File: type_double.py

package info (click to toggle)
onnxscript 0.2.0%2Bdfsg-1
  • links: PTS, VCS
  • area: main
  • in suites: experimental
  • size: 12,384 kB
  • sloc: python: 75,957; sh: 41; makefile: 6
file content (73 lines) | stat: -rw-r--r-- 2,028 bytes parent folder | download
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
# Copyright (c) Microsoft Corporation.
# Licensed under the MIT License.

from onnx import TensorProto
from onnx.helper import make_tensor

from onnxscript import script
from onnxscript.onnx_opset import opset15 as op
from onnxscript.onnx_types import DOUBLE, INT64


@script()
def sumprod_typed(x: DOUBLE["N"], N: INT64) -> DOUBLE["N"]:  # noqa: F821
    weight = op.Constant(value=make_tensor("zero", TensorProto.DOUBLE, [1], [0]))
    weighted_sum = op.Identity(weight)
    for i in range(N):
        b: DOUBLE[...] = op.Cast(i, to=11)
        a: DOUBLE[...] = x * b
        weighted_sum = weighted_sum + a
        weight = weight + b
    return op.Div(weighted_sum, weight)


@script()
def sumprod(x: DOUBLE["N"], N: INT64) -> DOUBLE["N"]:  # noqa: F821
    weight = op.Constant(value=make_tensor("zero", TensorProto.DOUBLE, [1], [0]))
    weighted_sum = op.Identity(weight)
    for i in range(N):
        b = op.Cast(i, to=11)
        a = x * b
        weighted_sum = weighted_sum + a
        weight = weight + b
    return op.Div(weighted_sum, weight)


@script()
def notype_abs_subgraph(A):
    zero = op.Constant(value=make_tensor("zero", TensorProto.FLOAT, [1], [0]))
    if op.Cast(op.ReduceSum(A), to=1) > zero:
        B = op.Identity(A)
    else:
        B = op.Neg(A)
    return B


@script()
def double_abs_subgraph(A: DOUBLE["N"]) -> DOUBLE["N"]:
    zero = op.Constant(value=make_tensor("zero", TensorProto.DOUBLE, [1], [0]))
    if op.ReduceSum(A) > zero:
        B: DOUBLE["N"] = op.Identity(A)
    else:
        B: DOUBLE["N"] = op.Neg(A)
    return B


@script()
def double_abs(A: DOUBLE["N"]) -> DOUBLE["N"]:
    return op.Abs(A)


@script()
def double_cast(A: INT64["N"]) -> DOUBLE["N"]:
    return op.Cast(A, to=11)


# Does not work.
# @script()
# def double_abs_subgraph_direct_return(A: DOUBLE["N"]) -> DOUBLE["N"]:
#     zero = op.Constant(value=make_tensor('zero', TensorProto.INT64, [1], [0]))
#     if op.ReduceSum(A) > zero:
#         return op.Identity(A)
#     else:
#         return op.Neg(A)