Fixes#130087
This patch tries to provide a built-in id function implementation for TensorVariable when the id function is called on tensors like module parameters. The id function call on intermediate tensors is not supported.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/130100
Approved by: https://github.com/anijain2305
Add similar semantics for creating a buffer object similar to creating a parameter. This is done by introducing a new Buffer class that can be used for type disambiguation. The underlying functionality of registering a buffer remains the same as the register_buffer method has not been changed. The persistent parameter in the Buffer type is to indicate whether a buffer object should be persistent or not. Other non-test changes have to do with getting the new Buffer type recognized by inductor and dynamo. Remaining changes are test changes to make sure that the Buffer type can be used as a drop in replacement for register_buffer as it just leads to register_buffer being called. The addition of this new functionality still allows for normal tensors to be used as buffers so these changes are intended to be backwards compatible.
Fixes#35735
Co-authored-by: Mikayla Gawarecki <mikaylagawarecki@gmail.com>
Pull Request resolved: https://github.com/pytorch/pytorch/pull/125971
Approved by: https://github.com/albanD, https://github.com/anijain2305, https://github.com/mlazos
I didn't test this path when creating the orchestrator. This PR fixes
that path to work in the capture_triton path. The problem is that we are
handling a value that is an int (in the capture_triton path) and a
ConstantVariable (in the Dynamo triton path) so we abstract that out in
the orchestrator.
Test Plan:
- new tests
Pull Request resolved: https://github.com/pytorch/pytorch/pull/132143
Approved by: https://github.com/oulgen
Fixes https://github.com/pytorch/pytorch/issues/121353
our handle for `.data` in dynamo today basically just converts `y = x.data` into `y = x.detach()`. The semantics of these two ops are not quite the same, because:
(1) any future mutations on `x.data` will be fully ignored by autograd
(2) any mutations on `x.detach()` will bump x's version counter
the linked model does a .data mutation that is hidden from autograd in eager, but ends up erroring during AOTDispatcher tracing.
I updated dynamo's handling so that:
(1) when dynamo sees a call to `getattr(tensor, "data")` and calls `.detach()` we set a flag on the returned `TensorVariable` indicating it came from `.data`
(2) on any tensor method that we call with an input `TensorVariable` with this flag turned on, we proxy autograd's `preserve_version_counter` logic into the graph, to properly reset the VC after the op is run.
One thing to note is that I don't actually do this on every op that we pass the tensor to: I only do it for tensor methods that appear to be mutations (by checking for a trailing underscore). My thought was that:
(1) I didn't want to do this for **every** op that you pass `y` into, since that will e.g. triple the number of nodes in the graph, and could cause compile time regressions if you use .data
(2) this situation is pretty rare in general, and I'm hoping that "tensor method mutations" cover most reasonable mutation cases. If we manage to miss a case, you will get a loud error during tracing anyway, so there is not a safety issue.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/131403
Approved by: https://github.com/anijain2305, https://github.com/zou3519
Fixes https://github.com/pytorch/pytorch/issues/130750.
Repro of lazy/eager `map` discrepancy without `islice`:
```python
def fn(a, b):
y = 1
def f(x):
nonlocal y
y += 1
return x
l = list(zip([a, b], map(f, [1, 2, 3, 4])))
return a + y
```
The major change is that we implement `MapVariable` and `ZipVariable` based on `IteratorVariable`. Before, `map` and `zip` were being traced by immediately unpacking the result as a `TupleVariable`, which is wrong in cases such as the example above.
`MapVariable`s are not allowed to be unpacked while `ZipVariable`s can only be unpacked if all of its iterables can also be unpacked.
We also add new `[has_]force_unpack_var_sequence` methods to `VariableTracker` for the case where it is safe to unpack the entire sequence lazily, e.g., when building a list from a map (i.e. `list(map(f, ...))`).
Pull Request resolved: https://github.com/pytorch/pytorch/pull/131413
Approved by: https://github.com/anijain2305
Fixes https://github.com/pytorch/pytorch/issues/103602.
This PR implements the idea of "if someone creates a string and then ends up not using it, we would prefer to NOT have specialized." mentioned in above issue. Specifically, we create a lazy variable tracker instead of ConstantVariable when we're in FORMAT_VALUE, and when the lazy variable tracker is realized (i.e. it's going to be used), we create a ConstantVariable and the specialization/guarding happens at the time of realization.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/131529
Approved by: https://github.com/ezyang
This PR implements an opt-in configuration option for synchronizing compilation across all ranks at the end of Dynamo tracing (and potentially, other places in the future). There are two pieces to this PR:
1. Implementing infrastructure for compiler collectives (DistributedState/LocalState, the actual collective)
2. Using this infrastructure to synchronize automatic dynamic choices across all ranks
The infrastructure in part one can be used for other purposes, just add more (serializable) fields to LocalState.
Here is how automatic dynamic synchronization works:
1. Preflight in "torch/_dynamo/variables/builder.py": On the first Dynamo trace run, we trace without automatic dynamic at all; we assume all Tensor inputs that are not otherwise marked are static. This run is purely to collect all Tensor input sizes in the program.
2. torch/_dynamo/output_graph.py: At the end of the first Dynamo trace run, we perform a compiler collective to distribute all Tensor input sizes to all ranks. Then, we restart Dynamo
3. Apply the updates in "torch/_dynamo/variables/builder.py": Now that we have all sizes for every rank, we now update frame state with the observed sizes for all ranks, in rank order. Under the assumption that frame state is consistent on all ranks, this series of updates will preserve consistency.
For future work, it would be safer if we force a consistent hint on all ranks; this is more involved as we have to interpose in fakification.
Signed-off-by: Edward Z. Yang <ezyang@meta.com>
Pull Request resolved: https://github.com/pytorch/pytorch/pull/130935
Approved by: https://github.com/jansel
The issue addressed is that compiled autograd changes the calling convention of the FX graph to only have a single placeholder which contains a list of inputs. In this case, the meta of the tensor input nodes don't contain the `tensor_dict` meta. This adds them.
The context is that `tensor_dict` is used to convey if a tensor is an input with a static address.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/131556
Approved by: https://github.com/anijain2305
Summary:
- Log export errors to Scuba and mark them with "classified" and "unclassified"
- Classify errors by exception type (ALLOW_LIST) and a `case_name` attribute
- Add `case_name` for some exceptions.
Test Plan:
Running the code below logs a classified error to `torch_export_usage` table in Scuba.
```
import torch
from torch._export.db.case import SupportLevel
class TorchSymMin(torch.nn.Module):
"""
torch.sym_min operator is not supported in export.
"""
def forward(self, x):
return x.sum() + torch.sym_min(x.size(0), 100)
example_args = (torch.randn(3, 2),)
tags = {"torch.operator"}
support_level = SupportLevel.NOT_SUPPORTED_YET
model = TorchSymMin()
torch.export.export(model, example_args)
``
Differential Revision: D59981459
Pull Request resolved: https://github.com/pytorch/pytorch/pull/131327
Approved by: https://github.com/zhxchen17
All the changes brought by the original PR have been addressed in alternative ways in the stack. Why the original PR has to be reverted requires more effort because there is some bad interaction with export.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/131058
Approved by: https://github.com/williamwen42
Uses `dict.fromkeys` whenever possible as covered by flake8-comprehensions rule C420. While the ruff rule RUF025 is still in preview, flake8-comprehensions have added a new rule which covers this. Use dict.fromkeys is faster when the value being added to the dictionary is the same at every iteration and is immutable, it also removes an unnecessary dict comprehension.
This rule will be enabled with our current ruleset in RUF in 0.6 as C420.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/130699
Approved by: https://github.com/lezcano, https://github.com/ezyang
This PR marks all buffers and parameters of an NNModule as static using the `mark_static_address` API. As a result, when tensors are passed to AOT, the `tensor_dict` metadata of placeholder nodes will contain the `static_address_type` key, indicating which graph argument positions are static for cudagraphs.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/130391
Approved by: https://github.com/anijain2305
Reland https://github.com/pytorch/pytorch/pull/128709.
When the input predicate is a python constant, we specialize into one of the branches and warn users that torch.cond is not preserving the dynamism. The previous behavior is that we baked in True/False in the cond operator. This can be confusing. In this PR, we change it to be specializing into one of the branches when the inputs are constants.
We additionally change the naming of cond operator to default one without overriding its name. This allows better testing on de-serialized graph.
Test Plan:
The predicate in some existing tests is the result of a shape comparison. When no dynamic shape is involved, the predicate is a python bool. To fix them, we either change the predicate to be some data-dependent tensor or change the test to check cond is specialized as one of the branches,
Pull Request resolved: https://github.com/pytorch/pytorch/pull/130493
Approved by: https://github.com/BoyuanFeng
Sometimes, it could be difficult to write a fake class e.g. when the original implementation is using some third-party libraries or users are certain that the class is safe to trace with the real object.
This PR allows user to specify their intention by implementing a "safe_to_trace_with_real_obj" method on their script class.
Test Plan:
`pytest test/export/test_torchbind.py -k safe`
Pull Request resolved: https://github.com/pytorch/pytorch/pull/129586
Approved by: https://github.com/zou3519
When the input predicate is a python constant, we specialize into one of the branches and warn users that torch.cond is not preserving the dynamism. The previous behavior is that we baked in True/False in the cond operator. This can be confusing. In this PR, we change it to be specializing into one of the branches when the inputs are constants.
We additionally change the naming of cond operator to default one without overriding its name. This allows better testing on de-serialized graph.
Test Plan:
The predicate in some existing tests is the result of a shape comparison. When no dynamic shape is involved, the predicate is a python bool. To fix them, we either change the predicate to be some data-dependent tensor or change the test to check cond is specialized as one of the branches,
Differential Revision: [D59589709](https://our.internmc.facebook.com/intern/diff/D59589709)
Pull Request resolved: https://github.com/pytorch/pytorch/pull/128709
Approved by: https://github.com/zou3519
TritonKernelVariable's logic tells us how to go from a user-defined
triton kernel and a grid to a call to the triton_kernel_wrapper_mutation
HOP. We want to re-use this in a setting without Dynamo; in the next PR
up, we create a new decorator (capture_triton) that, when applied to a
triton kernel, transforms a call to the triton kernel into a call
to the triton_kernel_wrapper_mutation HOP.
Test Plan:
- existing tests
Pull Request resolved: https://github.com/pytorch/pytorch/pull/130177
Approved by: https://github.com/oulgen, https://github.com/ydwu4
Previously, subgraph input names were whatever the input proxies were,
which were confusing. This PR changes those names to be
whatever the names of the arguments the functions being
speculate_subgraph'ed are. This is best-effort: if we can't figure it
out then we go back to the previous strategy.
Test Plan:
- existing expecttests
Pull Request resolved: https://github.com/pytorch/pytorch/pull/130255
Approved by: https://github.com/ydwu4
Re-organize ```block_mask``` related arguments a tuple to reduce the individual argument number. I was trying to use named tuple, but aot autograd doesn't work well with named tuple. The only downside of using tuple rather than named tuple is we need to use index to access its element. But we only need this at one place, it should be fine.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/129831
Approved by: https://github.com/Chillee, https://github.com/drisspg
Hard to write tests. This PR makes many test pass in the stack such as
`PYTORCH_TEST_WITH_DYNAMO=1 pytest test/test_ao_sparsity.py::TestComposability::test_convert_without_squash_mask`
Pull Request resolved: https://github.com/pytorch/pytorch/pull/129858
Approved by: https://github.com/mlazos
ghstack dependencies: #129830
Fixes#129601
Background: it's possible that a traceable wrapper subclass will have an optional inner tensor constituent (e.g. NJT's cached min / max sequence lengths). To specify this, the subclass's `__tensor_flatten__()` impl should leave out any unspecified optional inner tensors in the returned list of `attrs`.
This PR guards on the list of inner tensor `attrs` returned in `subclass.__tensor_flatten__()[0]`.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/129618
Approved by: https://github.com/anijain2305
When the input predicate is a python constant, we specialize into one of the branches and warn users that torch.cond is not preserving the dynamism. The previous behavior is that we baked in True/False in the cond operator. This can be confusing. In this PR, we change it to be specializing into one of the branches when the inputs are constants.
We additionally change the naming of cond operator to default one without overriding its name. This allows better testing on de-serialized graph.
Test Plan:
The predicate in some existing tests is the result of a shape comparison. When no dynamic shape is involved, the predicate is a python bool. To fix them, we either change the predicate to be some data-dependent tensor or change the test to check cond is specialized as one of the branches,
Pull Request resolved: https://github.com/pytorch/pytorch/pull/128709
Approved by: https://github.com/zou3519
Summary: Somehow the delegate returns a real tensor result even though we pass in fake tensors. So here we need to convert the result to fake.
Test Plan: `buck2 run @//mode/dev-nosan //on_device_ai/helios/multi_zion:multi_zion_test -- -r test_single_delegate_dsp_only`
Differential Revision: D58617091
Pull Request resolved: https://github.com/pytorch/pytorch/pull/128752
Approved by: https://github.com/ydwu4
# Compile time for eager backend
## AlbertForMaskedLM
No inlining - 3.65 seconds
Inlining on main - 7.48 seconds
Inlining + this PR - 6.70 seconds
## MobileBertForMaskedLM
No inlining - 26.90 seconds
Inlining on main - 48.21 seconds
Inlining + this PR - 43.85 seconds
*Next PR in the stack makes the total compile time better/comparable to no inlining*
Pull Request resolved: https://github.com/pytorch/pytorch/pull/129315
Approved by: https://github.com/jansel
ghstack dependencies: #129316
This code is unused because we just inline the `.parameters` call. The code was also wrong because side-effects only track the first level of mutations. An object might not marked mutated if one of the child objects (like a dict) is mutated.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/129316
Approved by: https://github.com/jansel
Significant bytecode generation API change!
The new suggested convention to generating bytecode to call a function is now to wrap instructions that push a callable to the stack with `add_push_null`, then that callable is called with `create_call_function` with `push_null=False` (see diff for examples).
In Python 3.13, NULL is now expected to be pushed after the callable. In <=3.12, the NULL was pushed before the callable. This change abstracts away the exact placement of the NULL, but the developer must be aware that a NULL may be needed when codegen'ing a callable.
This abstraction also reduces the need for the `push_null=True` option in `create_call_function`, which removes the need to rotate a NULL to the right place on the stack with a sequence of `SWAP` instructions.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/129172
Approved by: https://github.com/jansel
Adds support for `Variable._execution_engine.queue_callback()`, which is used in FSDP2.
Important tests:
- `pytest -rA test/inductor/test_compiled_autograd.py::TestCompiledAutograd::test_callback_graph_break_throws_error`
- `pytest -rA test/inductor/test_compiled_autograd.py::TestAutogradWithCompiledAutograd::test_callback_adds_callback`
- `PYTORCH_TEST_WITH_DYNAMO=1 python test/test_autograd.py -k TestAutograd.test_callback_adds_callback`
Pull Request resolved: https://github.com/pytorch/pytorch/pull/126366
Approved by: https://github.com/xmfan
Improve Dynamo to support the FSDP2 `use_training_state()` context manager.
Test command:
`
pytest -rA test/distributed/_composable/fsdp/test_fully_shard_compile.py::TestFullyShardCompile::test_dynamo_trace_use_training_state
`
Pull Request resolved: https://github.com/pytorch/pytorch/pull/127854
Approved by: https://github.com/yanboliang
Fixes https://github.com/pytorch/pytorch/issues/125720
I was earlier worried that DELETE_* or STORE_* on referent values should result in a graph break, because they could invalidate the weak ref. But then @zou3519 pointed out that weakref invalidation will happen EVENTUALLY, CPython provides no guarantees when the weakref will be invalidated (even when the user calls del x and x is the last reference).
So any code that relies on del x to invalidate the weakref of x right away is BAD code. CPython provide no guarantees. Therefore we can (ab)use this nuance, and can just ignore DELETE_* or STORE_* on the referent objects.
The only corner case is when Dynamo is reconstructing the weakref object. Dynamo will have a hard time being correct here, so just SKIP_FRAME on such a case. This is rare.
Cpython notes
1) https://docs.python.org/3/library/weakref.html
2) https://docs.python.org/3/reference/datamodel.html#index-2
Pull Request resolved: https://github.com/pytorch/pytorch/pull/128533
Approved by: https://github.com/jansel
FIXES#113263. Same idea as in https://github.com/pytorch/pytorch/pull/113417, but we need a more intrusive C API to silently nop default saved tensor hooks, in order to support user-code that use torch.autograd.disable_saved_tensors_hooks (see test_unpack_hooks_can_be_disabled). We mock the output of get_hooks while leaving push/pop untouched.
For compiled autograd, we're firing pack hooks once and unpack hooks twice right now, I'll look into this separately from this issue.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/123196
Approved by: https://github.com/soulitzer
This is a short-term fix (for 2.4). In the longer term we should
fix https://github.com/pytorch/pytorch/issues/128430
The problem is that warnings.warn that are inside Dynamo print
all the time. Python warnings are supposed to print once, unless their
cache is reset: Dynamo ends up resetting that cache everytime it runs.
As a workaround we provide our own warn_once cache that is keyed on the
warning msg. I am not worried about this increasing memory usage because
that's effectively what python's warnings.warn cache does.
Test Plan:
- fix tests.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/128456
Approved by: https://github.com/anijain2305
Today inlining builtin nn modules is not compatible with parameter freezing. Freezing parameters and then constant folding them through the graph relies on the assumption that they will not be inputs and will be static across calls to the same graph. When inlining builtin nn modules this assumption is broken and we reuse the same graph for different instances of the same nn module. There are three options 1) abandon constant folding, 2) create a dispatcher layer (like cudagraphs) which will dispatch to the correct constant-folded graph for each distinct set of parameters or 3) recompile
This PR implements 3 by introducing guards on the parameter pointers. This was due to freezing being relatively rare and performance sensistive. 2 Had many more unknowns and 1 is not a viable option due to the drop in performance.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/128355
Approved by: https://github.com/anijain2305
Tracing through `__init__` is important because it initializes (calls STORE_ATTR) on members. By doing that, we kick in the mutation tracking for these objects. So, things like mutating `_modules` etc is tracked automatically.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/126578
Approved by: https://github.com/jansel
ghstack dependencies: #128001
Before the pr, we have a graph break for `callable(nn_module)`:
```python
class M(nn.Module):
def forward(self, x):
return x.sin()
def f(m):
return callable(m)
res = torch.compile(f, fullgraph=True)(M())
```
```
Traceback (most recent call last):
File "/data/users/yidi/pytorch/t.py", line 17, in <module>
out = torch.compile(f, backend="eager", fullgraph=True)(M())
File "/data/users/yidi/pytorch/torch/_dynamo/eval_frame.py", line 414, in _fn
return fn(*args, **kwargs)
File "/data/users/yidi/pytorch/torch/_dynamo/convert_frame.py", line 1077, in catch_errors
return callback(frame, cache_entry, hooks, frame_state, skip=1)
File "/data/users/yidi/pytorch/torch/_dynamo/convert_frame.py", line 456, in _convert_frame_assert
return _compile(
File "/data/users/yidi/pytorch/torch/_utils_internal.py", line 74, in wrapper_function
return function(*args, **kwargs)
File "/home/yidi/.conda/envs/pytorch/lib/python3.10/contextlib.py", line 79, in inner
return func(*args, **kwds)
File "/data/users/yidi/pytorch/torch/_dynamo/convert_frame.py", line 799, in _compile
guarded_code = compile_inner(code, one_graph, hooks, transform)
File "/data/users/yidi/pytorch/torch/_dynamo/utils.py", line 210, in time_wrapper
r = func(*args, **kwargs)
File "/data/users/yidi/pytorch/torch/_dynamo/convert_frame.py", line 618, in compile_inner
out_code = transform_code_object(code, transform)
File "/data/users/yidi/pytorch/torch/_dynamo/bytecode_transformation.py", line 1167, in transform_code_object
transformations(instructions, code_options)
File "/data/users/yidi/pytorch/torch/_dynamo/convert_frame.py", line 177, in _fn
return fn(*args, **kwargs)
File "/data/users/yidi/pytorch/torch/_dynamo/convert_frame.py", line 564, in transform
tracer.run()
File "/data/users/yidi/pytorch/torch/_dynamo/symbolic_convert.py", line 2244, in run
super().run()
File "/data/users/yidi/pytorch/torch/_dynamo/symbolic_convert.py", line 886, in run
while self.step():
File "/data/users/yidi/pytorch/torch/_dynamo/symbolic_convert.py", line 801, in step
self.dispatch_table[inst.opcode](self, inst)
File "/data/users/yidi/pytorch/torch/_dynamo/symbolic_convert.py", line 496, in wrapper
return inner_fn(self, inst)
File "/data/users/yidi/pytorch/torch/_dynamo/symbolic_convert.py", line 1255, in CALL_FUNCTION
self.call_function(fn, args, {})
File "/data/users/yidi/pytorch/torch/_dynamo/symbolic_convert.py", line 739, in call_function
self.push(fn.call_function(self, args, kwargs))
File "/data/users/yidi/pytorch/torch/_dynamo/variables/builtin.py", line 948, in call_function
return handler(tx, args, kwargs)
File "/data/users/yidi/pytorch/torch/_dynamo/variables/builtin.py", line 711, in <lambda>
return lambda tx, args, kwargs: obj.call_function(
File "/data/users/yidi/pytorch/torch/_dynamo/variables/builtin.py", line 948, in call_function
return handler(tx, args, kwargs)
File "/data/users/yidi/pytorch/torch/_dynamo/variables/builtin.py", line 835, in builtin_dipatch
unimplemented(error_msg)
File "/data/users/yidi/pytorch/torch/_dynamo/exc.py", line 216, in unimplemented
raise Unsupported(msg)
torch._dynamo.exc.Unsupported: builtin: callable [<class 'torch._dynamo.variables.nn_module.NNModuleVariable'>] False
```
Pull Request resolved: https://github.com/pytorch/pytorch/pull/127026
Approved by: https://github.com/jansel
This PR requires a little justification, but let's start with what it does first:
1. When you have a 0d CPU scalar int64/float64 tensor input to a graph, we will preallocate a backed SymInt/SymFloat corresponding to what you would get if you call item() on this tensor. This means you can freely change your input to be a Python int/float or a Tensor with an item() call and end up with exactly the same level of expressivity (specifically, you can guard on the internal SymInt/SymFloat no matter what). By default, the source of the backed SymInt/SymFloat is `L['tensor'].item()`, but if you have promoted a float input into a Tensor, we will cancel out `torch.as_tensor(L['float']).item()` into just `L['float']`.
2. We switch wrap_symfloat to use this, instead of hand crafting the new SymNodeVariable. Everything works out, except that we carefully pass the item() result to tracked fakes (and not the fake Tensor argument)
OK, so why do this at all? There is some marginal benefit where now some item() calls on scalar inputs can be guarded on, but IMO this is a pretty marginal benefit, and if it was the only reason, I wouldn't do this. The real reason for this is that I need to be able to propagate fake tensors through the graphs that are produced by Dynamo, and if I am doing the old custom wrap_symfloat logic, there's no way I can do this, because ordinarily an item() call will cause an unbacked SymInt when I reallocate.
The other obvious way to solve the problem above is to make a HOP alternative that item() that "bakes in" the backed SymInt its supposed to return. But this strategy seems more parsimonious, and it does have the marginal benefit I mentioned above. The main downside is that what I have to do next, is make it so that when I run tensor computation, I also apply the equivalent operations to the SymInt/SymFloat as well. That's next PR.
Signed-off-by: Edward Z. Yang <ezyang@meta.com>
Pull Request resolved: https://github.com/pytorch/pytorch/pull/126245
Approved by: https://github.com/eellison
ghstack dependencies: #126637
I wasn't paying enough attention and didn't notice that LOAD_DEREF is
defined differently for InliningInstructionTranslator. Match it up with
the code there.
This also fixes comptime.print(), which was broken, because closing over
an argument turned it into a cell rather than a regular local.
Signed-off-by: Edward Z. Yang <ezyang@meta.com>
Pull Request resolved: https://github.com/pytorch/pytorch/pull/126637
Approved by: https://github.com/yanboliang
The big idea is that floats are treated as Tensors on input/output to the FX graph, but on the inside, we immediately call item() on the synthetic Tensor and record regular float operations on it. Canonicalization to Tensor operations will happen in a standalone FX pass. This behavior is controlled by `specialize_float` config variable when set to False.
The generated graph looks like this for the test `test_unspec_float_output`:
```
def forward(self, L_x_: "f32[3]", L_y_: "f32[]"):
l_x_ = L_x_
l_y_ = L_y_
# File: /data/users/ezyang/a/pytorch/test/dynamo/test_unspec.py:511 in f, code: return x + 1, y * 2
add: "f32[3]" = l_x_ + 1; l_x_ = None
item: "Sym(zf0)" = l_y_.item(); l_y_ = None
mul: "Sym(2*zf0)" = item * 2; item = None
scalar_tensor: "f32[]" = torch.scalar_tensor(mul); mul = None
return (add, scalar_tensor)
```
The ingredients:
* **torch/_dynamo/variables/builder.py** When `specialize_float` is False, we wrap float literals with `wrap_symfloat`. This is an unholy mashup of `wrap_symint` and `wrap_unspecialized_primitive`. The overall strategy is that we first generate a tensor argument (because that's what we want to show up into the FX graph), but then immediately call item() on the tensor argument to get a SymNodeVariable, which we will do the rest of the tracing with. Importantly, this SymNodeVariable is backed with the source of the original float: this means we can guard on the resulting value (something we could NOT do with UnspecializedPythonVariable). This has to be done manually, because if you literally call item() on the tensor, you will end up with an unbacked float. There is a bit of copy paste from wrap_symint and wrap_unspecialized_primitive which we can try to factor out, but this really is its own thing and you should review every line of code in the function.
* **torch/fx/experimental/symbolic_shapes.py** We now can generate guards on float inputs, and these guards are handled inside of ShapeEnv. So we need to be able to allocate (backed!) float symbols, and produce guards for them. Fairly straightforward generalization.
* **torch/_dynamo/codegen.py** I also need to maintain the invariant that there are no float outputs to the FX graph. I chose to do this at codegen time. When we detect a SymNodeVariable on the return stack for a float, we on the fly convert it (via `as_tensor`) to a TensorVariable, which is the true output. We then special case the output bytecode to call item() on it again. The tensor conversion is memoized on SymNodeVariable since we typically run the code generation process twice.
Signed-off-by: Edward Z. Yang <ezyang@meta.com>
Pull Request resolved: https://github.com/pytorch/pytorch/pull/125325
Approved by: https://github.com/lezcano, https://github.com/jansel
- `FakeContext` hides all fields other than ctx.saved_tensors, this dynamo errors when the autograd.Function.backward uses other attrs on ctx and it also doesn't allow fallback to eager.
- If we remove it, we still can't fallback to eager: node variables are already freed (ctx.saved_tensors throws)
- However, we can fallback to "pseudo-eager" by using a duck-typed ctx and routing the ctx.saved_tensors to lifted tensors
- Dynamo tries to inline external_utils.call_backward, treats BackwardCFunction as a AutogradFunctionContextVariable (only used up until we create the fake context: FakeBackwardCFunction)
- we call_function backward from the forward class AutogradFunctionVariable, and we still pass in the fake context as a UserDefinedObjectVariable (can later use AutogradFunctionContextVariable + HOO graph speculate)
Fixes#125489#124827
Pull Request resolved: https://github.com/pytorch/pytorch/pull/125661
Approved by: https://github.com/jansel
This allows `associative_scan` to take an arbitrary pytree of tensors,
which is flattened to their leaves before calling the `associative_scan`
higher order operator.
I also add support in inductor to generate code for scanning over sequences
of tensors.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/122137
Approved by: https://github.com/lezcano, https://github.com/Chillee
ghstack dependencies: #119430
While there are some similarities, they are also quite different (one
handles Numpy numbers while the other handles ints. I am also going to
add a wrap_symfloat soon which will do even more different behavior.
So split these out for clarity.
Signed-off-by: Edward Z. Yang <ezyang@meta.com>
Pull Request resolved: https://github.com/pytorch/pytorch/pull/125483
Approved by: https://github.com/lezcano
ghstack dependencies: #125395, #125419
A common complaint when working with data-dependent code in PyTorch is that it's hard to tell how far you are from the finish line: every time a GuardOnDataDependentSymNode error is hit, you have to somehow fix or workaround it to see the next one.
This PR adds a new mode `torch._functorch.config.fake_tensor_propagate_real_tensors` which modifies fake tensors to also propagate real tensors. This means that when we try to guard on a data-dependent SymNode, we can actually produce a real result. We also produce a warning which you should consult to figure out what the crux points are.
I ran this on vision_maskrcnn. In the baseline (without this mode), the model has 27 graph breaks, resulting in 40 graphs. With this mode on, the model has only 11 graph breaks, resulting in 15 graphs (the remaining graph breaks are due to missing functionality for item() on float tensor and some other Dynamo missing features.) You get a list of things that would have errored like this:
```
WARNING:torch.fx.experimental.symbolic_shapes:propagate_real_tensors evaluate_expr(Max(1, u1) < 2) -> True
WARNING:torch.fx.experimental.symbolic_shapes:propagate_real_tensors evaluate_expr(Eq(Max(1, u1), 1)) -> True
WARNING:torch.fx.experimental.symbolic_shapes:propagate_real_tensors evaluate_expr(Eq(Max(1, u1), 1)) -> True
WARNING:torch.fx.experimental.symbolic_shapes:propagate_real_tensors evaluate_expr(Ne(Max(1, u1), 1)) -> False
WARNING:torch.fx.experimental.symbolic_shapes:propagate_real_tensors evaluate_expr(Max(1, u0) < 2) -> True
WARNING:torch.fx.experimental.symbolic_shapes:propagate_real_tensors evaluate_expr(Eq(Max(1, u0), 1)) -> True
WARNING:torch.fx.experimental.symbolic_shapes:propagate_real_tensors evaluate_expr(Eq(Max(1, u0), 1)) -> True
WARNING:torch.fx.experimental.symbolic_shapes:propagate_real_tensors evaluate_expr(Ne(Max(1, u0), 1)) -> False
WARNING:torch.fx.experimental.symbolic_shapes:propagate_real_tensors evaluate_expr(Max(1, u1) < 2) -> True
WARNING:torch.fx.experimental.symbolic_shapes:propagate_real_tensors evaluate_expr(Eq(Max(1, u1), 1)) -> True
WARNING:torch.fx.experimental.symbolic_shapes:propagate_real_tensors evaluate_expr(Eq(Max(1, u1), 1)) -> True
WARNING:torch.fx.experimental.symbolic_shapes:propagate_real_tensors evaluate_expr(Ne(Max(1, u1), 1)) -> False
WARNING:torch.fx.experimental.symbolic_shapes:propagate_real_tensors evaluate_expr(Max(1, u0) < 2) -> True
WARNING:torch.fx.experimental.symbolic_shapes:propagate_real_tensors evaluate_expr(Eq(Max(1, u0), 1)) -> True
WARNING:torch.fx.experimental.symbolic_shapes:propagate_real_tensors evaluate_expr(Eq(Max(1, u0), 1)) -> True
WARNING:torch.fx.experimental.symbolic_shapes:propagate_real_tensors evaluate_expr(Ne(Max(1, u0), 1)) -> False
WARNING:torch.fx.experimental.symbolic_shapes:propagate_real_tensors evaluate_expr(Max(1, u1) < 2) -> False
WARNING:torch.fx.experimental.symbolic_shapes:propagate_real_tensors evaluate_expr(Eq(Max(1, u1), 1)) -> False
WARNING:torch.fx.experimental.symbolic_shapes:propagate_real_tensors evaluate_expr(Ne(Max(1, u1), 1)) -> True
WARNING:torch.fx.experimental.symbolic_shapes:propagate_real_tensors evaluate_expr(Max(1, u0) < 2) -> False
WARNING:torch.fx.experimental.symbolic_shapes:propagate_real_tensors evaluate_expr(Eq(Max(1, u0), 1)) -> False
```
Potential later follow ups:
* Improve the warning messages (in particular, should provide user frames)
* GC real tensors when they are no longer needed by tracing. Right now, this will use A LOT of memory, equal to as if your GC was broken and every intermediate tensor was kept live
Signed-off-by: Edward Z. Yang <ezyang@meta.com>
Pull Request resolved: https://github.com/pytorch/pytorch/pull/125115
Approved by: https://github.com/IvanKobzarev
To fix data-dependent errors we want to recommend that people use `torch._check*` APIs. The `constrain_as*` APIs should be fully subsumed by them, and in the future we should kill them entirely.
Differential Revision: D56774333
Pull Request resolved: https://github.com/pytorch/pytorch/pull/125253
Approved by: https://github.com/ezyang
Fix https://github.com/pytorch/pytorch/issues/124900.
When we reconstruct `ContextWrappingVariables`s, we only reconstruct the context class, not the object. Normally, contexts are active (via `with ctx:`) and we initialize the context object in the resume function. But for the case of inactive contexts (contexts declared ahead of time before the `with` block), we do not reconstruct them properly in the optimized bytecode or resume function. So this PR adds initialization for inactive contexts in the resume function.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/125203
Approved by: https://github.com/jansel
# Summary
This is part one of adding backwards support to FlexAttention.
This PR focuses on the eager implementation and wiring up enough of the templated_attention_backward(name change soon 😉) to get through aot_eager.
Notably this does not actually wire up the triton template just yet in order to make this PR easier to review. That will be the next follow up PR.
#### Structure
We pass both the forward and backward graph to the backwardsHOP since these are both needed to be inlined into the calculation for backwards:
- the forward graph is needed in order to re-compute the scores
- the joint graph is needed in order to construct the correct gradients post softmax_grad calc
### Attatched AOT Graph
https://gist.github.com/drisspg/ce4c041f8df8a5a7983c5174705cf2b5
Pull Request resolved: https://github.com/pytorch/pytorch/pull/123902
Approved by: https://github.com/Chillee
This PR is to add support for tensor's is_complex method in dynamo. Take the following code as an example:
```python
def test_tensor_is_complex(x):
if x.is_complex():
return x + 1
else:
return x - 1
```
Before this fix, the is_complex() call will cause a graph break "torch.* op returned non-Tensor bool call_method is_complex". After this fix, the graph break can be avoided.
Fixes#122692
Pull Request resolved: https://github.com/pytorch/pytorch/pull/124927
Approved by: https://github.com/ezyang
We guard on key order
1) When a key is a non-constant object
2) When we actually need key order - like .values, .items etc
For dicts/OrderedDicts that do not require key order guarding, we just rely on usual `GuardManger + DictGetItemGuardAccessor`. This is faster than going through the `list(d.keys())` based design for OrderedDicts.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/124779
Approved by: https://github.com/jansel
