [here](33e0860c9c/flash_attn/flash_attn_interface.py (L126)) natively in PyTorch. One thing that we don't have access to is ctx.save_for_backward in native PyTorch so in order to save these variables I extended the returned objects from the forward functions.
### MetaFunctions
I also updated the FlashAttention meta functions to mirror the real outputs now. As well I added a meta registration for backwards. I have an XLMR training script and while eager training now works with FlashAttention compiling this module fails with the inductor error down below.
### Questions?
Performance issues vs mem efficient when using torch.nn.mha_forward
TorchCompile -> See purposed solution below.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/92917
Approved by: https://github.com/cpuhrsch
This makes it easier to narrow down who is throwing the error,
instead of having to use gdb.
Signed-off-by: Edward Z. Yang <ezyang@fb.com>
Differential Revision: [D42088781](https://our.internmc.facebook.com/intern/diff/D42088781)
A retry of #89487. Accidentally closed.
## Split `torchgen.api.types` into `types_base`, `types` and `signatures`.
In `types_base`:
* Created base class `CType`. `BaseCType` and `ConstRefCType` etc are inheriting `CType`.
* Only keep abstract type model definitions, such as `BaseCppType`.
In `types`:
* Define `BaseCppType` with `at` and `c10` namespaces.
* All the signatures using these types.
In `signatures`:
* Define all the signatures.
In `__init__`:
* `from ... import *`, suppress flake8 error.
Differential Revision: [D41455634](https://our.internmc.facebook.com/intern/diff/D41455634/)
**NOTE FOR REVIEWERS**: This PR has internal Meta-specific changes or comments, please review them on [Phabricator](https://our.internmc.facebook.com/intern/diff/D41455634/)!
Pull Request resolved: https://github.com/pytorch/pytorch/pull/90589
Approved by: https://github.com/iseeyuan
Our prevailing strategy for symbolic shapes in C++ is to only
write the SymInt version of the code, and pay a slight performance
tax from not knowing if it is symbolic or not. However, there are
some fastpath functions where this tax is unacceptable, and we want
to specialize for the int case. Sometimes, it is easy to template
the function; but when the function involves Tensors, it is not,
because the functions you may want to call are not templated,
e.g., t.view vs t.view_symint
This PR adds an at::symint:: namespace which contains templated
functions for all functions in PyTorch which you can use in this
way. To show this works, I refactored sum_to to stop incorrectly
reinterpret casting and instead use a template. Instead of
t.sizes(), we call at::symint::sizes<T>(t), and so forth.
The template functions are SFINAE'd using a template argument that
is not otherwise used. As such, deduction is impossible. Typically, deduction
is hard anyway, because many of the constructors are ambiguous (this
is why we split foo and foo_symint in the first place). So you must pass
a template argument to these functions.
These functions are codegened into Functions.h so they are subject
to per-operator headers. This matters most for methods, which likely
didn't include the per-operator header, so you will have to add an
include in that case. We never generate method variants for these.
Signed-off-by: Edward Z. Yang <ezyang@fb.com>
Pull Request resolved: https://github.com/pytorch/pytorch/pull/86329
Approved by: https://github.com/bdhirsh, https://github.com/voznesenskym
Now, we also avoid translating SymInt to valueT if you haven't asked
for a SymInt implementation. This makes embedding_dense_backward
work without changes to LTC.
Signed-off-by: Edward Z. Yang <ezyang@fb.com>
Pull Request resolved: https://github.com/pytorch/pytorch/pull/86043
Approved by: https://github.com/wconstab
- Make toIValue accept SymIntNode and SymFloatNode where number (aka Scalar) is
expected
- Binding for symintlistOptional in python arg parser
- Teach translate to convert from IntArrayRef to ArrayRef<int64_t>
- Don't query _symint function for meta info in LTC unless LTC is
code generating a symint function
Signed-off-by: Edward Z. Yang <ezyang@fb.com>
Pull Request resolved: https://github.com/pytorch/pytorch/pull/86042
Approved by: https://github.com/Chillee
Partially fixes: #66328
This PR:
- adds support for `ITensorList` to the dispatcher for:
- computing the dispatch key
- boxing and unboxing `ITensorList`
- modified the codegen for structured kernels:
- codegen APIs use `ITensorList` instead of `ArrayRef<Tensor>`
**Changes summary:**
- Signature changes due to the different APIs:
- dispatcher API (e.g. `BatchingRegistrations.cpp`)
- C++ API (e.g. `TensorShape.cpp`)
- Miscelaneous functions used by codegen'd functions (e.g. `FunctionalTensorWrapper.*`)
- Dispatcher changes for handling `ITensorList` correctly (e.g. `DispatchKeyExtractor.h`)
- Signature changes of `at::cat` due to the need of `const` inside `TensorBody.h`
- Forward declarations of `ITensorList` (e.g. `MethodOperators.h`)
- Codegen changes, special casing structured kernels (e.g. `gen.py`)
**Short description of structured kernels special casing:**
I introduced, mainly, 5 types of changes to the codegen for generating code depending on
whether the kernel is structured or not:
1. Added a `structured_type_override` flag to the `argument_type` function definition of
the affected APIs (mainly the dispatcher and C++ APIs).
- `api/cpp.py`, `api/dispatcher.py`, `api/native.py`
2. Added a `structured_type_override` member to the signature
classes (e.g. `CppSignature`), since `FunctionSchema` doesn't really know whether the
function is structured or not
- `api/types.py`
3. Added a `part_of_structured_group` to `NativeFunction` class, which is just a
convenient function to forward to `structured_type_override` wherever needed
- `model.py`
4. Appropriately changed the rest of the codegen, whenever it used either the signature
classes or the `arguments` function directly
5. Added a check for `const ITensorList&` type wherever there was a check for `TensorList`
Pull Request resolved: https://github.com/pytorch/pytorch/pull/73350
Approved by: https://github.com/bdhirsh
Since we separated at::foo and at::foo_symint there is no benefit
to trying to make initializer lists work in both cases. So we can
get rid of the special different struct.
Signed-off-by: Edward Z. Yang <ezyang@fb.com>
Pull Request resolved: https://github.com/pytorch/pytorch/pull/84837
Approved by: https://github.com/kit1980
Since we separated at::foo and at::foo_symint there is no benefit
to trying to make initializer lists work in both cases. So we can
get rid of the special different struct.
Signed-off-by: Edward Z. Yang <ezyang@fb.com>
Pull Request resolved: https://github.com/pytorch/pytorch/pull/84837
Approved by: https://github.com/kit1980
This fixes two problems:
- First, shape signature didn't respect the symint property (so it
would always mark the operator as symint). This was relatively
easy to fix.
- Second, the call to fallback goes directly through at::_ops, so
it must always be SymInt-aware, even if SymInt is disabled externally.
This was a bit more difficult, because the current LTC codegen
is poorly factored. First, I needed to make it so individual
arguments knew if they were going to be SymInt in LTC or not; second,
I need to plumb enough information about the enclosing bindings so
that I could use translate to do the expressions (previously, it was
just assumed the signatures matched.)
The LTC codegen would do well to have a complete rewrite, but this will
have to do for now, I suppose.
Signed-off-by: Edward Z. Yang <ezyang@fb.com>
Pull Request resolved: https://github.com/pytorch/pytorch/pull/84832
Approved by: https://github.com/wconstab
Something people found confusing was that whether or not a native::
signature would get SymInt or not in its type was based on the dispatch
key. This changes it so that SymInt or not in type is based on whether
or not you have _symint in the name of the kernel or not. This means
that even when we make operators support SymInt, you no longer have to
go and update all the preexisting definitions; instead, you now
selectively write _symint to opt individual kernels into SymInt support.
I then go and update a bunch of kernels that don't have proper SymInt
support to make use of this convention. There is some hacking around
for view generation code.
I also add support for external backends to specify 'symint' operators, for which we generate SymInt signatures instead of regular signatures.
Signed-off-by: Edward Z. Yang <ezyang@fb.com>
Differential Revision: [D39310060](https://our.internmc.facebook.com/intern/diff/D39310060)
Pull Request resolved: https://github.com/pytorch/pytorch/pull/84579
Approved by: https://github.com/wconstab
Also Back out "Revert D39075159: [acc_tensor] Use SymIntArrayRef for overloaded empty.memory_format's signature"
Original commit changeset: dab4a9dba4fa
Original commit changeset: dcaf16c037a9
Original Phabricator Diff: D38984222
Original Phabricator Diff: D39075159
Also update Metal registrations for C++ registration changes.
Also update NNPI registration to account for tightened schema checking
Differential Revision: [D39084762](https://our.internmc.facebook.com/intern/diff/D39084762/)
**NOTE FOR REVIEWERS**: This PR has internal Facebook specific changes or comments, please review them on [Phabricator](https://our.internmc.facebook.com/intern/diff/D39084762/)!
Pull Request resolved: https://github.com/pytorch/pytorch/pull/84173
Approved by: https://github.com/Krovatkin
Previously, we introduced new SymInt overloads for every function we wanted. This led to a lot of boilerplate, and also a lot of confusion about how the overloads needed to be implemented.
This PR takes a simpler but more risky approach: just take the original function and changes its ints to SymInts.
This is BC-breaking in the following ways:
* The C++ API for registering implementations for aten operators will change from int64_t to SymInt whenever you make this change. Code generated registrations in PyTorch do not change as codegen handles the translation automatically, but manual registrations will need to follow the change. Typically, if you now accept a SymInt where you previously only took int64_t, you have to convert it back manually. This will definitely break XLA, see companion PR https://github.com/pytorch/xla/pull/3914 Note that not all dispatch keys get the automatic translation; all the composite keys and Meta keys are modified to take SymInt directly (because they should handle them directly), and so there are adjustments for this.
This is not BC-breaking in the following ways:
* The user facing C++ API remains compatible. Even if a function changes from int to SymInt, the default C++ binding still takes only ints. (e.g., at::empty(IntArrayRef, ...). To call with SymInts, you must call at::empty_symint instead. This involved adding two more signatures to CppSignatureGroup; in many cases I refactored code to iterate over all signatures in the group instead of hard-coding the two that previously existed.
* This is TorchScript compatible; internally we treat SymInts as ints so there is no change to what happens at runtime in TorchScript. In particular, it's OK to reference an empty schema by its old type (using int types), as long as you're not doing string equality (which you shouldn't be), these parse to the same underyling type.
Structure of the PR:
* The general strategy of this PR is that, even when you write `SymInt` inside `native_functions.yaml`, sometimes, we will treat it *as if* it were an `int`. This idea pervades the codegen changes, where we have a translation from SymInt to c10::SymInt or int64_t, and this is controlled by a symint kwarg which I added and then audited all call sites to decide which I wanted. Here are some of the major places where we pick one or the other:
* The C++ FunctionSchema representation represents `SymInt` as `int`. There are a few places we do need to know that we actually have a SymInt and we consult `real_type()` to get the real type in this case. In particular:
* When we do schema validation of C++ operator registration, we must compare against true schema (as the C++ API will provide `c10::SymInt`, and this will only be accepted if the schema is `SymInt`. This is handled with cloneWithRealTypes before we check for schema differences.
* In `toIValue` argument parsing, we parse against the true schema value. For backwards compatibility reasons, I do still accept ints in many places where Layout/SymInt/etc were expected. (Well, accepting int where SymInt is expected is not BC, it's just the right logic!)
* In particular, because SymInt never shows up as type() in FunctionSchema, this means that we no longer need a dedicated Tag::SymInt. This is good, because SymInts never show up in mobile anyway.
* Changes to functorch/aten are mostly about tracking changes to the C++ API registration convention. Additionally, since SymInt overloads no longer exist, registrations for SymInt implementations are deleted. In many cases, the old implementations did not properly support SymInts; I did not add any new functionality with this PR, but I did try to annotate with TODOs where this is work to do. Finally, because the signature of `native::` API changed from int to SymInt, I need to find alternative APIs for people who were directly calling these functions to call. Typically, I insert a new dispatch call when perf doesn't matter, or use `at::compositeexplicitautograd` namespace to handle other caes.
* The change to `make_boxed_from_unboxed_functor.h` is so that we accept a plain IntList IValue anywhere a SymIntList is expected; these are read-only arguments so covariant typing is OK.
* I change how unboxing logic works slightly. Previously, we interpret the C++ type for Layout/etc directly as IntType JIT type, which works well because the incoming IValue is tagged as an integer. Now, we interpret the C++ type for Layout as its true type, e.g., LayoutType (change to `jit_type.h`), but then we accept an int IValue for it anyway. This makes it symmetric with SymInt, where we interpret the C++ type as SymIntType, and then accept SymInt and int IValues for it.
* I renamed the `empty.names` overload to `empty_names` to make it less confusing (I kept mixing it up with the real empty overload)
* I deleted the `empty.SymInt` overload, which ended up killing a pile of functions. (This was originally a separate PR but the profiler expect test was giving me grief so I folded it in.)
* I deleted the LazyDynamicOpsTest tests. These were failing after these changes, and I couldn't figure out why they used to be passing: they make use of `narrow_copy` which didn't actually support SymInts; they were immediately converted to ints.
* I bashed LTC into working. The patches made here are not the end of the story. The big problem is that SymInt translates into Value, but what if you have a list of SymInt? This cannot be conveniently represented in the IR today, since variadic Values are not supported. To work around this, I translate SymInt[] into plain int[] (this is fine for tests because LTC dynamic shapes never actually worked); but this will need to be fixed for proper LTC SymInt support. The LTC codegen also looked somewhat questionable; I added comments based on my code reading.
Signed-off-by: Edward Z. Yang <ezyang@fb.com>
Pull Request resolved: https://github.com/pytorch/pytorch/pull/83628
Approved by: https://github.com/albanD, https://github.com/bdhirsh
Fix use-dict-literal pylint suggestions by changing `dict()` to `{}`. This PR should do the change for every Python file except test/jit/test_list_dict.py, where I think the intent is to test the constructor.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/83718
Approved by: https://github.com/albanD
Previously, we introduced new SymInt overloads for every function we wanted. This led to a lot of boilerplate, and also a lot of confusion about how the overloads needed to be implemented.
This PR takes a simpler but more risky approach: just take the original function and changes its ints to SymInts.
This is BC-breaking in the following ways:
* The C++ API for registering implementations for aten operators will change from int64_t to SymInt whenever you make this change. Code generated registrations in PyTorch do not change as codegen handles the translation automatically, but manual registrations will need to follow the change. Typically, if you now accept a SymInt where you previously only took int64_t, you have to convert it back manually. This will definitely break XLA, see companion PR https://github.com/pytorch/xla/pull/3914 Note that not all dispatch keys get the automatic translation; all the composite keys and Meta keys are modified to take SymInt directly (because they should handle them directly), and so there are adjustments for this.
This is not BC-breaking in the following ways:
* The user facing C++ API remains compatible. Even if a function changes from int to SymInt, the default C++ binding still takes only ints. (e.g., at::empty(IntArrayRef, ...). To call with SymInts, you must call at::empty_symint instead. This involved adding two more signatures to CppSignatureGroup; in many cases I refactored code to iterate over all signatures in the group instead of hard-coding the two that previously existed.
* This is TorchScript compatible; internally we treat SymInts as ints so there is no change to what happens at runtime in TorchScript. In particular, it's OK to reference an empty schema by its old type (using int types), as long as you're not doing string equality (which you shouldn't be), these parse to the same underyling type.
Structure of the PR:
* The general strategy of this PR is that, even when you write `SymInt` inside `native_functions.yaml`, sometimes, we will treat it *as if* it were an `int`. This idea pervades the codegen changes, where we have a translation from SymInt to c10::SymInt or int64_t, and this is controlled by a symint kwarg which I added and then audited all call sites to decide which I wanted. Here are some of the major places where we pick one or the other:
* The C++ FunctionSchema representation represents `SymInt` as `int`. There are a few places we do need to know that we actually have a SymInt and we consult `real_type()` to get the real type in this case. In particular:
* When we do schema validation of C++ operator registration, we must compare against true schema (as the C++ API will provide `c10::SymInt`, and this will only be accepted if the schema is `SymInt`. This is handled with cloneWithRealTypes before we check for schema differences.
* In `toIValue` argument parsing, we parse against the true schema value. For backwards compatibility reasons, I do still accept ints in many places where Layout/SymInt/etc were expected. (Well, accepting int where SymInt is expected is not BC, it's just the right logic!)
* In particular, because SymInt never shows up as type() in FunctionSchema, this means that we no longer need a dedicated Tag::SymInt. This is good, because SymInts never show up in mobile anyway.
* Changes to functorch/aten are mostly about tracking changes to the C++ API registration convention. Additionally, since SymInt overloads no longer exist, registrations for SymInt implementations are deleted. In many cases, the old implementations did not properly support SymInts; I did not add any new functionality with this PR, but I did try to annotate with TODOs where this is work to do. Finally, because the signature of `native::` API changed from int to SymInt, I need to find alternative APIs for people who were directly calling these functions to call. Typically, I insert a new dispatch call when perf doesn't matter, or use `at::compositeexplicitautograd` namespace to handle other caes.
* The change to `make_boxed_from_unboxed_functor.h` is so that we accept a plain IntList IValue anywhere a SymIntList is expected; these are read-only arguments so covariant typing is OK.
* I change how unboxing logic works slightly. Previously, we interpret the C++ type for Layout/etc directly as IntType JIT type, which works well because the incoming IValue is tagged as an integer. Now, we interpret the C++ type for Layout as its true type, e.g., LayoutType (change to `jit_type.h`), but then we accept an int IValue for it anyway. This makes it symmetric with SymInt, where we interpret the C++ type as SymIntType, and then accept SymInt and int IValues for it.
* I renamed the `empty.names` overload to `empty_names` to make it less confusing (I kept mixing it up with the real empty overload)
* I deleted the `empty.SymInt` overload, which ended up killing a pile of functions. (This was originally a separate PR but the profiler expect test was giving me grief so I folded it in.)
* I deleted the LazyDynamicOpsTest tests. These were failing after these changes, and I couldn't figure out why they used to be passing: they make use of `narrow_copy` which didn't actually support SymInts; they were immediately converted to ints.
* I bashed LTC into working. The patches made here are not the end of the story. The big problem is that SymInt translates into Value, but what if you have a list of SymInt? This cannot be conveniently represented in the IR today, since variadic Values are not supported. To work around this, I translate SymInt[] into plain int[] (this is fine for tests because LTC dynamic shapes never actually worked); but this will need to be fixed for proper LTC SymInt support. The LTC codegen also looked somewhat questionable; I added comments based on my code reading.
Signed-off-by: Edward Z. Yang <ezyang@fb.com>
Pull Request resolved: https://github.com/pytorch/pytorch/pull/83628
Approved by: https://github.com/albanD, https://github.com/bdhirsh
Summary:
Previously we don't generate out variant (both schema and kernel) for an operator with functional variant only. This adds support for that and adds test.
## Changes on `native_function_generation.py`
We are generating out variant for all functional variants if possible. This PR introduces a lot of newly generated out variants and `native_functions.yaml` needs to incorporate the changes by adding `autogen` keywords.
The logic for determining what operators we should generate an out variant for is the following:
1. No existing out variant for this `NativeFunction`
2. Contains an existing in place, mutable or functional variant
3. Contains at least 1 tensor like return(s)
For operators matching the first two conditions but failing the third, I listed them in `FUNCTIONAL_OPS_THAT_CANNOT_GET_AN_OUT_VARIANT`.
## Special handling
The following operators satisfy all 3 criteria above but we chose to not autogen them, with some reasons.
* `mkldnn_adaptive_avg_pool2d`, the generated out variant `mkldnn_adaptive_avg_pool2d.out` is colliding with the `mkldnn_adaptive_avg_pool2d_out` kernel in `adaptive_avg_pool2d.out` operator. I manually created `mkldnn_adaptive_avg_pool2d.out` and renamed `mkldnn_adaptive_avg_pool2d_out` to `mkldnn_adaptive_avg_pool2d_out_stub`.
* `min`, `max` and `mean`. There already exist `min.out`, `max.out` and `mean.out` but they are having different semantics with the functional ones. I manually created `min.unary_out`, `max.unary_out` and `mean.dtype_out` to disambiguate.
## Autograd Changes
We introduced a logic to not match derivatives info in `derivatives.yaml` to out variant, since we are generating `NOT_IMPLEMENTED` kernels for those out variants anyway. The issue we are seeing with the original logic is that it doesn't handle `TensorOption` arguments really well. For example we have these two operators:
* `_to_copy(Tensor self, *, ScalarType? dtype=None, Layout? layout=None, Device? device=None, bool? pin_memory=None, bool non_blocking=False, MemoryFormat? memory_format=None) -> Tensor`
* `_to_copy.out(Tensor self, *, bool non_blocking=False, MemoryFormat? memory_format=None, Tensor(a!) out) -> Tensor(a!)`
If we uses `_to_copy` derivative info, there will be compilation error since `dtype` is missing from `_to_copy.out` signature.
Test Plan: Rely on unit test
Differential Revision: D37832342
Pull Request resolved: https://github.com/pytorch/pytorch/pull/81437
Approved by: https://github.com/iseeyuan, https://github.com/bdhirsh
`derivatives.yaml` can now take a `dispatch` entry which registers per-autograd dispatch key derivatives such as
```
name: foo(Tensor self, Tensor y) -> Tensor
dispatch:
Default:
x: grad
y: grad.expand(y.sizes())
AutogradNestedTensor:
x: grad
y: NestedTensor_foo_backward(grad, y)
output_differentiabilty: [True]
```
However the old schema where there is no `dispatch` entry is still supported.
Would greatly appreciate feedback on *how to improve the testing strategy* of this PR, currently have registered an aten test op in TestOps.cpp with dummy gradients in derivatives.yaml and have some tests in test_autograd.py:TestAutogradMultipleDispatch but I am not sure whether these are sufficiently rigorous.
Additionally, this PR also makes the assumption that sets like [VIEW_FUNCTIONS](ff5399e528/tools/autograd/gen_inplace_or_view_type.py (L60)) are per-native-function and not per-native-function-and-dispatch-key. I'm not sure whether this is necessarily the case, *would there ever be a situation where (e.g. a nested_tensor op is a view op but the aten function is not or vice versa?)*
* __->__ #82801
Pull Request resolved: https://github.com/pytorch/pytorch/pull/82801
Approved by: https://github.com/bhosmer, https://github.com/albanD
Fixes#81774
`TensorOptions` arguments in the JIT schema are optional, but in the Python API these were being translated to non-optional but with a default value. This change makes the arguments accept `None` for consistency with the JIT schema. However, it also means that `dtype=c10::nullopt` was previously completely untested so this also fixes several related bugs.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/82241
Approved by: https://github.com/ngimel
Deprecated signatures are currently "parsed" manually to find the
relative order of the argument names and all other information is
inferred from the aten schema for the non-deprecated overload.
However, this leads to problems if the argument names don't match or
if there are multiple candidates that match the ATen function call.
Instead, this makes the deprecated function a full FunctionSchema and
so the entire python signature comes solely from the deprecated
schema, with the `aten:` clause only used for the dispatch lambda call.
I have confirmed locally that there is no change to
`python_torch_functionsEverything.cpp`.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/82179
Approved by: https://github.com/albanD
Currently any function with a default dtype other than None has to be
manually entered into this function. Instead, this reads the default
directly from `native_functions.yaml`. In order to do this, I also
change `PythonSignatureGroup` to take `tensor_options_args` from the
functional variant since the out variant doesn't actually have tensor
options arguments to take the default values from.
Also note that we need to use `default_init` instead of `default`
because the out argument version doesn't have a `tensor_options`
argument to extract the default value from and so the PythonSignature
objects wouldn't match.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/81479
Approved by: https://github.com/albanD
This should fix the last issue that @anijain2305 hit when running ResNet with TorchDynamo <> functionalization.
Today if you try to call an `OpOverloadPacket` from python with some arguments, we will use the types of those arguments to perform overload resolution. With some functional variants of ops, this can be ambiguous.
Today this affects just one op: `_fused_moving_avg_obs_fq_helper`, although it would potentially affect e.g. `native_batch_norm` in the future.
Example:
```
# There are technically two overloads:
# torch.ops.aten._fused_moving_avg_obs_fq_helper.default (returns 2 argument, mutates 4 of its inputs inplace)
# torch.ops.aten._fused_moving_avg_obs_fq_helper.functional (returns 6 argument, mutates none of its inputs)
# We pick the wrong one - no way to know that we should pick the functional one, just from the call site.
outs = torch.ops.aten._fused_moving_avg_obs_fq_helper(a, a, a, a, a, a, a, 1.0, 0, 1, 0)
# raises an error - tries to call the overload with only 2 returns
return _fused_moving_avg_obs_fq_helper_functional[5]
```
Specifically, functionalization will bake `_fused_moving_avg_obs_fq_helper.functional` into the graph, but when AOTAutograd tries to compile with TorchScript, it needs to remove the overload name (TS doesn't know how to parse overload names directly, so we need to remove the overload name and let it infer the right overload at runtime later- so it picks the wrong one).
The situation is pretty similar to inplace; `ops.aten.add` and `ops.aten.add_` represent two different `OverloadPacket` objects; they can't be overloads of the same op, because their schemas would be ambiguous - the alias annotations are different, but that isn't enough to disambiguate).
In this PR, I try to fix the situation in a pretty similar way to how we handle `inplace` in the data model: `inplace` ops get their own base operator name, but they are represented as a flag inside of `BaseOperatorName` in the data model.
Two other important changes that I made as part of this PR:
(1) Originally, there were ~100 different `*_functional` operators: e.g. we had operators named `resize.functional` and `zero.functional`. The `_functional` bit isn't actually necessary in most cases: it's only necessary for operators that **also** have a `SchemaKind.mutable` variant, where `_fused_moving_avg_obs_fq_helper` is the only op that fits that description today. So I removed the unnecessary notion of "functional" from those other ops. I also added a bunch of assertions to force this restriction.
I think that makes more sense in the long run, because it eliminates an unnecessary difference in the model. E.g. we don't have `add_.Tensor` and `add.Tensor_functional`. We just have `add_.Tensor` and `add.Tensor`.
(2) I noticed that we actually still weren't pairing up a bunch of `_foreach` operators correctly, because their input arguments were different (`self` vs. `tensors`). Since they're private API's, I went ahead and changed the argument names directly so they get matched up. Before this PR, we were generating a separate `_foreach_add` and `_foreach_add.functional` variant in a bunch of cases, that really did the same thing (but happened to have a different name for the first argument).
Pull Request resolved: https://github.com/pytorch/pytorch/pull/80556
Approved by: https://github.com/ezyang, https://github.com/albanD
Due to implicit conversion shenanigans, having both IntArrayRef
and SymIntArrayRef overloads makes {} ambiguous. While we could
fix this by making a single unified type that accepts all the overloads
we want, an easier fix was to just push the SymIntArrayRef overload
to its own name.
Signed-off-by: Edward Z. Yang <ezyangfb.com>
Pull Request resolved: https://github.com/pytorch/pytorch/pull/79281
Approved by: https://github.com/suo
