If

If - 16

Version

• name: If (GitHub)

• domain: main

• since_version: 16

• function: False

• support_level: SupportType.COMMON

• shape inference: True

This version of the operator has been available since version 16.

Summary

If conditional

Attributes

• else_branch (required): Graph to run if condition is false. Has N outputs: values you wish to be live-out to the enclosing scope. The number of outputs must match the number of outputs in the then_branch.

• then_branch (required): Graph to run if condition is true. Has N outputs: values you wish to be live-out to the enclosing scope. The number of outputs must match the number of outputs in the else_branch.

Inputs

• cond (heterogeneous) - B: Condition for the if

Outputs

Between 1 and 2147483647 outputs.

• outputs (variadic) - V: Values that are live-out to the enclosing scope. The return values in the then_branch and else_branch must be of the same data type. The then_branch and else_branch may produce tensors with the same element type and different shapes. If corresponding outputs from the then-branch and the else-branch have static shapes S1 and S2, then the shape of the corresponding output variable of the if- node (if present) must be compatible with both S1 and S2 as it represents the union of both possible shapes.For example, if in a model file, the the first output of then_branch is typed float tensor with shape [2] and the first output of else_branch is another float tensor with shape [3], If’s first output should have (a) no shape set, or (b) a shape of rank 1 with neither dim_value nor dim_param set, or (c) a shape of rank 1 with a unique dim_param. In contrast, the first output cannot have the shape [2] since [2] and [3] are not compatible.

Type Constraints

• V in ( optional(seq(tensor(bfloat16))), optional(seq(tensor(bool))), optional(seq(tensor(complex128))), optional(seq(tensor(complex64))), optional(seq(tensor(double))), optional(seq(tensor(float))), optional(seq(tensor(float16))), optional(seq(tensor(int16))), optional(seq(tensor(int32))), optional(seq(tensor(int64))), optional(seq(tensor(int8))), optional(seq(tensor(string))), optional(seq(tensor(uint16))), optional(seq(tensor(uint32))), optional(seq(tensor(uint64))), optional(seq(tensor(uint8))), optional(tensor(bfloat16)), optional(tensor(bool)), optional(tensor(complex128)), optional(tensor(complex64)), optional(tensor(double)), optional(tensor(float)), optional(tensor(float16)), optional(tensor(int16)), optional(tensor(int32)), optional(tensor(int64)), optional(tensor(int8)), optional(tensor(string)), optional(tensor(uint16)), optional(tensor(uint32)), optional(tensor(uint64)), optional(tensor(uint8)), seq(tensor(bfloat16)), seq(tensor(bool)), seq(tensor(complex128)), seq(tensor(complex64)), seq(tensor(double)), seq(tensor(float)), seq(tensor(float16)), seq(tensor(int16)), seq(tensor(int32)), seq(tensor(int64)), seq(tensor(int8)), seq(tensor(string)), seq(tensor(uint16)), seq(tensor(uint32)), seq(tensor(uint64)), seq(tensor(uint8)), tensor(bfloat16), tensor(bool), tensor(complex128), tensor(complex64), tensor(double), tensor(float), tensor(float16), tensor(int16), tensor(int32), tensor(int64), tensor(int8), tensor(string), tensor(uint16), tensor(uint32), tensor(uint64), tensor(uint8) ): All Tensor, Sequence(Tensor), Optional(Tensor), and Optional(Sequence(Tensor)) types

• B in ( tensor(bool) ): Only bool

Examples

if

# Given a bool scalar input cond.
# return constant tensor x if cond is True, otherwise return constant tensor y.

then_out = onnx.helper.make_tensor_value_info('then_out', onnx.TensorProto.FLOAT, [5])
else_out = onnx.helper.make_tensor_value_info('else_out', onnx.TensorProto.FLOAT, [5])

x = np.array([1, 2, 3, 4, 5]).astype(np.float32)
y = np.array([5, 4, 3, 2, 1]).astype(np.float32)

then_const_node = onnx.helper.make_node(
    'Constant',
    inputs=[],
    outputs=['then_out'],
    value=onnx.numpy_helper.from_array(x)
)

else_const_node = onnx.helper.make_node(
    'Constant',
    inputs=[],
    outputs=['else_out'],
    value=onnx.numpy_helper.from_array(y)
)

then_body = onnx.helper.make_graph(
    [then_const_node],
    'then_body',
    [],
    [then_out]
)

else_body = onnx.helper.make_graph(
    [else_const_node],
    'else_body',
    [],
    [else_out]
)

if_node = onnx.helper.make_node(
    'If',
    inputs=['cond'],
    outputs=['res'],
    then_branch=then_body,
    else_branch=else_body
)

cond = np.array(1).astype(bool)
res = x if cond else y
expect(if_node, inputs=[cond], outputs=[res], name='test_if',
       opset_imports=[onnx.helper.make_opsetid("", 11)])

if_seq

# Given a bool scalar input cond.
# return constant sequence x if cond is True, otherwise return constant sequence y.

then_out = onnx.helper.make_tensor_sequence_value_info('then_out', onnx.TensorProto.FLOAT, shape=[5])
else_out = onnx.helper.make_tensor_sequence_value_info('else_out', onnx.TensorProto.FLOAT, shape=[5])

x = [np.array([1, 2, 3, 4, 5]).astype(np.float32)]
y = [np.array([5, 4, 3, 2, 1]).astype(np.float32)]

then_const_node = onnx.helper.make_node(
    'Constant',
    inputs=[],
    outputs=['x'],
    value=onnx.numpy_helper.from_array(x[0])
)

then_seq_node = onnx.helper.make_node(
    'SequenceConstruct',
    inputs=['x'],
    outputs=['then_out']
)

else_const_node = onnx.helper.make_node(
    'Constant',
    inputs=[],
    outputs=['y'],
    value=onnx.numpy_helper.from_array(y[0])
)

else_seq_node = onnx.helper.make_node(
    'SequenceConstruct',
    inputs=['y'],
    outputs=['else_out']
)

then_body = onnx.helper.make_graph(
    [then_const_node, then_seq_node],
    'then_body',
    [],
    [then_out]
)

else_body = onnx.helper.make_graph(
    [else_const_node, else_seq_node],
    'else_body',
    [],
    [else_out]
)

if_node = onnx.helper.make_node(
    'If',
    inputs=['cond'],
    outputs=['res'],
    then_branch=then_body,
    else_branch=else_body
)

cond = np.array(1).astype(bool)
res = x if cond else y
expect(if_node, inputs=[cond], outputs=[res], name='test_if_seq',
       opset_imports=[onnx.helper.make_opsetid("", 13)])

if_optional

# Given a bool scalar input cond, return an empty optional sequence of
# tensor if True, return an optional sequence with value x
# (the input optional sequence) otherwise.

ten_in_tp = onnx.helper.make_tensor_type_proto(onnx.TensorProto.FLOAT, shape=[5])
seq_in_tp = onnx.helper.make_sequence_type_proto(ten_in_tp)

then_out_tensor_tp = onnx.helper.make_tensor_type_proto(onnx.TensorProto.FLOAT, shape=[5])
then_out_seq_tp = onnx.helper.make_sequence_type_proto(then_out_tensor_tp)
then_out_opt_tp = onnx.helper.make_optional_type_proto(then_out_seq_tp)
then_out = onnx.helper.make_value_info('optional_empty', then_out_opt_tp)

else_out_tensor_tp = onnx.helper.make_tensor_type_proto(onnx.TensorProto.FLOAT, shape=[5])
else_out_seq_tp = onnx.helper.make_sequence_type_proto(else_out_tensor_tp)
else_out_opt_tp = onnx.helper.make_optional_type_proto(else_out_seq_tp)
else_out = onnx.helper.make_value_info('else_opt', else_out_opt_tp)

x = [np.array([1, 2, 3, 4, 5]).astype(np.float32)]
cond = np.array(0).astype(bool)
res = compute_if_outputs(x, cond)

opt_empty_in = onnx.helper.make_node(
    'Optional',
    inputs=[],
    outputs=['optional_empty'],
    type=seq_in_tp
)

then_body = onnx.helper.make_graph(
    [opt_empty_in],
    'then_body',
    [],
    [then_out]
)

else_const_node = onnx.helper.make_node(
    'Constant',
    inputs=[],
    outputs=['x'],
    value=onnx.numpy_helper.from_array(x[0])
)

else_seq_node = onnx.helper.make_node(
    'SequenceConstruct',
    inputs=['x'],
    outputs=['else_seq']
)

else_optional_seq_node = onnx.helper.make_node(
    'Optional',
    inputs=['else_seq'],
    outputs=['else_opt']
)

else_body = onnx.helper.make_graph(
    [else_const_node, else_seq_node, else_optional_seq_node],
    'else_body',
    [],
    [else_out]
)

if_node = onnx.helper.make_node(
    'If',
    inputs=['cond'],
    outputs=['sequence'],
    then_branch=then_body,
    else_branch=else_body
)

expect(if_node, inputs=[cond], outputs=[res], name='test_if_opt',
       output_type_protos=[else_out_opt_tp],
       opset_imports=[onnx.helper.make_opsetid("", 16)])

Differences

0	0	If conditional	If conditional
1	1
2	2	**Attributes**	**Attributes**
3	3
4	4	* **else_branch** (required):	* **else_branch** (required):
5	5	Graph to run if condition is false. Has N outputs: values you wish	Graph to run if condition is false. Has N outputs: values you wish
6	6	to be live-out to the enclosing scope. The number of outputs must	to be live-out to the enclosing scope. The number of outputs must
7	7	match the number of outputs in the then_branch.	match the number of outputs in the then_branch.
8	8	* **then_branch** (required):	* **then_branch** (required):
9	9	Graph to run if condition is true. Has N outputs: values you wish to	Graph to run if condition is true. Has N outputs: values you wish to
10	10	be live-out to the enclosing scope. The number of outputs must match	be live-out to the enclosing scope. The number of outputs must match
11	11	the number of outputs in the else_branch.	the number of outputs in the else_branch.
12	12
13	13	**Inputs**	**Inputs**
14	14
15	15	* **cond** (heterogeneous) - **B**:	* **cond** (heterogeneous) - **B**:
16	16	Condition for the if	Condition for the if
17	17
18	18	**Outputs**	**Outputs**
19	19
20	20	Between 1 and 2147483647 outputs.	Between 1 and 2147483647 outputs.
21	21
22	22	* **outputs** (variadic) - **V**:	* **outputs** (variadic) - **V**:
23	23	Values that are live-out to the enclosing scope. The return values	Values that are live-out to the enclosing scope. The return values
24	24	in the then_branch and else_branch must be of the same data	in the then_branch and else_branch must be of the same data
25	25	type. The then_branch and else_branch may produce tensors with	type. The then_branch and else_branch may produce tensors with
26	26	the same element type and different shapes. If corresponding outputs	the same element type and different shapes. If corresponding outputs
27	27	from the then-branch and the else-branch have static shapes S1 and	from the then-branch and the else-branch have static shapes S1 and
28	28	S2, then the shape of the corresponding output variable of the if-	S2, then the shape of the corresponding output variable of the if-
29	29	node (if present) must be compatible with both S1 and S2 as it	node (if present) must be compatible with both S1 and S2 as it
30	30	represents the union of both possible shapes.For example, if in a	represents the union of both possible shapes.For example, if in a
31	31	model file, the the first output of then_branch is typed float	model file, the the first output of then_branch is typed float
32	32	tensor with shape [2] and the first output of else_branch is	tensor with shape [2] and the first output of else_branch is
33	33	another float tensor with shape [3], If's first output should have	another float tensor with shape [3], If's first output should have
34	34	(a) no shape set, or (b) a shape of rank 1 with neither dim_value	(a) no shape set, or (b) a shape of rank 1 with neither dim_value
35	35	nor dim_param set, or (c) a shape of rank 1 with a unique	nor dim_param set, or (c) a shape of rank 1 with a unique
36	36	dim_param. In contrast, the first output cannot have the shape [2]	dim_param. In contrast, the first output cannot have the shape [2]
37	37	since [2] and [3] are not compatible.	since [2] and [3] are not compatible.
38	38
39	39	**Type Constraints**	**Type Constraints**
40	40
41	41	* **V** in (	* **V** in (
	42		optional(seq(tensor(bfloat16))),
	43		optional(seq(tensor(bool))),
	44		optional(seq(tensor(complex128))),
	45		optional(seq(tensor(complex64))),
	46		optional(seq(tensor(double))),
	47		optional(seq(tensor(float))),
	48		optional(seq(tensor(float16))),
	49		optional(seq(tensor(int16))),
	50		optional(seq(tensor(int32))),
	51		optional(seq(tensor(int64))),
	52		optional(seq(tensor(int8))),
	53		optional(seq(tensor(string))),
	54		optional(seq(tensor(uint16))),
	55		optional(seq(tensor(uint32))),
	56		optional(seq(tensor(uint64))),
	57		optional(seq(tensor(uint8))),
	58		optional(tensor(bfloat16)),
	59		optional(tensor(bool)),
	60		optional(tensor(complex128)),
	61		optional(tensor(complex64)),
	62		optional(tensor(double)),
	63		optional(tensor(float)),
	64		optional(tensor(float16)),
	65		optional(tensor(int16)),
	66		optional(tensor(int32)),
	67		optional(tensor(int64)),
	68		optional(tensor(int8)),
	69		optional(tensor(string)),
	70		optional(tensor(uint16)),
	71		optional(tensor(uint32)),
	72		optional(tensor(uint64)),
	73		optional(tensor(uint8)),
	74		seq(tensor(bfloat16)),
42	75	seq(tensor(bool)),	seq(tensor(bool)),
43	76	seq(tensor(complex128)),	seq(tensor(complex128)),
44	77	seq(tensor(complex64)),	seq(tensor(complex64)),
45	78	seq(tensor(double)),	seq(tensor(double)),
46	79	seq(tensor(float)),	seq(tensor(float)),
47	80	seq(tensor(float16)),	seq(tensor(float16)),
48	81	seq(tensor(int16)),	seq(tensor(int16)),
49	82	seq(tensor(int32)),	seq(tensor(int32)),
50	83	seq(tensor(int64)),	seq(tensor(int64)),
51	84	seq(tensor(int8)),	seq(tensor(int8)),
52	85	seq(tensor(string)),	seq(tensor(string)),
53	86	seq(tensor(uint16)),	seq(tensor(uint16)),
54	87	seq(tensor(uint32)),	seq(tensor(uint32)),
55	88	seq(tensor(uint64)),	seq(tensor(uint64)),
56	89	seq(tensor(uint8)),	seq(tensor(uint8)),
	90		tensor(bfloat16),
57	91	tensor(bool),	tensor(bool),
58	92	tensor(complex128),	tensor(complex128),
59	93	tensor(complex64),	tensor(complex64),
60	94	tensor(double),	tensor(double),
61	95	tensor(float),	tensor(float),
62	96	tensor(float16),	tensor(float16),
63	97	tensor(int16),	tensor(int16),
64	98	tensor(int32),	tensor(int32),
65	99	tensor(int64),	tensor(int64),
66	100	tensor(int8),	tensor(int8),
67	101	tensor(string),	tensor(string),
68	102	tensor(uint16),	tensor(uint16),
69	103	tensor(uint32),	tensor(uint32),
70	104	tensor(uint64),	tensor(uint64),
71	105	tensor(uint8)	tensor(uint8)
72	106	):	):
73	107	All Tensor and Sequence types	All Tensor, Sequence(Tensor), Optional(Tensor), and
	108		Optional(Sequence(Tensor)) types
74	109	* **B** in (	* **B** in (
75	110	tensor(bool)	tensor(bool)
76	111	):	):
77	112	Only bool	Only bool

If - 13

Version

• name: If (GitHub)

• domain: main

• since_version: 13

• function: False

• support_level: SupportType.COMMON

• shape inference: True

This version of the operator has been available since version 13.

Summary

If conditional

Attributes

• else_branch (required): Graph to run if condition is false. Has N outputs: values you wish to be live-out to the enclosing scope. The number of outputs must match the number of outputs in the then_branch.

• then_branch (required): Graph to run if condition is true. Has N outputs: values you wish to be live-out to the enclosing scope. The number of outputs must match the number of outputs in the else_branch.

Inputs

• cond (heterogeneous) - B: Condition for the if

Outputs

Between 1 and 2147483647 outputs.

• outputs (variadic) - V: Values that are live-out to the enclosing scope. The return values in the then_branch and else_branch must be of the same data type. The then_branch and else_branch may produce tensors with the same element type and different shapes. If corresponding outputs from the then-branch and the else-branch have static shapes S1 and S2, then the shape of the corresponding output variable of the if- node (if present) must be compatible with both S1 and S2 as it represents the union of both possible shapes.For example, if in a model file, the the first output of then_branch is typed float tensor with shape [2] and the first output of else_branch is another float tensor with shape [3], If’s first output should have (a) no shape set, or (b) a shape of rank 1 with neither dim_value nor dim_param set, or (c) a shape of rank 1 with a unique dim_param. In contrast, the first output cannot have the shape [2] since [2] and [3] are not compatible.

Type Constraints

• V in ( seq(tensor(bool)), seq(tensor(complex128)), seq(tensor(complex64)), seq(tensor(double)), seq(tensor(float)), seq(tensor(float16)), seq(tensor(int16)), seq(tensor(int32)), seq(tensor(int64)), seq(tensor(int8)), seq(tensor(string)), seq(tensor(uint16)), seq(tensor(uint32)), seq(tensor(uint64)), seq(tensor(uint8)), tensor(bool), tensor(complex128), tensor(complex64), tensor(double), tensor(float), tensor(float16), tensor(int16), tensor(int32), tensor(int64), tensor(int8), tensor(string), tensor(uint16), tensor(uint32), tensor(uint64), tensor(uint8) ): All Tensor and Sequence types

• B in ( tensor(bool) ): Only bool

Differences

0	0	If conditional	If conditional
1	1
2	2	**Attributes**	**Attributes**
3	3
4	4	* **else_branch** (required):	* **else_branch** (required):
5	5	Graph to run if condition is false. Has N outputs: values you wish	Graph to run if condition is false. Has N outputs: values you wish
6	6	to be live-out to the enclosing scope. The number of outputs must	to be live-out to the enclosing scope. The number of outputs must
7	7	match the number of outputs in the then_branch.	match the number of outputs in the then_branch.
8	8	* **then_branch** (required):	* **then_branch** (required):
9	9	Graph to run if condition is true. Has N outputs: values you wish to	Graph to run if condition is true. Has N outputs: values you wish to
10	10	be live-out to the enclosing scope. The number of outputs must match	be live-out to the enclosing scope. The number of outputs must match
11	11	the number of outputs in the else_branch.	the number of outputs in the else_branch.
12	12
13	13	**Inputs**	**Inputs**
14	14
15	15	* **cond** (heterogeneous) - **B**:	* **cond** (heterogeneous) - **B**:
16	16	Condition for the if	Condition for the if
17	17
18	18	**Outputs**	**Outputs**
19	19
20	20	Between 1 and 2147483647 outputs.	Between 1 and 2147483647 outputs.
21	21
22	22	* **outputs** (variadic) - **V**:	* **outputs** (variadic) - **V**:
23	23	Values that are live-out to the enclosing scope. The return values	Values that are live-out to the enclosing scope. The return values
24	24	in the then_branch and else_branch must be of the same data	in the then_branch and else_branch must be of the same data
25	25	type. The then_branch and else_branch may produce tensors with	type. The then_branch and else_branch may produce tensors with
26	26	the same element type and different shapes. If corresponding outputs	the same element type and different shapes. If corresponding outputs
27	27	from the then-branch and the else-branch have static shapes S1 and	from the then-branch and the else-branch have static shapes S1 and
28	28	S2, then the shape of the corresponding output variable of the if-	S2, then the shape of the corresponding output variable of the if-
29	29	node (if present) must be compatible with both S1 and S2 as it	node (if present) must be compatible with both S1 and S2 as it
30	30	represents the union of both possible shapes.For example, if in a	represents the union of both possible shapes.For example, if in a
31	31	model file, the the first output of then_branch is typed float	model file, the the first output of then_branch is typed float
32	32	tensor with shape [2] and the first output of else_branch is	tensor with shape [2] and the first output of else_branch is
33	33	another float tensor with shape [3], If's first output should have	another float tensor with shape [3], If's first output should have
34	34	(a) no shape set, or (b) a shape of rank 1 with neither dim_value	(a) no shape set, or (b) a shape of rank 1 with neither dim_value
35	35	nor dim_param set, or (c) a shape of rank 1 with a unique	nor dim_param set, or (c) a shape of rank 1 with a unique
36	36	dim_param. In contrast, the first output cannot have the shape [2]	dim_param. In contrast, the first output cannot have the shape [2]
37	37	since [2] and [3] are not compatible.	since [2] and [3] are not compatible.
38	38
39	39	**Type Constraints**	**Type Constraints**
40	40
41	41	* **V** in (	* **V** in (
	42		seq(tensor(bool)),
	43		seq(tensor(complex128)),
	44		seq(tensor(complex64)),
	45		seq(tensor(double)),
	46		seq(tensor(float)),
	47		seq(tensor(float16)),
	48		seq(tensor(int16)),
	49		seq(tensor(int32)),
	50		seq(tensor(int64)),
	51		seq(tensor(int8)),
	52		seq(tensor(string)),
	53		seq(tensor(uint16)),
	54		seq(tensor(uint32)),
	55		seq(tensor(uint64)),
	56		seq(tensor(uint8)),
42	57	tensor(bool),	tensor(bool),
43	58	tensor(complex128),	tensor(complex128),
44	59	tensor(complex64),	tensor(complex64),
45	60	tensor(double),	tensor(double),
46	61	tensor(float),	tensor(float),
47	62	tensor(float16),	tensor(float16),
48	63	tensor(int16),	tensor(int16),
49	64	tensor(int32),	tensor(int32),
50	65	tensor(int64),	tensor(int64),
51	66	tensor(int8),	tensor(int8),
52	67	tensor(string),	tensor(string),
53	68	tensor(uint16),	tensor(uint16),
54	69	tensor(uint32),	tensor(uint32),
55	70	tensor(uint64),	tensor(uint64),
56	71	tensor(uint8)	tensor(uint8)
57	72	):	):
58	73	All Tensor types	All Tensor and Sequence types
59	74	* **B** in (	* **B** in (
60	75	tensor(bool)	tensor(bool)
61	76	):	):
62	77	Only bool	Only bool

If - 11

Version

• name: If (GitHub)

• domain: main

• since_version: 11

• function: False

• support_level: SupportType.COMMON

• shape inference: True

This version of the operator has been available since version 11.

Summary

If conditional

Attributes

• else_branch (required): Graph to run if condition is false. Has N outputs: values you wish to be live-out to the enclosing scope. The number of outputs must match the number of outputs in the then_branch.

• then_branch (required): Graph to run if condition is true. Has N outputs: values you wish to be live-out to the enclosing scope. The number of outputs must match the number of outputs in the else_branch.

Inputs

• cond (heterogeneous) - B: Condition for the if

Outputs

Between 1 and 2147483647 outputs.

• outputs (variadic) - V: Values that are live-out to the enclosing scope. The return values in the then_branch and else_branch must be of the same data type. The then_branch and else_branch may produce tensors with the same element type and different shapes. If corresponding outputs from the then-branch and the else-branch have static shapes S1 and S2, then the shape of the corresponding output variable of the if- node (if present) must be compatible with both S1 and S2 as it represents the union of both possible shapes.For example, if in a model file, the the first output of then_branch is typed float tensor with shape [2] and the first output of else_branch is another float tensor with shape [3], If’s first output should have (a) no shape set, or (b) a shape of rank 1 with neither dim_value nor dim_param set, or (c) a shape of rank 1 with a unique dim_param. In contrast, the first output cannot have the shape [2] since [2] and [3] are not compatible.

Type Constraints

• V in ( tensor(bool), tensor(complex128), tensor(complex64), tensor(double), tensor(float), tensor(float16), tensor(int16), tensor(int32), tensor(int64), tensor(int8), tensor(string), tensor(uint16), tensor(uint32), tensor(uint64), tensor(uint8) ): All Tensor types

• B in ( tensor(bool) ): Only bool

Differences

0	0	If conditional	If conditional
1	1
2	2	**Attributes**	**Attributes**
3	3
4	4	* **else_branch** (required):	* **else_branch** (required):
5	5	Graph to run if condition is false. Has N outputs: values you wish	Graph to run if condition is false. Has N outputs: values you wish
6	6	to be live-out to the enclosing scope. The number of outputs must	to be live-out to the enclosing scope. The number of outputs must
7	7	match the number of outputs in the then_branch.	match the number of outputs in the then_branch.
8	8	* **then_branch** (required):	* **then_branch** (required):
9	9	Graph to run if condition is true. Has N outputs: values you wish to	Graph to run if condition is true. Has N outputs: values you wish to
10	10	be live-out to the enclosing scope. The number of outputs must match	be live-out to the enclosing scope. The number of outputs must match
11	11	the number of outputs in the else_branch.	the number of outputs in the else_branch.
12	12
13	13	**Inputs**	**Inputs**
14	14
15	15	* **cond** (heterogeneous) - **B**:	* **cond** (heterogeneous) - **B**:
16	16	Condition for the if	Condition for the if
17	17
18	18	**Outputs**	**Outputs**
19	19
20	20	Between 1 and 2147483647 outputs.	Between 1 and 2147483647 outputs.
21	21
22	22	* **outputs** (variadic) - **V**:	* **outputs** (variadic) - **V**:
23	23	Values that are live-out to the enclosing scope. The return values	Values that are live-out to the enclosing scope. The return values
24	24	in the then_branch and else_branch must be of the same shape and	in the then_branch and else_branch must be of the same data
	25		type. The then_branch and else_branch may produce tensors with
	26		the same element type and different shapes. If corresponding outputs
	27		from the then-branch and the else-branch have static shapes S1 and
	28		S2, then the shape of the corresponding output variable of the if-
	29		node (if present) must be compatible with both S1 and S2 as it
	30		represents the union of both possible shapes.For example, if in a
25	31	same data type.	model file, the the first output of then_branch is typed float
	32		tensor with shape [2] and the first output of else_branch is
	33		another float tensor with shape [3], If's first output should have
	34		(a) no shape set, or (b) a shape of rank 1 with neither dim_value
	35		nor dim_param set, or (c) a shape of rank 1 with a unique
	36		dim_param. In contrast, the first output cannot have the shape [2]
	37		since [2] and [3] are not compatible.
26	38
27	39	**Type Constraints**	**Type Constraints**
28	40
29	41	* **V** in (	* **V** in (
30	42	tensor(bool),	tensor(bool),
31	43	tensor(complex128),	tensor(complex128),
32	44	tensor(complex64),	tensor(complex64),
33	45	tensor(double),	tensor(double),
34	46	tensor(float),	tensor(float),
35	47	tensor(float16),	tensor(float16),
36	48	tensor(int16),	tensor(int16),
37	49	tensor(int32),	tensor(int32),
38	50	tensor(int64),	tensor(int64),
39	51	tensor(int8),	tensor(int8),
40	52	tensor(string),	tensor(string),
41	53	tensor(uint16),	tensor(uint16),
42	54	tensor(uint32),	tensor(uint32),
43	55	tensor(uint64),	tensor(uint64),
44	56	tensor(uint8)	tensor(uint8)
45	57	):	):
46	58	All Tensor types	All Tensor types
47	59	* **B** in (	* **B** in (
48	60	tensor(bool)	tensor(bool)
49	61	):	):
50	62	Only bool	Only bool

If - 1

Version

• name: If (GitHub)

• domain: main

• since_version: 1

• function: False

• support_level: SupportType.COMMON

• shape inference: True

This version of the operator has been available since version 1.

Summary

If conditional

Attributes

• else_branch (required): Graph to run if condition is false. Has N outputs: values you wish to be live-out to the enclosing scope. The number of outputs must match the number of outputs in the then_branch.

• then_branch (required): Graph to run if condition is true. Has N outputs: values you wish to be live-out to the enclosing scope. The number of outputs must match the number of outputs in the else_branch.

Inputs

• cond (heterogeneous) - B: Condition for the if

Outputs

Between 1 and 2147483647 outputs.

• outputs (variadic) - V: Values that are live-out to the enclosing scope. The return values in the then_branch and else_branch must be of the same shape and same data type.

Type Constraints

• V in ( tensor(bool), tensor(complex128), tensor(complex64), tensor(double), tensor(float), tensor(float16), tensor(int16), tensor(int32), tensor(int64), tensor(int8), tensor(string), tensor(uint16), tensor(uint32), tensor(uint64), tensor(uint8) ): All Tensor types

• B in ( tensor(bool) ): Only bool

previous

Identity

next

InstanceNormalization