torch.dynamic-shape¶
cond_branch_class_method¶
Original source code:
import torch
from functorch.experimental.control_flow import cond
class MySubModule(torch.nn.Module):
def foo(self, x):
return x.cos()
def forward(self, x):
return self.foo(x)
class CondBranchClassMethod(torch.nn.Module):
"""
The branch functions (`true_fn` and `false_fn`) passed to cond() must follow these rules:
- both branches must take the same args, which must also match the branch args passed to cond.
- both branches must return a single tensor
- returned tensor must have the same tensor metadata, e.g. shape and dtype
- branch function can be free function, nested function, lambda, class methods
- branch function can not have closure variables
- no inplace mutations on inputs or global variables
This example demonstrates using class method in cond().
NOTE: If the `pred` is test on a dim with batch size < 2, it will be specialized.
"""
def __init__(self):
super().__init__()
self.subm = MySubModule()
def bar(self, x):
return x.sin()
def forward(self, x):
return cond(x.shape[0] <= 2, self.subm.forward, self.bar, [x])
Result:
ExportedProgram:
class GraphModule(torch.nn.Module):
def forward(self, arg0_1: "f32[3]"):
true_graph_0 = self.true_graph_0
false_graph_0 = self.false_graph_0
conditional = torch.ops.higher_order.cond(False, true_graph_0, false_graph_0, [arg0_1]); true_graph_0 = false_graph_0 = arg0_1 = None
getitem: "f32[3]" = conditional[0]; conditional = None
return (getitem,)
class <lambda>(torch.nn.Module):
def forward(self, arg0_1: "f32[3]"):
cos: "f32[3]" = torch.ops.aten.cos.default(arg0_1); arg0_1 = None
return (cos,)
class <lambda>(torch.nn.Module):
def forward(self, arg0_1: "f32[3]"):
sin: "f32[3]" = torch.ops.aten.sin.default(arg0_1); arg0_1 = None
return (sin,)
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='getitem'), target=None)])
Range constraints: {}
cond_branch_nested_function¶
Original source code:
import torch
from functorch.experimental.control_flow import cond
class CondBranchNestedFunction(torch.nn.Module):
"""
The branch functions (`true_fn` and `false_fn`) passed to cond() must follow these rules:
- both branches must take the same args, which must also match the branch args passed to cond.
- both branches must return a single tensor
- returned tensor must have the same tensor metadata, e.g. shape and dtype
- branch function can be free function, nested function, lambda, class methods
- branch function can not have closure variables
- no inplace mutations on inputs or global variables
This example demonstrates using nested function in cond().
NOTE: If the `pred` is test on a dim with batch size < 2, it will be specialized.
"""
def __init__(self):
super().__init__()
def forward(self, x):
def true_fn(x):
def inner_true_fn(y):
return x + y
return inner_true_fn(x)
def false_fn(x):
def inner_false_fn(y):
return x - y
return inner_false_fn(x)
return cond(x.shape[0] < 10, true_fn, false_fn, [x])
Result:
ExportedProgram:
class GraphModule(torch.nn.Module):
def forward(self, arg0_1: "f32[3]"):
true_graph_0 = self.true_graph_0
false_graph_0 = self.false_graph_0
conditional = torch.ops.higher_order.cond(True, true_graph_0, false_graph_0, [arg0_1]); true_graph_0 = false_graph_0 = arg0_1 = None
getitem: "f32[3]" = conditional[0]; conditional = None
return (getitem,)
class <lambda>(torch.nn.Module):
def forward(self, arg0_1: "f32[3]"):
add: "f32[3]" = torch.ops.aten.add.Tensor(arg0_1, arg0_1); arg0_1 = None
return (add,)
class <lambda>(torch.nn.Module):
def forward(self, arg0_1: "f32[3]"):
sub: "f32[3]" = torch.ops.aten.sub.Tensor(arg0_1, arg0_1); arg0_1 = None
return (sub,)
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='getitem'), target=None)])
Range constraints: {}
cond_branch_nonlocal_variables¶
Original source code:
import torch
from functorch.experimental.control_flow import cond
class CondBranchNonlocalVariables(torch.nn.Module):
"""
The branch functions (`true_fn` and `false_fn`) passed to cond() must follow these rules:
- both branches must take the same args, which must also match the branch args passed to cond.
- both branches must return a single tensor
- returned tensor must have the same tensor metadata, e.g. shape and dtype
- branch function can be free function, nested function, lambda, class methods
- branch function can not have closure variables
- no inplace mutations on inputs or global variables
This example demonstrates how to rewrite code to avoid capturing closure variables in branch functions.
The code below will not work because capturing closure variables is not supported.
```
my_tensor_var = x + 100
my_primitive_var = 3.14
def true_fn(y):
nonlocal my_tensor_var, my_primitive_var
return y + my_tensor_var + my_primitive_var
def false_fn(y):
nonlocal my_tensor_var, my_primitive_var
return y - my_tensor_var - my_primitive_var
return cond(x.shape[0] > 5, true_fn, false_fn, [x])
```
NOTE: If the `pred` is test on a dim with batch size < 2, it will be specialized.
"""
def __init__(self):
super().__init__()
def forward(self, x):
my_tensor_var = x + 100
my_primitive_var = 3.14
def true_fn(x, y, z):
return x + y + z
def false_fn(x, y, z):
return x - y - z
return cond(
x.shape[0] > 5,
true_fn,
false_fn,
[x, my_tensor_var, torch.tensor(my_primitive_var)],
)
Result:
ExportedProgram:
class GraphModule(torch.nn.Module):
def forward(self, _lifted_tensor_constant0: "f32[]", arg0_1: "f32[6]"):
add: "f32[6]" = torch.ops.aten.add.Tensor(arg0_1, 100)
lift_fresh_copy: "f32[]" = torch.ops.aten.lift_fresh_copy.default(_lifted_tensor_constant0); _lifted_tensor_constant0 = None
true_graph_0 = self.true_graph_0
false_graph_0 = self.false_graph_0
conditional = torch.ops.higher_order.cond(True, true_graph_0, false_graph_0, [arg0_1, add, lift_fresh_copy]); true_graph_0 = false_graph_0 = arg0_1 = add = lift_fresh_copy = None
getitem: "f32[6]" = conditional[0]; conditional = None
return (getitem,)
class <lambda>(torch.nn.Module):
def forward(self, arg0_1: "f32[6]", arg1_1: "f32[6]", arg2_1: "f32[]"):
add: "f32[6]" = torch.ops.aten.add.Tensor(arg0_1, arg1_1); arg0_1 = arg1_1 = None
add_1: "f32[6]" = torch.ops.aten.add.Tensor(add, arg2_1); add = arg2_1 = None
return (add_1,)
class <lambda>(torch.nn.Module):
def forward(self, arg0_1: "f32[6]", arg1_1: "f32[6]", arg2_1: "f32[]"):
sub: "f32[6]" = torch.ops.aten.sub.Tensor(arg0_1, arg1_1); arg0_1 = arg1_1 = None
sub_1: "f32[6]" = torch.ops.aten.sub.Tensor(sub, arg2_1); sub = arg2_1 = None
return (sub_1,)
Graph signature: ExportGraphSignature(input_specs=[InputSpec(kind=<InputKind.CONSTANT_TENSOR: 4>, arg=TensorArgument(name='_lifted_tensor_constant0'), target='_lifted_tensor_constant0', persistent=None), 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='getitem'), target=None)])
Range constraints: {}
cond_operands¶
Original source code:
import torch
from torch.export import Dim
from functorch.experimental.control_flow import cond
x = torch.randn(3, 2)
y = torch.ones(2)
dim0_x = Dim("dim0_x")
class CondOperands(torch.nn.Module):
"""
The operands passed to cond() must be:
- a list of tensors
- match arguments of `true_fn` and `false_fn`
NOTE: If the `pred` is test on a dim with batch size < 2, it will be specialized.
"""
def __init__(self):
super().__init__()
def forward(self, x, y):
def true_fn(x, y):
return x + y
def false_fn(x, y):
return x - y
return cond(x.shape[0] > 2, true_fn, false_fn, [x, y])
Result:
ExportedProgram:
class GraphModule(torch.nn.Module):
def forward(self, arg0_1: "f32[s0, 2]", arg1_1: "f32[2]"):
sym_size_int: "Sym(s0)" = torch.ops.aten.sym_size.int(arg0_1, 0)
gt: "Sym(s0 > 2)" = sym_size_int > 2; sym_size_int = None
true_graph_0 = self.true_graph_0
false_graph_0 = self.false_graph_0
conditional = torch.ops.higher_order.cond(gt, true_graph_0, false_graph_0, [arg0_1, arg1_1]); gt = true_graph_0 = false_graph_0 = arg0_1 = arg1_1 = None
getitem: "f32[s0, 2]" = conditional[0]; conditional = None
return (getitem,)
class <lambda>(torch.nn.Module):
def forward(self, arg0_1: "f32[s0, 2]", arg1_1: "f32[2]"):
add: "f32[s0, 2]" = torch.ops.aten.add.Tensor(arg0_1, arg1_1); arg0_1 = arg1_1 = None
return (add,)
class <lambda>(torch.nn.Module):
def forward(self, arg0_1: "f32[s0, 2]", arg1_1: "f32[2]"):
sub: "f32[s0, 2]" = torch.ops.aten.sub.Tensor(arg0_1, arg1_1); arg0_1 = arg1_1 = None
return (sub,)
Graph signature: ExportGraphSignature(input_specs=[InputSpec(kind=<InputKind.USER_INPUT: 1>, arg=TensorArgument(name='arg0_1'), target=None, persistent=None), InputSpec(kind=<InputKind.USER_INPUT: 1>, arg=TensorArgument(name='arg1_1'), target=None, persistent=None)], output_specs=[OutputSpec(kind=<OutputKind.USER_OUTPUT: 1>, arg=TensorArgument(name='getitem'), target=None)])
Range constraints: {s0: ValueRanges(lower=2, upper=oo, is_bool=False)}
cond_predicate¶
Original source code:
import torch
from functorch.experimental.control_flow import cond
class CondPredicate(torch.nn.Module):
"""
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.
"""
def __init__(self):
super().__init__()
def forward(self, x):
pred = x.dim() > 2 and x.shape[2] > 10
return cond(pred, lambda x: x.cos(), lambda y: y.sin(), [x])
Result:
ExportedProgram:
class GraphModule(torch.nn.Module):
def forward(self, arg0_1: "f32[6, 4, 3]"):
true_graph_0 = self.true_graph_0
false_graph_0 = self.false_graph_0
conditional = torch.ops.higher_order.cond(False, true_graph_0, false_graph_0, [arg0_1]); true_graph_0 = false_graph_0 = arg0_1 = None
getitem: "f32[6, 4, 3]" = conditional[0]; conditional = None
return (getitem,)
class <lambda>(torch.nn.Module):
def forward(self, arg0_1: "f32[6, 4, 3]"):
cos: "f32[6, 4, 3]" = torch.ops.aten.cos.default(arg0_1); arg0_1 = None
return (cos,)
class <lambda>(torch.nn.Module):
def forward(self, arg0_1: "f32[6, 4, 3]"):
sin: "f32[6, 4, 3]" = torch.ops.aten.sin.default(arg0_1); arg0_1 = None
return (sin,)
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='getitem'), target=None)])
Range constraints: {}
dynamic_shape_constructor¶
Original source code:
import torch
class DynamicShapeConstructor(torch.nn.Module):
"""
Tensor constructors should be captured with dynamic shape inputs rather
than being baked in with static shape.
"""
def __init__(self):
super().__init__()
def forward(self, x):
return torch.ones(x.shape[0] * 2)
Result:
ExportedProgram:
class GraphModule(torch.nn.Module):
def forward(self, arg0_1: "f32[3, 2]"):
ones: "f32[6]" = torch.ops.aten.ones.default([6], device = device(type='cpu'), pin_memory = False)
return (ones,)
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='ones'), target=None)])
Range constraints: {}
dynamic_shape_if_guard¶
Original source code:
import torch
class DynamicShapeIfGuard(torch.nn.Module):
"""
`if` statement with backed dynamic shape predicate will be specialized into
one particular branch and generate a guard. However, export will fail if the
the dimension is marked as dynamic shape from higher level API.
"""
def forward(self, x):
if x.shape[0] == 3:
return x.cos()
return x.sin()
Result:
ExportedProgram:
class GraphModule(torch.nn.Module):
def forward(self, arg0_1: "f32[3, 2, 2]"):
cos: "f32[3, 2, 2]" = torch.ops.aten.cos.default(arg0_1); arg0_1 = None
return (cos,)
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='cos'), target=None)])
Range constraints: {}
dynamic_shape_map¶
Original source code:
import torch
from functorch.experimental.control_flow import map
class DynamicShapeMap(torch.nn.Module):
"""
functorch map() maps a function over the first tensor dimension.
"""
def __init__(self):
super().__init__()
def forward(self, xs, y):
def body(x, y):
return x + y
return map(body, xs, y)
Result:
ExportedProgram:
class GraphModule(torch.nn.Module):
def forward(self, arg0_1: "f32[3, 2]", arg1_1: "f32[2]"):
body_graph_0 = self.body_graph_0
map_impl = torch.ops.higher_order.map_impl(body_graph_0, [arg0_1], [arg1_1]); body_graph_0 = arg0_1 = arg1_1 = None
getitem: "f32[3, 2]" = map_impl[0]; map_impl = None
return (getitem,)
class <lambda>(torch.nn.Module):
def forward(self, arg0_1: "f32[2]", arg1_1: "f32[2]"):
add: "f32[2]" = torch.ops.aten.add.Tensor(arg0_1, arg1_1); arg0_1 = arg1_1 = None
return (add,)
Graph signature: ExportGraphSignature(input_specs=[InputSpec(kind=<InputKind.USER_INPUT: 1>, arg=TensorArgument(name='arg0_1'), target=None, persistent=None), InputSpec(kind=<InputKind.USER_INPUT: 1>, arg=TensorArgument(name='arg1_1'), target=None, persistent=None)], output_specs=[OutputSpec(kind=<OutputKind.USER_OUTPUT: 1>, arg=TensorArgument(name='getitem'), target=None)])
Range constraints: {}
dynamic_shape_round¶
Original source code:
import torch
from torch.export import Dim
x = torch.ones(3, 2)
dim0_x = Dim("dim0_x")
class DynamicShapeRound(torch.nn.Module):
"""
Calling round on dynamic shapes is not supported.
"""
def __init__(self):
super().__init__()
def forward(self, x):
return x[: round(x.shape[0] / 2)]
Result:
AssertionError:
dynamic_shape_slicing¶
Original source code:
import torch
class DynamicShapeSlicing(torch.nn.Module):
"""
Slices with dynamic shape arguments should be captured into the graph
rather than being baked in.
"""
def __init__(self):
super().__init__()
def forward(self, x):
return x[: x.shape[0] - 2, x.shape[1] - 1 :: 2]
Result:
ExportedProgram:
class GraphModule(torch.nn.Module):
def forward(self, arg0_1: "f32[3, 2]"):
slice_1: "f32[1, 2]" = torch.ops.aten.slice.Tensor(arg0_1, 0, 0, 1); arg0_1 = None
slice_2: "f32[1, 1]" = torch.ops.aten.slice.Tensor(slice_1, 1, 1, 9223372036854775807, 2); slice_1 = None
return (slice_2,)
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='slice_2'), target=None)])
Range constraints: {}
dynamic_shape_view¶
Original source code:
import torch
class DynamicShapeView(torch.nn.Module):
"""
Dynamic shapes should be propagated to view arguments instead of being
baked into the exported graph.
"""
def __init__(self):
super().__init__()
def forward(self, x):
new_x_shape = x.size()[:-1] + (2, 5)
x = x.view(*new_x_shape)
return x.permute(0, 2, 1)
Result:
ExportedProgram:
class GraphModule(torch.nn.Module):
def forward(self, arg0_1: "f32[10, 10]"):
view: "f32[10, 2, 5]" = torch.ops.aten.view.default(arg0_1, [10, 2, 5]); arg0_1 = None
permute: "f32[10, 5, 2]" = torch.ops.aten.permute.default(view, [0, 2, 1]); view = None
return (permute,)
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='permute'), target=None)])
Range constraints: {}
list_contains¶
Original source code:
import torch
class ListContains(torch.nn.Module):
"""
List containment relation can be checked on a dynamic shape or constants.
"""
def __init__(self):
super().__init__()
def forward(self, x):
assert x.size(-1) in [6, 2]
assert x.size(0) not in [4, 5, 6]
assert "monkey" not in ["cow", "pig"]
return x + x
Result:
ExportedProgram:
class GraphModule(torch.nn.Module):
def forward(self, arg0_1: "f32[3, 2]"):
add: "f32[3, 2]" = torch.ops.aten.add.Tensor(arg0_1, arg0_1); arg0_1 = None
return (add,)
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'), target=None)])
Range constraints: {}
scalar_output¶
Original source code:
import torch
from torch.export import Dim
x = torch.ones(3, 2)
dim1_x = Dim("dim1_x")
class ScalarOutput(torch.nn.Module):
"""
Returning scalar values from the graph is supported, in addition to Tensor
outputs. Symbolic shapes are captured and rank is specialized.
"""
def __init__(self):
super().__init__()
def forward(self, x):
return x.shape[1] + 1
Result:
ExportedProgram:
class GraphModule(torch.nn.Module):
def forward(self, arg0_1: "f32[3, s0]"):
# No stacktrace found for following nodes
sym_size_int: "Sym(s0)" = torch.ops.aten.sym_size.int(arg0_1, 1); arg0_1 = None
add: "Sym(s0 + 1)" = sym_size_int + 1; sym_size_int = None
return (add,)
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=SymIntArgument(name='add'), target=None)])
Range constraints: {s0: ValueRanges(lower=2, upper=oo, is_bool=False)}
