참고
Click here to download the full example code
torch.export Tutorial¶
Author: William Wen, Zhengxu Chen, Angela Yi
경고
torch.export and its related features are in prototype status and are subject to backwards compatibility
breaking changes. This tutorial provides a snapshot of torch.export usage as of PyTorch 2.3.
torch.export() is the PyTorch 2.X way to export PyTorch models into
standardized model representations, intended
to be run on different (i.e. Python-less) environments. The official
documentation can be found here.
In this tutorial, you will learn how to use torch.export() to extract
ExportedProgram’s (i.e. single-graph representations) from PyTorch programs.
We also detail some considerations/modifications that you may need
to make in order to make your model compatible with torch.export.
Contents
Basic Usage¶
torch.export extracts single-graph representations from PyTorch programs
by tracing the target function, given example inputs.
torch.export.export() is the main entry point for torch.export.
In this tutorial, torch.export and torch.export.export() are practically synonymous,
though torch.export generally refers to the PyTorch 2.X export process, and torch.export.export()
generally refers to the actual function call.
The signature of torch.export.export() is:
export(
f: Callable,
args: Tuple[Any, ...],
kwargs: Optional[Dict[str, Any]] = None,
*,
dynamic_shapes: Optional[Dict[str, Dict[int, Dim]]] = None
) -> ExportedProgram
torch.export.export() traces the tensor computation graph from calling f(*args, **kwargs)
and wraps it in an ExportedProgram, which can be serialized or executed later with
different inputs. Note that while the output ExportedGraph is callable and can be
called in the same way as the original input callable, it is not a torch.nn.Module.
We will detail the dynamic_shapes argument later in the tutorial.
import torch
from torch.export import export
class MyModule(torch.nn.Module):
def __init__(self):
super().__init__()
self.lin = torch.nn.Linear(100, 10)
def forward(self, x, y):
return torch.nn.functional.relu(self.lin(x + y), inplace=True)
mod = MyModule()
exported_mod = export(mod, (torch.randn(8, 100), torch.randn(8, 100)))
print(type(exported_mod))
print(exported_mod.module()(torch.randn(8, 100), torch.randn(8, 100)))
<class 'torch.export.exported_program.ExportedProgram'>
tensor([[0.8632, 0.8407, 0.0407, 0.0000, 0.4132, 0.0000, 0.0000, 0.1538, 0.6111,
0.0000],
[0.0000, 0.0000, 0.0273, 0.8057, 0.0000, 1.0162, 0.8042, 0.0000, 0.2660,
0.0000],
[0.9481, 0.1396, 1.0225, 0.9563, 0.5832, 0.2546, 0.4095, 0.4591, 0.0000,
2.0053],
[1.1300, 0.4873, 0.0000, 0.9663, 1.2275, 1.4015, 0.0000, 0.9444, 0.0000,
0.0000],
[0.0000, 0.8724, 1.1648, 0.6867, 0.0000, 0.2833, 0.3202, 0.5848, 0.0000,
0.0833],
[1.1311, 0.1324, 0.0000, 1.7842, 0.0000, 0.3474, 0.9916, 0.3571, 0.0000,
0.0000],
[1.4348, 1.0570, 0.1771, 0.0000, 0.9510, 0.0000, 0.0000, 0.0000, 0.2618,
0.0000],
[0.8853, 0.0000, 0.0000, 0.4486, 0.0000, 0.0000, 0.5841, 0.7604, 0.0000,
0.0000]], grad_fn=<ReluBackward0>)
Let’s review some attributes of ExportedProgram that are of interest.
The graph attribute is an FX graph
traced from the function we exported, that is, the computation graph of all PyTorch operations.
The FX graph has some important properties:
The operations are 《ATen-level》 operations.
The graph is 《functionalized》, meaning that no operations are mutations.
The graph_module attribute is the GraphModule that wraps the graph attribute
so that it can be ran as a torch.nn.Module.
print(exported_mod)
print(exported_mod.graph_module)
ExportedProgram:
class GraphModule(torch.nn.Module):
def forward(self, arg0_1: "f32[10, 100]", arg1_1: "f32[10]", arg2_1: "f32[8, 100]", arg3_1: "f32[8, 100]"):
# File: /workspace/tutorials-kr/intermediate_source/torch_export_tutorial.py:70 in forward, code: return torch.nn.functional.relu(self.lin(x + y), inplace=True)
add: "f32[8, 100]" = torch.ops.aten.add.Tensor(arg2_1, arg3_1); arg2_1 = arg3_1 = None
t: "f32[100, 10]" = torch.ops.aten.t.default(arg0_1); arg0_1 = None
addmm: "f32[8, 10]" = torch.ops.aten.addmm.default(arg1_1, add, t); arg1_1 = add = t = None
relu: "f32[8, 10]" = torch.ops.aten.relu.default(addmm); addmm = None
return (relu,)
Graph signature: ExportGraphSignature(input_specs=[InputSpec(kind=<InputKind.PARAMETER: 2>, arg=TensorArgument(name='arg0_1'), target='lin.weight', persistent=None), InputSpec(kind=<InputKind.PARAMETER: 2>, arg=TensorArgument(name='arg1_1'), target='lin.bias', persistent=None), InputSpec(kind=<InputKind.USER_INPUT: 1>, arg=TensorArgument(name='arg2_1'), target=None, persistent=None), InputSpec(kind=<InputKind.USER_INPUT: 1>, arg=TensorArgument(name='arg3_1'), target=None, persistent=None)], output_specs=[OutputSpec(kind=<OutputKind.USER_OUTPUT: 1>, arg=TensorArgument(name='relu'), target=None)])
Range constraints: {}
GraphModule()
def forward(self, arg0_1, arg1_1, arg2_1, arg3_1):
add = torch.ops.aten.add.Tensor(arg2_1, arg3_1); arg2_1 = arg3_1 = None
t = torch.ops.aten.t.default(arg0_1); arg0_1 = None
addmm = torch.ops.aten.addmm.default(arg1_1, add, t); arg1_1 = add = t = None
relu = torch.ops.aten.relu.default(addmm); addmm = None
return (relu,)
# To see more debug info, please use `graph_module.print_readable()`
The printed code shows that FX graph only contains ATen-level ops (such as torch.ops.aten)
and that mutations were removed. For example, the mutating op torch.nn.functional.relu(..., inplace=True)
is represented in the printed code by torch.ops.aten.relu.default, which does not mutate.
Future uses of input to the original mutating relu op are replaced by the additional new output
of the replacement non-mutating relu op.
Other attributes of interest in ExportedProgram include:
graph_signature– the inputs, outputs, parameters, buffers, etc. of the exported graph.range_constraints– constraints, covered later
print(exported_mod.graph_signature)
ExportGraphSignature(input_specs=[InputSpec(kind=<InputKind.PARAMETER: 2>, arg=TensorArgument(name='arg0_1'), target='lin.weight', persistent=None), InputSpec(kind=<InputKind.PARAMETER: 2>, arg=TensorArgument(name='arg1_1'), target='lin.bias', persistent=None), InputSpec(kind=<InputKind.USER_INPUT: 1>, arg=TensorArgument(name='arg2_1'), target=None, persistent=None), InputSpec(kind=<InputKind.USER_INPUT: 1>, arg=TensorArgument(name='arg3_1'), target=None, persistent=None)], output_specs=[OutputSpec(kind=<OutputKind.USER_OUTPUT: 1>, arg=TensorArgument(name='relu'), target=None)])
See the torch.export documentation
for more details.
Graph Breaks¶
Although torch.export shares components with torch.compile,
the key limitation of torch.export, especially when compared to
torch.compile, is that it does not support graph breaks. This is because
handling graph breaks involves interpreting the unsupported operation with
default Python evaluation, which is incompatible with the export use case.
Therefore, in order to make your model code compatible with torch.export,
you will need to modify your code to remove graph breaks.
A graph break is necessary in cases such as:
data-dependent control flow
class Bad1(torch.nn.Module):
def forward(self, x):
if x.sum() > 0:
return torch.sin(x)
return torch.cos(x)
import traceback as tb
try:
export(Bad1(), (torch.randn(3, 3),))
except Exception:
tb.print_exc()
Traceback (most recent call last):
File "/workspace/tutorials-kr/intermediate_source/torch_export_tutorial.py", line 136, in <module>
export(Bad1(), (torch.randn(3, 3),))
File "/usr/local/lib/python3.10/dist-packages/torch/export/__init__.py", line 174, in export
return _export(
File "/usr/local/lib/python3.10/dist-packages/torch/export/_trace.py", line 635, in wrapper
raise e
File "/usr/local/lib/python3.10/dist-packages/torch/export/_trace.py", line 618, in wrapper
ep = fn(*args, **kwargs)
File "/usr/local/lib/python3.10/dist-packages/torch/export/exported_program.py", line 83, in wrapper
return fn(*args, **kwargs)
File "/usr/local/lib/python3.10/dist-packages/torch/export/_trace.py", line 860, in _export
gm_torch_level = _export_to_torch_ir(
File "/usr/local/lib/python3.10/dist-packages/torch/export/_trace.py", line 347, in _export_to_torch_ir
gm_torch_level, _ = torch._dynamo.export(
File "/usr/local/lib/python3.10/dist-packages/torch/_dynamo/eval_frame.py", line 1311, in inner
result_traced = opt_f(*args, **kwargs)
File "/usr/local/lib/python3.10/dist-packages/torch/nn/modules/module.py", line 1532, in _wrapped_call_impl
return self._call_impl(*args, **kwargs)
File "/usr/local/lib/python3.10/dist-packages/torch/nn/modules/module.py", line 1541, in _call_impl
return forward_call(*args, **kwargs)
File "/usr/local/lib/python3.10/dist-packages/torch/_dynamo/eval_frame.py", line 451, in _fn
return fn(*args, **kwargs)
File "/usr/local/lib/python3.10/dist-packages/torch/nn/modules/module.py", line 1532, in _wrapped_call_impl
return self._call_impl(*args, **kwargs)
File "/usr/local/lib/python3.10/dist-packages/torch/nn/modules/module.py", line 1541, in _call_impl
return forward_call(*args, **kwargs)
File "/usr/local/lib/python3.10/dist-packages/torch/_dynamo/convert_frame.py", line 921, in catch_errors
return callback(frame, cache_entry, hooks, frame_state, skip=1)
File "/usr/local/lib/python3.10/dist-packages/torch/_dynamo/convert_frame.py", line 400, in _convert_frame_assert
return _compile(
File "/usr/lib/python3.10/contextlib.py", line 79, in inner
return func(*args, **kwds)
File "/usr/local/lib/python3.10/dist-packages/torch/_dynamo/convert_frame.py", line 676, in _compile
guarded_code = compile_inner(code, one_graph, hooks, transform)
File "/usr/local/lib/python3.10/dist-packages/torch/_dynamo/utils.py", line 262, in time_wrapper
r = func(*args, **kwargs)
File "/usr/local/lib/python3.10/dist-packages/torch/_dynamo/convert_frame.py", line 535, in compile_inner
out_code = transform_code_object(code, transform)
File "/usr/local/lib/python3.10/dist-packages/torch/_dynamo/bytecode_transformation.py", line 1036, in transform_code_object
transformations(instructions, code_options)
File "/usr/local/lib/python3.10/dist-packages/torch/_dynamo/convert_frame.py", line 165, in _fn
return fn(*args, **kwargs)
File "/usr/local/lib/python3.10/dist-packages/torch/_dynamo/convert_frame.py", line 500, in transform
tracer.run()
File "/usr/local/lib/python3.10/dist-packages/torch/_dynamo/symbolic_convert.py", line 2149, in run
super().run()
File "/usr/local/lib/python3.10/dist-packages/torch/_dynamo/symbolic_convert.py", line 810, in run
and self.step()
File "/usr/local/lib/python3.10/dist-packages/torch/_dynamo/symbolic_convert.py", line 773, in step
getattr(self, inst.opname)(inst)
File "/usr/local/lib/python3.10/dist-packages/torch/_dynamo/symbolic_convert.py", line 464, in inner
raise exc.UserError(
torch._dynamo.exc.UserError: Dynamic control flow is not supported at the moment. Please use functorch.experimental.control_flow.cond to explicitly capture the control flow. For more information about this error, see: https://pytorch.org/docs/main/generated/exportdb/index.html#cond-operands
from user code:
File "/workspace/tutorials-kr/intermediate_source/torch_export_tutorial.py", line 130, in forward
if x.sum() > 0:
Set TORCH_LOGS="+dynamo" and TORCHDYNAMO_VERBOSE=1 for more information
accessing tensor data with
.data
class Bad2(torch.nn.Module):
def forward(self, x):
x.data[0, 0] = 3
return x
try:
export(Bad2(), (torch.randn(3, 3),))
except Exception:
tb.print_exc()
calling unsupported functions (such as many built-in functions)
class Bad3(torch.nn.Module):
def forward(self, x):
x = x + 1
return x + id(x)
try:
export(Bad3(), (torch.randn(3, 3),))
except Exception:
tb.print_exc()
Traceback (most recent call last):
File "/workspace/tutorials-kr/intermediate_source/torch_export_tutorial.py", line 162, in <module>
export(Bad3(), (torch.randn(3, 3),))
File "/usr/local/lib/python3.10/dist-packages/torch/export/__init__.py", line 174, in export
return _export(
File "/usr/local/lib/python3.10/dist-packages/torch/export/_trace.py", line 635, in wrapper
raise e
File "/usr/local/lib/python3.10/dist-packages/torch/export/_trace.py", line 618, in wrapper
ep = fn(*args, **kwargs)
File "/usr/local/lib/python3.10/dist-packages/torch/export/exported_program.py", line 83, in wrapper
return fn(*args, **kwargs)
File "/usr/local/lib/python3.10/dist-packages/torch/export/_trace.py", line 860, in _export
gm_torch_level = _export_to_torch_ir(
File "/usr/local/lib/python3.10/dist-packages/torch/export/_trace.py", line 347, in _export_to_torch_ir
gm_torch_level, _ = torch._dynamo.export(
File "/usr/local/lib/python3.10/dist-packages/torch/_dynamo/eval_frame.py", line 1311, in inner
result_traced = opt_f(*args, **kwargs)
File "/usr/local/lib/python3.10/dist-packages/torch/nn/modules/module.py", line 1532, in _wrapped_call_impl
return self._call_impl(*args, **kwargs)
File "/usr/local/lib/python3.10/dist-packages/torch/nn/modules/module.py", line 1541, in _call_impl
return forward_call(*args, **kwargs)
File "/usr/local/lib/python3.10/dist-packages/torch/_dynamo/eval_frame.py", line 451, in _fn
return fn(*args, **kwargs)
File "/usr/local/lib/python3.10/dist-packages/torch/nn/modules/module.py", line 1532, in _wrapped_call_impl
return self._call_impl(*args, **kwargs)
File "/usr/local/lib/python3.10/dist-packages/torch/nn/modules/module.py", line 1541, in _call_impl
return forward_call(*args, **kwargs)
File "/usr/local/lib/python3.10/dist-packages/torch/_dynamo/convert_frame.py", line 921, in catch_errors
return callback(frame, cache_entry, hooks, frame_state, skip=1)
File "/usr/local/lib/python3.10/dist-packages/torch/_dynamo/convert_frame.py", line 400, in _convert_frame_assert
return _compile(
File "/usr/lib/python3.10/contextlib.py", line 79, in inner
return func(*args, **kwds)
File "/usr/local/lib/python3.10/dist-packages/torch/_dynamo/convert_frame.py", line 676, in _compile
guarded_code = compile_inner(code, one_graph, hooks, transform)
File "/usr/local/lib/python3.10/dist-packages/torch/_dynamo/utils.py", line 262, in time_wrapper
r = func(*args, **kwargs)
File "/usr/local/lib/python3.10/dist-packages/torch/_dynamo/convert_frame.py", line 535, in compile_inner
out_code = transform_code_object(code, transform)
File "/usr/local/lib/python3.10/dist-packages/torch/_dynamo/bytecode_transformation.py", line 1036, in transform_code_object
transformations(instructions, code_options)
File "/usr/local/lib/python3.10/dist-packages/torch/_dynamo/convert_frame.py", line 165, in _fn
return fn(*args, **kwargs)
File "/usr/local/lib/python3.10/dist-packages/torch/_dynamo/convert_frame.py", line 500, in transform
tracer.run()
File "/usr/local/lib/python3.10/dist-packages/torch/_dynamo/symbolic_convert.py", line 2149, in run
super().run()
File "/usr/local/lib/python3.10/dist-packages/torch/_dynamo/symbolic_convert.py", line 810, in run
and self.step()
File "/usr/local/lib/python3.10/dist-packages/torch/_dynamo/symbolic_convert.py", line 773, in step
getattr(self, inst.opname)(inst)
File "/usr/local/lib/python3.10/dist-packages/torch/_dynamo/symbolic_convert.py", line 489, in wrapper
return inner_fn(self, inst)
File "/usr/local/lib/python3.10/dist-packages/torch/_dynamo/symbolic_convert.py", line 1219, in CALL_FUNCTION
self.call_function(fn, args, {})
File "/usr/local/lib/python3.10/dist-packages/torch/_dynamo/symbolic_convert.py", line 674, in call_function
self.push(fn.call_function(self, args, kwargs))
File "/usr/local/lib/python3.10/dist-packages/torch/_dynamo/variables/builtin.py", line 687, in call_function
result = handler(tx, *args, **kwargs)
File "/usr/local/lib/python3.10/dist-packages/torch/_dynamo/variables/builtin.py", line 1520, in call_id
unimplemented(f"call_id with args {args}")
File "/usr/local/lib/python3.10/dist-packages/torch/_dynamo/exc.py", line 190, in unimplemented
raise Unsupported(msg)
torch._dynamo.exc.Unsupported: call_id with args (TensorVariable(),)
from user code:
File "/workspace/tutorials-kr/intermediate_source/torch_export_tutorial.py", line 159, in forward
return x + id(x)
Set TORCH_LOGS="+dynamo" and TORCHDYNAMO_VERBOSE=1 for more information
unsupported Python language features (e.g. throwing exceptions, match statements)
class Bad4(torch.nn.Module):
def forward(self, x):
try:
x = x + 1
raise RuntimeError("bad")
except:
x = x + 2
return x
try:
export(Bad4(), (torch.randn(3, 3),))
except Exception:
tb.print_exc()
Traceback (most recent call last):
File "/workspace/tutorials-kr/intermediate_source/torch_export_tutorial.py", line 179, in <module>
export(Bad4(), (torch.randn(3, 3),))
File "/usr/local/lib/python3.10/dist-packages/torch/export/__init__.py", line 174, in export
return _export(
File "/usr/local/lib/python3.10/dist-packages/torch/export/_trace.py", line 635, in wrapper
raise e
File "/usr/local/lib/python3.10/dist-packages/torch/export/_trace.py", line 618, in wrapper
ep = fn(*args, **kwargs)
File "/usr/local/lib/python3.10/dist-packages/torch/export/exported_program.py", line 83, in wrapper
return fn(*args, **kwargs)
File "/usr/local/lib/python3.10/dist-packages/torch/export/_trace.py", line 860, in _export
gm_torch_level = _export_to_torch_ir(
File "/usr/local/lib/python3.10/dist-packages/torch/export/_trace.py", line 347, in _export_to_torch_ir
gm_torch_level, _ = torch._dynamo.export(
File "/usr/local/lib/python3.10/dist-packages/torch/_dynamo/eval_frame.py", line 1311, in inner
result_traced = opt_f(*args, **kwargs)
File "/usr/local/lib/python3.10/dist-packages/torch/nn/modules/module.py", line 1532, in _wrapped_call_impl
return self._call_impl(*args, **kwargs)
File "/usr/local/lib/python3.10/dist-packages/torch/nn/modules/module.py", line 1541, in _call_impl
return forward_call(*args, **kwargs)
File "/usr/local/lib/python3.10/dist-packages/torch/_dynamo/eval_frame.py", line 451, in _fn
return fn(*args, **kwargs)
File "/usr/local/lib/python3.10/dist-packages/torch/nn/modules/module.py", line 1532, in _wrapped_call_impl
return self._call_impl(*args, **kwargs)
File "/usr/local/lib/python3.10/dist-packages/torch/nn/modules/module.py", line 1541, in _call_impl
return forward_call(*args, **kwargs)
File "/usr/local/lib/python3.10/dist-packages/torch/_dynamo/convert_frame.py", line 921, in catch_errors
return callback(frame, cache_entry, hooks, frame_state, skip=1)
File "/usr/local/lib/python3.10/dist-packages/torch/_dynamo/convert_frame.py", line 400, in _convert_frame_assert
return _compile(
File "/usr/lib/python3.10/contextlib.py", line 79, in inner
return func(*args, **kwds)
File "/usr/local/lib/python3.10/dist-packages/torch/_dynamo/convert_frame.py", line 676, in _compile
guarded_code = compile_inner(code, one_graph, hooks, transform)
File "/usr/local/lib/python3.10/dist-packages/torch/_dynamo/utils.py", line 262, in time_wrapper
r = func(*args, **kwargs)
File "/usr/local/lib/python3.10/dist-packages/torch/_dynamo/convert_frame.py", line 535, in compile_inner
out_code = transform_code_object(code, transform)
File "/usr/local/lib/python3.10/dist-packages/torch/_dynamo/bytecode_transformation.py", line 1036, in transform_code_object
transformations(instructions, code_options)
File "/usr/local/lib/python3.10/dist-packages/torch/_dynamo/convert_frame.py", line 165, in _fn
return fn(*args, **kwargs)
File "/usr/local/lib/python3.10/dist-packages/torch/_dynamo/convert_frame.py", line 500, in transform
tracer.run()
File "/usr/local/lib/python3.10/dist-packages/torch/_dynamo/symbolic_convert.py", line 2149, in run
super().run()
File "/usr/local/lib/python3.10/dist-packages/torch/_dynamo/symbolic_convert.py", line 810, in run
and self.step()
File "/usr/local/lib/python3.10/dist-packages/torch/_dynamo/symbolic_convert.py", line 773, in step
getattr(self, inst.opname)(inst)
File "/usr/local/lib/python3.10/dist-packages/torch/_dynamo/symbolic_convert.py", line 489, in wrapper
return inner_fn(self, inst)
File "/usr/local/lib/python3.10/dist-packages/torch/_dynamo/symbolic_convert.py", line 1219, in CALL_FUNCTION
self.call_function(fn, args, {})
File "/usr/local/lib/python3.10/dist-packages/torch/_dynamo/symbolic_convert.py", line 674, in call_function
self.push(fn.call_function(self, args, kwargs))
File "/usr/local/lib/python3.10/dist-packages/torch/_dynamo/variables/builtin.py", line 730, in call_function
return super().call_function(tx, args, kwargs)
File "/usr/local/lib/python3.10/dist-packages/torch/_dynamo/variables/base.py", line 349, in call_function
unimplemented(f"call_function {self} {args} {kwargs}")
File "/usr/local/lib/python3.10/dist-packages/torch/_dynamo/exc.py", line 190, in unimplemented
raise Unsupported(msg)
torch._dynamo.exc.Unsupported: call_function BuiltinVariable(RuntimeError) [ConstantVariable()] {}
from user code:
File "/workspace/tutorials-kr/intermediate_source/torch_export_tutorial.py", line 173, in forward
raise RuntimeError("bad")
Set TORCH_LOGS="+dynamo" and TORCHDYNAMO_VERBOSE=1 for more information
Non-Strict Export¶
To trace the program, torch.export uses TorchDynamo, a byte code analysis
engine, to symbolically analyze the Python code and build a graph based on the
results. This analysis allows torch.export to provide stronger guarantees
about safety, but not all Python code is supported, causing these graph
breaks.
To address this issue, in PyTorch 2.3, we introduced a new mode of
exporting called non-strict mode, where we trace through the program using the
Python interpreter executing it exactly as it would in eager mode, allowing us
to skip over unsupported Python features. This is done through adding a
strict=False flag.
Looking at some of the previous examples which resulted in graph breaks:
Accessing tensor data with
.datanow works correctly
class Bad2(torch.nn.Module):
def forward(self, x):
x.data[0, 0] = 3
return x
bad2_nonstrict = export(Bad2(), (torch.randn(3, 3),), strict=False)
print(bad2_nonstrict.module()(torch.ones(3, 3)))
tensor([[3., 1., 1.],
[1., 1., 1.],
[1., 1., 1.]])
Calling unsupported functions (such as many built-in functions) traces
through, but in this case, id(x) gets specialized as a constant integer in
the graph. This is because id(x) is not a tensor operation, so the
operation is not recorded in the graph.
class Bad3(torch.nn.Module):
def forward(self, x):
x = x + 1
return x + id(x)
bad3_nonstrict = export(Bad3(), (torch.randn(3, 3),), strict=False)
print(bad3_nonstrict)
print(bad3_nonstrict.module()(torch.ones(3, 3)))
ExportedProgram:
class GraphModule(torch.nn.Module):
def forward(self, arg0_1: "f32[3, 3]"):
# No stacktrace found for following nodes
add: "f32[3, 3]" = torch.ops.aten.add.Tensor(arg0_1, 1); arg0_1 = None
add_1: "f32[3, 3]" = torch.ops.aten.add.Tensor(add, 128803024495904); add = None
return (add_1,)
Graph signature: ExportGraphSignature(input_specs=[InputSpec(kind=<InputKind.USER_INPUT: 1>, arg=TensorArgument(name='arg0_1'), target=None, persistent=None)], output_specs=[OutputSpec(kind=<OutputKind.USER_OUTPUT: 1>, arg=TensorArgument(name='add_1'), target=None)])
Range constraints: {}
tensor([[1.2880e+14, 1.2880e+14, 1.2880e+14],
[1.2880e+14, 1.2880e+14, 1.2880e+14],
[1.2880e+14, 1.2880e+14, 1.2880e+14]])
Unsupported Python language features (such as throwing exceptions, match
statements) now also get traced through.
class Bad4(torch.nn.Module):
def forward(self, x):
try:
x = x + 1
raise RuntimeError("bad")
except:
x = x + 2
return x
bad4_nonstrict = export(Bad4(), (torch.randn(3, 3),), strict=False)
print(bad4_nonstrict.module()(torch.ones(3, 3)))
tensor([[4., 4., 4.],
[4., 4., 4.],
[4., 4., 4.]])
However, there are still some features that require rewrites to the original module:
Control Flow Ops¶
torch.export actually does support data-dependent control flow.
But these need to be expressed using control flow ops. For example,
we can fix the control flow example above using the cond op, like so:
from functorch.experimental.control_flow import cond
class Bad1Fixed(torch.nn.Module):
def forward(self, x):
def true_fn(x):
return torch.sin(x)
def false_fn(x):
return torch.cos(x)
return cond(x.sum() > 0, true_fn, false_fn, [x])
exported_bad1_fixed = export(Bad1Fixed(), (torch.randn(3, 3),))
print(exported_bad1_fixed.module()(torch.ones(3, 3)))
print(exported_bad1_fixed.module()(-torch.ones(3, 3)))
tensor([[0.8415, 0.8415, 0.8415],
[0.8415, 0.8415, 0.8415],
[0.8415, 0.8415, 0.8415]])
tensor([[0.5403, 0.5403, 0.5403],
[0.5403, 0.5403, 0.5403],
[0.5403, 0.5403, 0.5403]])
There are limitations to cond that one should be aware of:
The predicate (i.e.
x.sum() > 0) must result in a boolean or a single-element tensor.The operands (i.e.
[x]) must be tensors.The branch function (i.e.
true_fnandfalse_fn) signature must match with the operands and they must both return a single tensor with the same metadata (for example,dtype,shape, etc.).Branch functions cannot mutate input or global variables.
Branch functions cannot access closure variables, except for
selfif the function is defined in the scope of a method.
For more details about cond, check out the cond documentation.
Constraints/Dynamic Shapes¶
Ops can have different specializations/behaviors for different tensor shapes, so by default,
torch.export requires inputs to ExportedProgram to have the same shape as the respective
example inputs given to the initial torch.export.export() call.
If we try to run the ExportedProgram in the example below with a tensor
with a different shape, we get an error:
class MyModule2(torch.nn.Module):
def __init__(self):
super().__init__()
self.lin = torch.nn.Linear(100, 10)
def forward(self, x, y):
return torch.nn.functional.relu(self.lin(x + y), inplace=True)
mod2 = MyModule2()
exported_mod2 = export(mod2, (torch.randn(8, 100), torch.randn(8, 100)))
try:
exported_mod2.module()(torch.randn(10, 100), torch.randn(10, 100))
except Exception:
tb.print_exc()
Traceback (most recent call last):
File "/workspace/tutorials-kr/intermediate_source/torch_export_tutorial.py", line 325, in <module>
exported_mod2.module()(torch.randn(10, 100), torch.randn(10, 100))
File "/usr/local/lib/python3.10/dist-packages/torch/fx/graph_module.py", line 737, in call_wrapped
return self._wrapped_call(self, *args, **kwargs)
File "/usr/local/lib/python3.10/dist-packages/torch/fx/graph_module.py", line 317, in __call__
raise e
File "/usr/local/lib/python3.10/dist-packages/torch/fx/graph_module.py", line 304, in __call__
return super(self.cls, obj).__call__(*args, **kwargs) # type: ignore[misc]
File "/usr/local/lib/python3.10/dist-packages/torch/nn/modules/module.py", line 1532, in _wrapped_call_impl
return self._call_impl(*args, **kwargs)
File "/usr/local/lib/python3.10/dist-packages/torch/nn/modules/module.py", line 1561, in _call_impl
args_kwargs_result = hook(self, args, kwargs) # type: ignore[misc]
File "/usr/local/lib/python3.10/dist-packages/torch/_dynamo/eval_frame.py", line 451, in _fn
return fn(*args, **kwargs)
File "/usr/local/lib/python3.10/dist-packages/torch/_dynamo/external_utils.py", line 36, in inner
return fn(*args, **kwargs)
File "/usr/local/lib/python3.10/dist-packages/torch/export/_unlift.py", line 32, in _check_input_constraints_pre_hook
return _check_input_constraints_for_graph(
File "/usr/local/lib/python3.10/dist-packages/torch/_export/utils.py", line 129, in _check_input_constraints_for_graph
raise RuntimeError(
RuntimeError: Expected input at *args[0].shape[0] to be equal to 8, but got 10
We can relax this constraint using the dynamic_shapes argument of
torch.export.export(), which allows us to specify, using torch.export.Dim
(documentation),
which dimensions of the input tensors are dynamic.
For each tensor argument of the input callable, we can specify a mapping from the dimension
to a torch.export.Dim.
A torch.export.Dim is essentially a named symbolic integer with optional
minimum and maximum bounds.
Then, the format of torch.export.export()〉s dynamic_shapes argument is a mapping
from the input callable’s tensor argument names, to dimension –> dim mappings as described above.
If there is no torch.export.Dim given to a tensor argument’s dimension, then that dimension is
assumed to be static.
The first argument of torch.export.Dim is the name for the symbolic integer, used for debugging.
Then we can specify an optional minimum and maximum bound (inclusive). Below, we show a usage example.
In the example below, our input
inp1 has an unconstrained first dimension, but the size of the second
dimension must be in the interval [4, 18].
from torch.export import Dim
inp1 = torch.randn(10, 10, 2)
class DynamicShapesExample1(torch.nn.Module):
def forward(self, x):
x = x[:, 2:]
return torch.relu(x)
inp1_dim0 = Dim("inp1_dim0")
inp1_dim1 = Dim("inp1_dim1", min=4, max=18)
dynamic_shapes1 = {
"x": {0: inp1_dim0, 1: inp1_dim1},
}
exported_dynamic_shapes_example1 = export(DynamicShapesExample1(), (inp1,), dynamic_shapes=dynamic_shapes1)
print(exported_dynamic_shapes_example1.module()(torch.randn(5, 5, 2)))
try:
exported_dynamic_shapes_example1.module()(torch.randn(8, 1, 2))
except Exception:
tb.print_exc()
try:
exported_dynamic_shapes_example1.module()(torch.randn(8, 20, 2))
except Exception:
tb.print_exc()
try:
exported_dynamic_shapes_example1.module()(torch.randn(8, 8, 3))
except Exception:
tb.print_exc()
tensor([[[0.0000, 0.0000],
[0.8850, 1.2371],
[0.0000, 0.0000]],
[[0.0000, 0.0000],
[0.0000, 0.3487],
[0.2520, 1.2545]],
[[0.5863, 0.2831],
[0.0000, 0.4669],
[0.1059, 0.0000]],
[[0.7833, 0.0000],
[0.4480, 0.0523],
[0.0000, 0.0000]],
[[0.9306, 0.0000],
[0.0000, 0.7895],
[0.1160, 0.0000]]])
Traceback (most recent call last):
File "/workspace/tutorials-kr/intermediate_source/torch_export_tutorial.py", line 372, in <module>
exported_dynamic_shapes_example1.module()(torch.randn(8, 1, 2))
File "/usr/local/lib/python3.10/dist-packages/torch/fx/graph_module.py", line 737, in call_wrapped
return self._wrapped_call(self, *args, **kwargs)
File "/usr/local/lib/python3.10/dist-packages/torch/fx/graph_module.py", line 317, in __call__
raise e
File "/usr/local/lib/python3.10/dist-packages/torch/fx/graph_module.py", line 304, in __call__
return super(self.cls, obj).__call__(*args, **kwargs) # type: ignore[misc]
File "/usr/local/lib/python3.10/dist-packages/torch/nn/modules/module.py", line 1532, in _wrapped_call_impl
return self._call_impl(*args, **kwargs)
File "/usr/local/lib/python3.10/dist-packages/torch/nn/modules/module.py", line 1561, in _call_impl
args_kwargs_result = hook(self, args, kwargs) # type: ignore[misc]
File "/usr/local/lib/python3.10/dist-packages/torch/_dynamo/eval_frame.py", line 451, in _fn
return fn(*args, **kwargs)
File "/usr/local/lib/python3.10/dist-packages/torch/_dynamo/external_utils.py", line 36, in inner
return fn(*args, **kwargs)
File "/usr/local/lib/python3.10/dist-packages/torch/export/_unlift.py", line 32, in _check_input_constraints_pre_hook
return _check_input_constraints_for_graph(
File "/usr/local/lib/python3.10/dist-packages/torch/_export/utils.py", line 117, in _check_input_constraints_for_graph
raise RuntimeError(
RuntimeError: Expected input at *args[0].shape[1] to be >= 4, but got 1
Traceback (most recent call last):
File "/workspace/tutorials-kr/intermediate_source/torch_export_tutorial.py", line 377, in <module>
exported_dynamic_shapes_example1.module()(torch.randn(8, 20, 2))
File "/usr/local/lib/python3.10/dist-packages/torch/fx/graph_module.py", line 737, in call_wrapped
return self._wrapped_call(self, *args, **kwargs)
File "/usr/local/lib/python3.10/dist-packages/torch/fx/graph_module.py", line 317, in __call__
raise e
File "/usr/local/lib/python3.10/dist-packages/torch/fx/graph_module.py", line 304, in __call__
return super(self.cls, obj).__call__(*args, **kwargs) # type: ignore[misc]
File "/usr/local/lib/python3.10/dist-packages/torch/nn/modules/module.py", line 1532, in _wrapped_call_impl
return self._call_impl(*args, **kwargs)
File "/usr/local/lib/python3.10/dist-packages/torch/nn/modules/module.py", line 1561, in _call_impl
args_kwargs_result = hook(self, args, kwargs) # type: ignore[misc]
File "/usr/local/lib/python3.10/dist-packages/torch/_dynamo/eval_frame.py", line 451, in _fn
return fn(*args, **kwargs)
File "/usr/local/lib/python3.10/dist-packages/torch/_dynamo/external_utils.py", line 36, in inner
return fn(*args, **kwargs)
File "/usr/local/lib/python3.10/dist-packages/torch/export/_unlift.py", line 32, in _check_input_constraints_pre_hook
return _check_input_constraints_for_graph(
File "/usr/local/lib/python3.10/dist-packages/torch/_export/utils.py", line 123, in _check_input_constraints_for_graph
raise RuntimeError(
RuntimeError: Expected input at *args[0].shape[1] to be <= 18, but got 20
Traceback (most recent call last):
File "/workspace/tutorials-kr/intermediate_source/torch_export_tutorial.py", line 382, in <module>
exported_dynamic_shapes_example1.module()(torch.randn(8, 8, 3))
File "/usr/local/lib/python3.10/dist-packages/torch/fx/graph_module.py", line 737, in call_wrapped
return self._wrapped_call(self, *args, **kwargs)
File "/usr/local/lib/python3.10/dist-packages/torch/fx/graph_module.py", line 317, in __call__
raise e
File "/usr/local/lib/python3.10/dist-packages/torch/fx/graph_module.py", line 304, in __call__
return super(self.cls, obj).__call__(*args, **kwargs) # type: ignore[misc]
File "/usr/local/lib/python3.10/dist-packages/torch/nn/modules/module.py", line 1532, in _wrapped_call_impl
return self._call_impl(*args, **kwargs)
File "/usr/local/lib/python3.10/dist-packages/torch/nn/modules/module.py", line 1561, in _call_impl
args_kwargs_result = hook(self, args, kwargs) # type: ignore[misc]
File "/usr/local/lib/python3.10/dist-packages/torch/_dynamo/eval_frame.py", line 451, in _fn
return fn(*args, **kwargs)
File "/usr/local/lib/python3.10/dist-packages/torch/_dynamo/external_utils.py", line 36, in inner
return fn(*args, **kwargs)
File "/usr/local/lib/python3.10/dist-packages/torch/export/_unlift.py", line 32, in _check_input_constraints_pre_hook
return _check_input_constraints_for_graph(
File "/usr/local/lib/python3.10/dist-packages/torch/_export/utils.py", line 129, in _check_input_constraints_for_graph
raise RuntimeError(
RuntimeError: Expected input at *args[0].shape[2] to be equal to 2, but got 3
Note that if our example inputs to torch.export do not satisfy the constraints
given by dynamic_shapes, then we get an error.
inp1_dim1_bad = Dim("inp1_dim1_bad", min=11, max=18)
dynamic_shapes1_bad = {
"x": {0: inp1_dim0, 1: inp1_dim1_bad},
}
try:
export(DynamicShapesExample1(), (inp1,), dynamic_shapes=dynamic_shapes1_bad)
except Exception:
tb.print_exc()
Traceback (most recent call last):
File "/usr/local/lib/python3.10/dist-packages/torch/export/_trace.py", line 347, in _export_to_torch_ir
gm_torch_level, _ = torch._dynamo.export(
File "/usr/local/lib/python3.10/dist-packages/torch/_dynamo/eval_frame.py", line 1354, in inner
raise constraint_violation_error
File "/usr/local/lib/python3.10/dist-packages/torch/_dynamo/eval_frame.py", line 1311, in inner
result_traced = opt_f(*args, **kwargs)
File "/usr/local/lib/python3.10/dist-packages/torch/nn/modules/module.py", line 1532, in _wrapped_call_impl
return self._call_impl(*args, **kwargs)
File "/usr/local/lib/python3.10/dist-packages/torch/nn/modules/module.py", line 1541, in _call_impl
return forward_call(*args, **kwargs)
File "/usr/local/lib/python3.10/dist-packages/torch/_dynamo/eval_frame.py", line 451, in _fn
return fn(*args, **kwargs)
File "/usr/local/lib/python3.10/dist-packages/torch/nn/modules/module.py", line 1532, in _wrapped_call_impl
return self._call_impl(*args, **kwargs)
File "/usr/local/lib/python3.10/dist-packages/torch/nn/modules/module.py", line 1541, in _call_impl
return forward_call(*args, **kwargs)
File "/usr/local/lib/python3.10/dist-packages/torch/_dynamo/convert_frame.py", line 921, in catch_errors
return callback(frame, cache_entry, hooks, frame_state, skip=1)
File "/usr/local/lib/python3.10/dist-packages/torch/_dynamo/convert_frame.py", line 400, in _convert_frame_assert
return _compile(
File "/usr/lib/python3.10/contextlib.py", line 79, in inner
return func(*args, **kwds)
File "/usr/local/lib/python3.10/dist-packages/torch/_dynamo/convert_frame.py", line 676, in _compile
guarded_code = compile_inner(code, one_graph, hooks, transform)
File "/usr/local/lib/python3.10/dist-packages/torch/_dynamo/utils.py", line 262, in time_wrapper
r = func(*args, **kwargs)
File "/usr/local/lib/python3.10/dist-packages/torch/_dynamo/convert_frame.py", line 535, in compile_inner
out_code = transform_code_object(code, transform)
File "/usr/local/lib/python3.10/dist-packages/torch/_dynamo/bytecode_transformation.py", line 1036, in transform_code_object
transformations(instructions, code_options)
File "/usr/local/lib/python3.10/dist-packages/torch/_dynamo/convert_frame.py", line 165, in _fn
return fn(*args, **kwargs)
File "/usr/local/lib/python3.10/dist-packages/torch/_dynamo/convert_frame.py", line 482, in transform
tracer = InstructionTranslator(
File "/usr/local/lib/python3.10/dist-packages/torch/_dynamo/symbolic_convert.py", line 2115, in __init__
self.symbolic_locals = VariableTracker.apply(
File "/usr/local/lib/python3.10/dist-packages/torch/_dynamo/variables/base.py", line 217, in apply
result = {
File "/usr/local/lib/python3.10/dist-packages/torch/_dynamo/variables/base.py", line 218, in <dictcomp>
k: cls.apply(fn, v, cache, skip_fn) for k, v in list(value.items())
File "/usr/local/lib/python3.10/dist-packages/torch/_dynamo/variables/base.py", line 203, in apply
result = fn(update_object_dict(value))
File "/usr/local/lib/python3.10/dist-packages/torch/_dynamo/symbolic_convert.py", line 2116, in <lambda>
lambda x: x.realize(), self.symbolic_locals
File "/usr/local/lib/python3.10/dist-packages/torch/_dynamo/variables/lazy.py", line 58, in realize
self._cache.realize(self.parents_tracker)
File "/usr/local/lib/python3.10/dist-packages/torch/_dynamo/variables/lazy.py", line 24, in realize
self.vt = VariableBuilder(tx, self.source)(self.value)
File "/usr/local/lib/python3.10/dist-packages/torch/_dynamo/variables/builder.py", line 269, in __call__
vt = self._wrap(value)
File "/usr/local/lib/python3.10/dist-packages/torch/_dynamo/variables/builder.py", line 402, in _wrap
return type_dispatch(self, value)
File "/usr/local/lib/python3.10/dist-packages/torch/_dynamo/variables/builder.py", line 1073, in wrap_tensor
tensor_variable = wrap_fx_proxy(
File "/usr/local/lib/python3.10/dist-packages/torch/_dynamo/variables/builder.py", line 1330, in wrap_fx_proxy
return wrap_fx_proxy_cls(target_cls=TensorVariable, **kwargs)
File "/usr/local/lib/python3.10/dist-packages/torch/_dynamo/variables/builder.py", line 1440, in wrap_fx_proxy_cls
example_value = wrap_to_fake_tensor_and_record(
File "/usr/local/lib/python3.10/dist-packages/torch/_dynamo/variables/builder.py", line 1880, in wrap_to_fake_tensor_and_record
fake_e = wrap_fake_exception(
File "/usr/local/lib/python3.10/dist-packages/torch/_dynamo/utils.py", line 1190, in wrap_fake_exception
return fn()
File "/usr/local/lib/python3.10/dist-packages/torch/_dynamo/variables/builder.py", line 1881, in <lambda>
lambda: tx.fake_mode.from_tensor(
File "/usr/local/lib/python3.10/dist-packages/torch/_subclasses/fake_tensor.py", line 1666, in from_tensor
return self.fake_tensor_converter(
File "/usr/local/lib/python3.10/dist-packages/torch/_subclasses/fake_tensor.py", line 349, in __call__
return self.from_real_tensor(
File "/usr/local/lib/python3.10/dist-packages/torch/_subclasses/fake_tensor.py", line 306, in from_real_tensor
out = self.meta_converter(
File "/usr/local/lib/python3.10/dist-packages/torch/_subclasses/meta_utils.py", line 967, in __call__
r = self.meta_tensor(
File "/usr/local/lib/python3.10/dist-packages/torch/_subclasses/meta_utils.py", line 782, in meta_tensor
) = sym_sizes_strides_storage_offset(t, source, symbolic_context)
File "/usr/local/lib/python3.10/dist-packages/torch/_subclasses/meta_utils.py", line 269, in sym_sizes_strides_storage_offset
return shape_env.create_symbolic_sizes_strides_storage_offset(
File "/usr/local/lib/python3.10/dist-packages/torch/fx/experimental/symbolic_shapes.py", line 2339, in create_symbolic_sizes_strides_storage_offset
return self._create_symbolic_sizes_strides_storage_offset(
File "/usr/local/lib/python3.10/dist-packages/torch/fx/experimental/recording.py", line 231, in wrapper
return fn(*args, **kwargs)
File "/usr/local/lib/python3.10/dist-packages/torch/fx/experimental/symbolic_shapes.py", line 2395, in _create_symbolic_sizes_strides_storage_offset
size: List[sympy.Expr] = self._produce_dyn_sizes_from_int_tuple(ex_size, source, symbolic_context)
File "/usr/local/lib/python3.10/dist-packages/torch/fx/experimental/symbolic_shapes.py", line 2271, in _produce_dyn_sizes_from_int_tuple
size.append(self.create_symbol(
File "/usr/local/lib/python3.10/dist-packages/torch/fx/experimental/recording.py", line 231, in wrapper
return fn(*args, **kwargs)
File "/usr/local/lib/python3.10/dist-packages/torch/fx/experimental/symbolic_shapes.py", line 2709, in create_symbol
raise ConstraintViolationError(f"{val} not in range [{vr.lower}, {vr.upper}]")
torch.fx.experimental.symbolic_shapes.ConstraintViolationError: 10 not in range [11, 18]
Set TORCH_LOGS="+dynamo" and TORCHDYNAMO_VERBOSE=1 for more information
During handling of the above exception, another exception occurred:
Traceback (most recent call last):
File "/workspace/tutorials-kr/intermediate_source/torch_export_tutorial.py", line 396, in <module>
export(DynamicShapesExample1(), (inp1,), dynamic_shapes=dynamic_shapes1_bad)
File "/usr/local/lib/python3.10/dist-packages/torch/export/__init__.py", line 174, in export
return _export(
File "/usr/local/lib/python3.10/dist-packages/torch/export/_trace.py", line 635, in wrapper
raise e
File "/usr/local/lib/python3.10/dist-packages/torch/export/_trace.py", line 618, in wrapper
ep = fn(*args, **kwargs)
File "/usr/local/lib/python3.10/dist-packages/torch/export/exported_program.py", line 83, in wrapper
return fn(*args, **kwargs)
File "/usr/local/lib/python3.10/dist-packages/torch/export/_trace.py", line 860, in _export
gm_torch_level = _export_to_torch_ir(
File "/usr/local/lib/python3.10/dist-packages/torch/export/_trace.py", line 359, in _export_to_torch_ir
raise UserError(UserErrorType.CONSTRAINT_VIOLATION, str(e)) # noqa: TRY200
torch._dynamo.exc.UserError: 10 not in range [11, 18]
Set TORCH_LOGS="+dynamo" and TORCHDYNAMO_VERBOSE=1 for more information
We can enforce that equalities between dimensions of different tensors
by using the same torch.export.Dim object, for example, in matrix multiplication:
inp2 = torch.randn(4, 8)
inp3 = torch.randn(8, 2)
class DynamicShapesExample2(torch.nn.Module):
def forward(self, x, y):
return x @ y
inp2_dim0 = Dim("inp2_dim0")
inner_dim = Dim("inner_dim")
inp3_dim1 = Dim("inp3_dim1")
dynamic_shapes2 = {
"x": {0: inp2_dim0, 1: inner_dim},
"y": {0: inner_dim, 1: inp3_dim1},
}
exported_dynamic_shapes_example2 = export(DynamicShapesExample2(), (inp2, inp3), dynamic_shapes=dynamic_shapes2)
print(exported_dynamic_shapes_example2.module()(torch.randn(2, 16), torch.randn(16, 4)))
try:
exported_dynamic_shapes_example2.module()(torch.randn(4, 8), torch.randn(4, 2))
except Exception:
tb.print_exc()
tensor([[-2.9354, -2.2066, -0.2080, 4.6121],
[-0.5658, -0.6108, 0.8887, 1.5908]])
Traceback (most recent call last):
File "/workspace/tutorials-kr/intermediate_source/torch_export_tutorial.py", line 425, in <module>
exported_dynamic_shapes_example2.module()(torch.randn(4, 8), torch.randn(4, 2))
File "/usr/local/lib/python3.10/dist-packages/torch/fx/graph_module.py", line 737, in call_wrapped
return self._wrapped_call(self, *args, **kwargs)
File "/usr/local/lib/python3.10/dist-packages/torch/fx/graph_module.py", line 317, in __call__
raise e
File "/usr/local/lib/python3.10/dist-packages/torch/fx/graph_module.py", line 304, in __call__
return super(self.cls, obj).__call__(*args, **kwargs) # type: ignore[misc]
File "/usr/local/lib/python3.10/dist-packages/torch/nn/modules/module.py", line 1532, in _wrapped_call_impl
return self._call_impl(*args, **kwargs)
File "/usr/local/lib/python3.10/dist-packages/torch/nn/modules/module.py", line 1561, in _call_impl
args_kwargs_result = hook(self, args, kwargs) # type: ignore[misc]
File "/usr/local/lib/python3.10/dist-packages/torch/_dynamo/eval_frame.py", line 451, in _fn
return fn(*args, **kwargs)
File "/usr/local/lib/python3.10/dist-packages/torch/_dynamo/external_utils.py", line 36, in inner
return fn(*args, **kwargs)
File "/usr/local/lib/python3.10/dist-packages/torch/export/_unlift.py", line 32, in _check_input_constraints_pre_hook
return _check_input_constraints_for_graph(
File "/usr/local/lib/python3.10/dist-packages/torch/_export/utils.py", line 85, in _check_input_constraints_for_graph
raise RuntimeError(
RuntimeError: Expected input at *args[1].shape[0] to be equal to 8, but got 4
We can also describe one dimension in terms of other. There are some
restrictions to how detailed we can specify one dimension in terms of another,
but generally, those in the form of A * Dim + B should work.
class DerivedDimExample1(torch.nn.Module):
def forward(self, x, y):
return x + y[1:]
foo = DerivedDimExample1()
x, y = torch.randn(5), torch.randn(6)
dimx = torch.export.Dim("dimx", min=3, max=6)
dimy = dimx + 1
derived_dynamic_shapes1 = ({0: dimx}, {0: dimy})
derived_dim_example1 = export(foo, (x, y), dynamic_shapes=derived_dynamic_shapes1)
print(derived_dim_example1.module()(torch.randn(4), torch.randn(5)))
try:
derived_dim_example1.module()(torch.randn(4), torch.randn(6))
except Exception:
tb.print_exc()
class DerivedDimExample2(torch.nn.Module):
def forward(self, z, y):
return z[1:] + y[1::3]
foo = DerivedDimExample2()
z, y = torch.randn(4), torch.randn(10)
dx = torch.export.Dim("dx", min=3, max=6)
dz = dx + 1
dy = dx * 3 + 1
derived_dynamic_shapes2 = ({0: dz}, {0: dy})
derived_dim_example2 = export(foo, (z, y), dynamic_shapes=derived_dynamic_shapes2)
print(derived_dim_example2.module()(torch.randn(7), torch.randn(19)))
tensor([ 0.3007, -1.7282, -0.0729, 0.1139])
Traceback (most recent call last):
File "/workspace/tutorials-kr/intermediate_source/torch_export_tutorial.py", line 450, in <module>
derived_dim_example1.module()(torch.randn(4), torch.randn(6))
File "/usr/local/lib/python3.10/dist-packages/torch/fx/graph_module.py", line 737, in call_wrapped
return self._wrapped_call(self, *args, **kwargs)
File "/usr/local/lib/python3.10/dist-packages/torch/fx/graph_module.py", line 317, in __call__
raise e
File "/usr/local/lib/python3.10/dist-packages/torch/fx/graph_module.py", line 304, in __call__
return super(self.cls, obj).__call__(*args, **kwargs) # type: ignore[misc]
File "/usr/local/lib/python3.10/dist-packages/torch/nn/modules/module.py", line 1532, in _wrapped_call_impl
return self._call_impl(*args, **kwargs)
File "/usr/local/lib/python3.10/dist-packages/torch/nn/modules/module.py", line 1561, in _call_impl
args_kwargs_result = hook(self, args, kwargs) # type: ignore[misc]
File "/usr/local/lib/python3.10/dist-packages/torch/_dynamo/eval_frame.py", line 451, in _fn
return fn(*args, **kwargs)
File "/usr/local/lib/python3.10/dist-packages/torch/_dynamo/external_utils.py", line 36, in inner
return fn(*args, **kwargs)
File "/usr/local/lib/python3.10/dist-packages/torch/export/_unlift.py", line 32, in _check_input_constraints_pre_hook
return _check_input_constraints_for_graph(
File "/usr/local/lib/python3.10/dist-packages/torch/_export/utils.py", line 85, in _check_input_constraints_for_graph
raise RuntimeError(
RuntimeError: Expected input at *args[1].shape[0] to be equal to 5, but got 6
tensor([ 2.5416, -0.2760, 0.9003, -1.7479, -2.9716, 0.1013])
We can actually use torch.export to guide us as to which dynamic_shapes constraints
are necessary. We can do this by relaxing all constraints (recall that if we
do not provide constraints for a dimension, the default behavior is to constrain
to the exact shape value of the example input) and letting torch.export
error out.
inp4 = torch.randn(8, 16)
inp5 = torch.randn(16, 32)
class DynamicShapesExample3(torch.nn.Module):
def forward(self, x, y):
if x.shape[0] <= 16:
return x @ y[:, :16]
return y
dynamic_shapes3 = {
"x": {i: Dim(f"inp4_dim{i}") for i in range(inp4.dim())},
"y": {i: Dim(f"inp5_dim{i}") for i in range(inp5.dim())},
}
try:
export(DynamicShapesExample3(), (inp4, inp5), dynamic_shapes=dynamic_shapes3)
except Exception:
tb.print_exc()
Traceback (most recent call last):
File "/usr/local/lib/python3.10/dist-packages/torch/export/_trace.py", line 347, in _export_to_torch_ir
gm_torch_level, _ = torch._dynamo.export(
File "/usr/local/lib/python3.10/dist-packages/torch/_dynamo/eval_frame.py", line 1354, in inner
raise constraint_violation_error
File "/usr/local/lib/python3.10/dist-packages/torch/_dynamo/eval_frame.py", line 1311, in inner
result_traced = opt_f(*args, **kwargs)
File "/usr/local/lib/python3.10/dist-packages/torch/nn/modules/module.py", line 1532, in _wrapped_call_impl
return self._call_impl(*args, **kwargs)
File "/usr/local/lib/python3.10/dist-packages/torch/nn/modules/module.py", line 1541, in _call_impl
return forward_call(*args, **kwargs)
File "/usr/local/lib/python3.10/dist-packages/torch/_dynamo/eval_frame.py", line 451, in _fn
return fn(*args, **kwargs)
File "/usr/local/lib/python3.10/dist-packages/torch/nn/modules/module.py", line 1532, in _wrapped_call_impl
return self._call_impl(*args, **kwargs)
File "/usr/local/lib/python3.10/dist-packages/torch/nn/modules/module.py", line 1541, in _call_impl
return forward_call(*args, **kwargs)
File "/usr/local/lib/python3.10/dist-packages/torch/_dynamo/convert_frame.py", line 921, in catch_errors
return callback(frame, cache_entry, hooks, frame_state, skip=1)
File "/usr/local/lib/python3.10/dist-packages/torch/_dynamo/convert_frame.py", line 400, in _convert_frame_assert
return _compile(
File "/usr/lib/python3.10/contextlib.py", line 79, in inner
return func(*args, **kwds)
File "/usr/local/lib/python3.10/dist-packages/torch/_dynamo/convert_frame.py", line 676, in _compile
guarded_code = compile_inner(code, one_graph, hooks, transform)
File "/usr/local/lib/python3.10/dist-packages/torch/_dynamo/utils.py", line 262, in time_wrapper
r = func(*args, **kwargs)
File "/usr/local/lib/python3.10/dist-packages/torch/_dynamo/convert_frame.py", line 634, in compile_inner
check_fn = CheckFunctionManager(
File "/usr/local/lib/python3.10/dist-packages/torch/_dynamo/guards.py", line 1048, in __init__
guard.create(builder)
File "/usr/local/lib/python3.10/dist-packages/torch/_guards.py", line 249, in create
return self.create_fn(builder, self)
File "/usr/local/lib/python3.10/dist-packages/torch/_dynamo/guards.py", line 705, in SHAPE_ENV
guards = output_graph.shape_env.produce_guards(
File "/usr/local/lib/python3.10/dist-packages/torch/fx/experimental/symbolic_shapes.py", line 3308, in produce_guards
raise ConstraintViolationError(
torch.fx.experimental.symbolic_shapes.ConstraintViolationError: Constraints violated (inp5_dim0, inp4_dim0, inp5_dim1)! For more information, run with TORCH_LOGS="+dynamic".
- The values of inp5_dim0 = L['y'].size()[0] and inp4_dim1 = L['x'].size()[1] must always be equal.
- Not all values of inp5_dim1 = L['y'].size()[1] in the specified range satisfy the generated guard Ne(L['y'].size()[1], 16).
- Not all values of inp4_dim0 = L['x'].size()[0] in the specified range satisfy the generated guard 2 <= L['x'].size()[0] and L['x'].size()[0] <= 16
- Not all values of inp5_dim1 = L['y'].size()[1] in the specified range satisfy the generated guard 16 <= L['y'].size()[1] and L['y'].size()[1] <= 9223372036854775806
Suggested fixes:
inp4_dim0 = Dim('inp4_dim0', max=16)
inp4_dim1 = Dim('inp4_dim1')
inp4_dim1 = Dim('inp4_dim1', min=2, max=9223372036854775806) # 2 <= inp4_dim1 <= 9223372036854775806
inp5_dim1 = Dim('inp5_dim1', min=16)
inp5_dim0 = inp4_dim1
During handling of the above exception, another exception occurred:
Traceback (most recent call last):
File "/workspace/tutorials-kr/intermediate_source/torch_export_tutorial.py", line 492, in <module>
export(DynamicShapesExample3(), (inp4, inp5), dynamic_shapes=dynamic_shapes3)
File "/usr/local/lib/python3.10/dist-packages/torch/export/__init__.py", line 174, in export
return _export(
File "/usr/local/lib/python3.10/dist-packages/torch/export/_trace.py", line 635, in wrapper
raise e
File "/usr/local/lib/python3.10/dist-packages/torch/export/_trace.py", line 618, in wrapper
ep = fn(*args, **kwargs)
File "/usr/local/lib/python3.10/dist-packages/torch/export/exported_program.py", line 83, in wrapper
return fn(*args, **kwargs)
File "/usr/local/lib/python3.10/dist-packages/torch/export/_trace.py", line 860, in _export
gm_torch_level = _export_to_torch_ir(
File "/usr/local/lib/python3.10/dist-packages/torch/export/_trace.py", line 359, in _export_to_torch_ir
raise UserError(UserErrorType.CONSTRAINT_VIOLATION, str(e)) # noqa: TRY200
torch._dynamo.exc.UserError: Constraints violated (inp5_dim0, inp4_dim0, inp5_dim1)! For more information, run with TORCH_LOGS="+dynamic".
- The values of inp5_dim0 = L['y'].size()[0] and inp4_dim1 = L['x'].size()[1] must always be equal.
- Not all values of inp5_dim1 = L['y'].size()[1] in the specified range satisfy the generated guard Ne(L['y'].size()[1], 16).
- Not all values of inp4_dim0 = L['x'].size()[0] in the specified range satisfy the generated guard 2 <= L['x'].size()[0] and L['x'].size()[0] <= 16
- Not all values of inp5_dim1 = L['y'].size()[1] in the specified range satisfy the generated guard 16 <= L['y'].size()[1] and L['y'].size()[1] <= 9223372036854775806
Suggested fixes:
inp4_dim0 = Dim('inp4_dim0', max=16)
inp4_dim1 = Dim('inp4_dim1')
inp4_dim1 = Dim('inp4_dim1', min=2, max=9223372036854775806) # 2 <= inp4_dim1 <= 9223372036854775806
inp5_dim1 = Dim('inp5_dim1', min=16)
inp5_dim0 = inp4_dim1
We can see that the error message gives us suggested fixes to our dynamic shape constraints. Let us follow those suggestions (exact suggestions may differ slightly):
def suggested_fixes():
inp4_dim1 = Dim('shared_dim')
# suggested fixes below
inp4_dim0 = Dim('inp4_dim0', max=16)
inp5_dim1 = Dim('inp5_dim1', min=17)
inp5_dim0 = inp4_dim1
# end of suggested fixes
return {
"x": {0: inp4_dim0, 1: inp4_dim1},
"y": {0: inp5_dim0, 1: inp5_dim1},
}
dynamic_shapes3_fixed = suggested_fixes()
exported_dynamic_shapes_example3 = export(DynamicShapesExample3(), (inp4, inp5), dynamic_shapes=dynamic_shapes3_fixed)
print(exported_dynamic_shapes_example3.module()(torch.randn(4, 32), torch.randn(32, 64)))
tensor([[ 12.5915, -0.7265, -1.8981, -8.0323, -2.1447, 14.4020, 13.7854,
1.5568, 3.7933, 7.2591, -2.2477, 1.6366, -5.9276, 8.5279,
7.9349, -1.1328],
[ -5.2210, 9.4576, -0.2372, 9.0035, 6.2572, -8.4716, 6.0191,
4.8424, -0.4486, 0.1885, -0.1749, 2.4314, 3.8271, 8.1822,
6.5064, 0.6512],
[ -3.6856, 7.5222, 4.8073, 13.1255, 3.6440, -4.1587, 2.9806,
0.3689, 1.1133, -1.7169, -2.1537, 1.1841, 6.7619, 9.3401,
-1.1372, -8.9628],
[ -4.3608, 5.1219, -0.6240, -6.8640, 2.3344, -2.0273, 0.2769,
-0.9930, 2.0298, -10.3922, -1.7186, -5.0928, 11.7383, 6.4864,
8.0827, 0.5863]])
Note that in the example above, because we constrained the value of x.shape[0] in
dynamic_shapes_example3, the exported program is sound even though there is a
raw if statement.
If you want to see why torch.export generated these constraints, you can
re-run the script with the environment variable TORCH_LOGS=dynamic,dynamo,
or use torch._logging.set_logs.
import logging
torch._logging.set_logs(dynamic=logging.INFO, dynamo=logging.INFO)
exported_dynamic_shapes_example3 = export(DynamicShapesExample3(), (inp4, inp5), dynamic_shapes=dynamic_shapes3_fixed)
# reset to previous values
torch._logging.set_logs(dynamic=logging.WARNING, dynamo=logging.WARNING)
I0613 07:10:17.674000 128806481463104 torch/_dynamo/logging.py:55] [18/0] Step 1: torchdynamo start tracing forward /workspace/tutorials-kr/intermediate_source/torch_export_tutorial.py:481
I0613 07:10:17.675000 128806481463104 torch/fx/experimental/symbolic_shapes.py:2724] [18/0] create_symbol s0 = 8 for L['x'].size()[0] [2, 16] (_dynamo/variables/builder.py:1881 in <lambda>)
I0613 07:10:17.675000 128806481463104 torch/fx/experimental/symbolic_shapes.py:2724] [18/0] create_symbol s1 = 16 for L['x'].size()[1] [2, 9223372036854775806] (_dynamo/variables/builder.py:1881 in <lambda>)
I0613 07:10:17.677000 128806481463104 torch/fx/experimental/symbolic_shapes.py:2724] [18/0] create_symbol s2 = 16 for L['y'].size()[0] [2, 9223372036854775806] (_dynamo/variables/builder.py:1881 in <lambda>)
I0613 07:10:17.677000 128806481463104 torch/fx/experimental/symbolic_shapes.py:2724] [18/0] create_symbol s3 = 32 for L['y'].size()[1] [17, 9223372036854775806] (_dynamo/variables/builder.py:1881 in <lambda>)
I0613 07:10:17.682000 128806481463104 torch/fx/experimental/symbolic_shapes.py:3809] [18/0] set_replacement s2 = s1 (solve_backed) ValueRanges(lower=2, upper=9223372036854775806, is_bool=False)
I0613 07:10:17.682000 128806481463104 torch/fx/experimental/symbolic_shapes.py:4035] [18/0] eval Eq(s1, s2) [guard added] at orkspace/tutorials-kr/intermediate_source/torch_export_tutorial.py:483 in forward (_meta_registrations.py:2014 in meta_mm)
I0613 07:10:17.683000 128806481463104 torch/_dynamo/logging.py:55] [18/0] Step 1: torchdynamo done tracing forward (RETURN_VALUE)
I0613 07:10:17.684000 128806481463104 torch/fx/experimental/symbolic_shapes.py:3809] [18/0] set_replacement s2 = s1 (find) ValueRanges(lower=2, upper=9223372036854775806, is_bool=False)
I0613 07:10:17.684000 128806481463104 torch/_dynamo/logging.py:55] [18/0] Step 2: calling compiler function dynamo_normalization_capturing_compiler
I0613 07:10:17.684000 128806481463104 torch/_dynamo/logging.py:55] [18/0] Step 2: done compiler function dynamo_normalization_capturing_compiler
I0613 07:10:17.685000 128806481463104 torch/fx/experimental/symbolic_shapes.py:2806] [18/0] produce_guards
I0613 07:10:17.700000 128806481463104 torch/_dynamo/eval_frame.py:1339] Summary of dimension constraints:
I0613 07:10:17.700000 128806481463104 torch/_dynamo/eval_frame.py:1339] Suggested fixes:
I0613 07:10:17.700000 128806481463104 torch/_dynamo/eval_frame.py:1339] inp4_dim0 = Dim('inp4_dim0', max=16)
I0613 07:10:17.700000 128806481463104 torch/_dynamo/eval_frame.py:1339] inp5_dim1 = Dim('inp5_dim1', min=17)
I0613 07:10:17.700000 128806481463104 torch/_dynamo/eval_frame.py:1339] shared_dim = Dim('shared_dim')
I0613 07:10:17.701000 128806481463104 torch/_dynamo/eval_frame.py:1363] Dynamo captured graph:
I0613 07:10:17.701000 128806481463104 torch/_dynamo/eval_frame.py:1363]
I0613 07:10:17.701000 128806481463104 torch/_dynamo/eval_frame.py:1363] class GraphModule(torch.nn.Module):
I0613 07:10:17.701000 128806481463104 torch/_dynamo/eval_frame.py:1363] def forward(self, L_x_ : torch.Tensor, L_y_ : torch.Tensor):
I0613 07:10:17.701000 128806481463104 torch/_dynamo/eval_frame.py:1363] l_x_ = L_x_
I0613 07:10:17.701000 128806481463104 torch/_dynamo/eval_frame.py:1363] l_y_ = L_y_
I0613 07:10:17.701000 128806481463104 torch/_dynamo/eval_frame.py:1363]
I0613 07:10:17.701000 128806481463104 torch/_dynamo/eval_frame.py:1363] # File: /workspace/tutorials-kr/intermediate_source/torch_export_tutorial.py:482 in forward, code: if x.shape[0] <= 16:
I0613 07:10:17.701000 128806481463104 torch/_dynamo/eval_frame.py:1363] size = l_x_.size()
I0613 07:10:17.701000 128806481463104 torch/_dynamo/eval_frame.py:1363] getitem = size[0]; size = None
I0613 07:10:17.701000 128806481463104 torch/_dynamo/eval_frame.py:1363] le = getitem <= 16; getitem = None
I0613 07:10:17.701000 128806481463104 torch/_dynamo/eval_frame.py:1363]
I0613 07:10:17.701000 128806481463104 torch/_dynamo/eval_frame.py:1363] # File: /workspace/tutorials-kr/intermediate_source/torch_export_tutorial.py:483 in forward, code: return x @ y[:, :16]
I0613 07:10:17.701000 128806481463104 torch/_dynamo/eval_frame.py:1363] getitem_2 = l_y_[(slice(None, None, None), slice(None, 16, None))]; l_y_ = None
I0613 07:10:17.701000 128806481463104 torch/_dynamo/eval_frame.py:1363] matmul = l_x_ @ getitem_2; l_x_ = getitem_2 = None
I0613 07:10:17.701000 128806481463104 torch/_dynamo/eval_frame.py:1363] return (matmul,)
I0613 07:10:17.701000 128806481463104 torch/_dynamo/eval_frame.py:1363]
We can view an ExportedProgram’s symbolic shape ranges using the
range_constraints field.
print(exported_dynamic_shapes_example3.range_constraints)
{s0: ValueRanges(lower=2, upper=16, is_bool=False), s1: ValueRanges(lower=2, upper=oo, is_bool=False), s3: ValueRanges(lower=17, upper=oo, is_bool=False)}
Custom Ops¶
torch.export can export PyTorch programs with custom operators.
Currently, the steps to register a custom op for use by torch.export are:
Define the custom op using
torch.library(reference) as with any other custom op
from torch.library import Library, impl, impl_abstract
m = Library("my_custom_library", "DEF")
m.define("custom_op(Tensor input) -> Tensor")
@impl(m, "custom_op", "CompositeExplicitAutograd")
def custom_op(x):
print("custom_op called!")
return torch.relu(x)
Define a
"Meta"implementation of the custom op that returns an empty tensor with the same shape as the expected output
@impl_abstract("my_custom_library::custom_op")
def custom_op_meta(x):
return torch.empty_like(x)
Call the custom op from the code you want to export using
torch.ops
class CustomOpExample(torch.nn.Module):
def forward(self, x):
x = torch.sin(x)
x = torch.ops.my_custom_library.custom_op(x)
x = torch.cos(x)
return x
Export the code as before
exported_custom_op_example = export(CustomOpExample(), (torch.randn(3, 3),))
exported_custom_op_example.graph_module.print_readable()
print(exported_custom_op_example.module()(torch.randn(3, 3)))
class GraphModule(torch.nn.Module):
def forward(self, arg0_1: "f32[3, 3]"):
# File: /workspace/tutorials-kr/intermediate_source/torch_export_tutorial.py:574 in forward, code: x = torch.sin(x)
sin: "f32[3, 3]" = torch.ops.aten.sin.default(arg0_1); arg0_1 = None
# File: /workspace/tutorials-kr/intermediate_source/torch_export_tutorial.py:575 in forward, code: x = torch.ops.my_custom_library.custom_op(x)
custom_op: "f32[3, 3]" = torch.ops.my_custom_library.custom_op.default(sin); sin = None
# File: /workspace/tutorials-kr/intermediate_source/torch_export_tutorial.py:576 in forward, code: x = torch.cos(x)
cos: "f32[3, 3]" = torch.ops.aten.cos.default(custom_op); custom_op = None
return (cos,)
custom_op called!
tensor([[0.6387, 0.5722, 1.0000],
[0.8776, 0.9846, 0.8223],
[0.9983, 1.0000, 0.9979]])
Note in the above outputs that the custom op is included in the exported graph.
And when we call the exported graph as a function, the original custom op is called,
as evidenced by the print call.
If you have a custom operator implemented in C++, please refer to
this document
to make it compatible with torch.export.
Decompositions¶
The graph produced by torch.export by default returns a graph containing
only functional ATen operators. This functional ATen operator set (or 《opset》) contains around 2000
operators, all of which are functional, that is, they do not
mutate or alias inputs. You can find a list of all ATen operators
here
and you can inspect if an operator is functional by checking
op._schema.is_mutable, for example:
print(torch.ops.aten.add.Tensor._schema.is_mutable)
print(torch.ops.aten.add_.Tensor._schema.is_mutable)
False
True
By default, the environment in which you want to run the exported graph should support all ~2000 of these operators. However, you can use the following API on the exported program if your specific environment is only able to support a subset of the ~2000 operators.
def run_decompositions(
self: ExportedProgram,
decomposition_table: Optional[Dict[torch._ops.OperatorBase, Callable]]
) -> ExportedProgram
run_decompositions takes in a decomposition table, which is a mapping of
operators to a function specifying how to reduce, or decompose, that operator
into an equivalent sequence of other ATen operators.
The default decomposition table for run_decompositions is the
Core ATen decomposition table
which will decompose the all ATen operators to the
Core ATen Operator Set
which consists of only ~180 operators.
class M(torch.nn.Module):
def __init__(self):
super().__init__()
self.linear = torch.nn.Linear(3, 4)
def forward(self, x):
return self.linear(x)
ep = export(M(), (torch.randn(2, 3),))
print(ep.graph)
core_ir_ep = ep.run_decompositions()
print(core_ir_ep.graph)
graph():
%arg0_1 : [num_users=1] = placeholder[target=arg0_1]
%arg1_1 : [num_users=1] = placeholder[target=arg1_1]
%arg2_1 : [num_users=1] = placeholder[target=arg2_1]
%t : [num_users=1] = call_function[target=torch.ops.aten.t.default](args = (%arg0_1,), kwargs = {})
%addmm : [num_users=1] = call_function[target=torch.ops.aten.addmm.default](args = (%arg1_1, %arg2_1, %t), kwargs = {})
return (addmm,)
graph():
%arg0_1 : [num_users=1] = placeholder[target=arg0_1]
%arg1_1 : [num_users=1] = placeholder[target=arg1_1]
%arg2_1 : [num_users=1] = placeholder[target=arg2_1]
%permute : [num_users=1] = call_function[target=torch.ops.aten.permute.default](args = (%arg0_1, [1, 0]), kwargs = {})
%addmm : [num_users=1] = call_function[target=torch.ops.aten.addmm.default](args = (%arg1_1, %arg2_1, %permute), kwargs = {})
return (addmm,)
Notice that after running run_decompositions the
torch.ops.aten.t.default operator, which is not part of the Core ATen
Opset, has been replaced with torch.ops.aten.permute.default which is part
of the Core ATen Opset.
Most ATen operators already have decompositions, which are located here. If you would like to use some of these existing decomposition functions, you can pass in a list of operators you would like to decompose to the get_decompositions function, which will return a decomposition table using existing decomposition implementations.
class M(torch.nn.Module):
def __init__(self):
super().__init__()
self.linear = torch.nn.Linear(3, 4)
def forward(self, x):
return self.linear(x)
ep = export(M(), (torch.randn(2, 3),))
print(ep.graph)
from torch._decomp import get_decompositions
decomp_table = get_decompositions([torch.ops.aten.t.default, torch.ops.aten.transpose.int])
core_ir_ep = ep.run_decompositions(decomp_table)
print(core_ir_ep.graph)
graph():
%arg0_1 : [num_users=1] = placeholder[target=arg0_1]
%arg1_1 : [num_users=1] = placeholder[target=arg1_1]
%arg2_1 : [num_users=1] = placeholder[target=arg2_1]
%t : [num_users=1] = call_function[target=torch.ops.aten.t.default](args = (%arg0_1,), kwargs = {})
%addmm : [num_users=1] = call_function[target=torch.ops.aten.addmm.default](args = (%arg1_1, %arg2_1, %t), kwargs = {})
return (addmm,)
graph():
%arg0_1 : [num_users=1] = placeholder[target=arg0_1]
%arg1_1 : [num_users=1] = placeholder[target=arg1_1]
%arg2_1 : [num_users=1] = placeholder[target=arg2_1]
%permute : [num_users=1] = call_function[target=torch.ops.aten.permute.default](args = (%arg0_1, [1, 0]), kwargs = {})
%addmm : [num_users=1] = call_function[target=torch.ops.aten.addmm.default](args = (%arg1_1, %arg2_1, %permute), kwargs = {})
return (addmm,)
If there is no existing decomposition function for an ATen operator that you would like to decompose, feel free to send a pull request into PyTorch implementing the decomposition!
ExportDB¶
torch.export will only ever export a single computation graph from a PyTorch program. Because of this requirement,
there will be Python or PyTorch features that are not compatible with torch.export, which will require users to
rewrite parts of their model code. We have seen examples of this earlier in the tutorial – for example, rewriting
if-statements using cond.
ExportDB is the standard reference that documents
supported and unsupported Python/PyTorch features for torch.export. It is essentially a list a program samples, each
of which represents the usage of one particular Python/PyTorch feature and its interaction with torch.export.
Examples are also tagged by category so that they can be more easily searched.
For example, let’s use ExportDB to get a better understanding of how the predicate works in the cond operator.
We can look at the example called cond_predicate, which has a torch.cond tag. The example code looks like:
def cond_predicate(x):
"""
The conditional statement (aka predicate) passed to ``cond()`` must be one of the following:
- ``torch.Tensor`` with a single element
- boolean expression
NOTE: If the `pred` is test on a dim with batch size < 2, it will be specialized.
"""
pred = x.dim() > 2 and x.shape[2] > 10
return cond(pred, lambda x: x.cos(), lambda y: y.sin(), [x])
More generally, ExportDB can be used as a reference when one of the following occurs:
Before attempting
torch.export, you know ahead of time that your model uses some tricky Python/PyTorch features and you want to know iftorch.exportcovers that feature.When attempting
torch.export, there is a failure and it’s unclear how to work around it.
ExportDB is not exhaustive, but is intended to cover all use cases found in typical PyTorch code. Feel free to reach
out if there is an important Python/PyTorch feature that should be added to ExportDB or supported by torch.export.
Running the Exported Program¶
As torch.export is only a graph capturing mechanism, calling the artifact
produced by torch.export eagerly will be equivalent to running the eager
module. To optimize the execution of the Exported Program, we can pass this
exported artifact to backends such as Inductor through torch.compile,
AOTInductor,
or TensorRT.
class M(torch.nn.Module):
def __init__(self):
super().__init__()
self.linear = torch.nn.Linear(3, 3)
def forward(self, x):
x = self.linear(x)
return x
inp = torch.randn(2, 3, device="cuda")
m = M().to(device="cuda")
ep = torch.export.export(m, (inp,))
# Run it eagerly
res = ep.module()(inp)
print(res)
# Run it with torch.compile
res = torch.compile(ep.module(), backend="inductor")(inp)
print(res)
tensor([[ 1.3676, -0.4303, -0.2113],
[-0.5053, -0.0877, 0.5134]], device='cuda:0',
grad_fn=<AddmmBackward0>)
tensor([[ 1.3676, -0.4303, -0.2113],
[-0.5053, -0.0877, 0.5134]], device='cuda:0',
grad_fn=<CompiledFunctionBackward>)
import torch._export
import torch._inductor
# Note: these APIs are subject to change
# Compile the exported program to a .so using ``AOTInductor``
with torch.no_grad():
so_path = torch._inductor.aot_compile(ep.module(), [inp])
# Load and run the .so file in Python.
# To load and run it in a C++ environment, see:
# https://pytorch.org/docs/main/torch.compiler_aot_inductor.html
res = torch._export.aot_load(so_path, device="cuda")(inp)
Conclusion¶
We introduced torch.export, the new PyTorch 2.X way to export single computation
graphs from PyTorch programs. In particular, we demonstrate several code modifications
and considerations (control flow ops, constraints, etc.) that need to be made in order to export a graph.
Total running time of the script: ( 0 minutes 1.206 seconds)