Coverage for mlprodict/onnxrt/ops_cpu/op_reduce_log_sum_exp.py: 96%

1# -*- encoding: utf-8 -*- 

2# pylint: disable=E0203,E1101,C0111 

3""" 

4@file 

5@brief Runtime operator. 

6""" 

7import numpy 

8from onnx.defs import onnx_opset_version 

9from ._op import OpRunReduceNumpy, OpRun 

10 

11 

12class ReduceLogSumExp_1(OpRunReduceNumpy): 

13 

14 atts = {'axes': [], 'keepdims': 1} 

15 

16 def __init__(self, onnx_node, desc=None, **options): 

17 OpRunReduceNumpy.__init__(self, onnx_node, desc=desc, 

18 expected_attributes=ReduceLogSumExp_1.atts, 

19 **options) 

20 

21 def _run(self, data, attributes=None, verbose=0, fLOG=None): # pylint: disable=W0221 

22 tax = tuple(self.axes) if self.axes else None 

23 data_max = data.copy() 

24 ind = numpy.isinf(data_max) 

25 data_max[ind] = -numpy.inf 

26 mx = data_max.max(axis=tax, keepdims=True) 

27 sub = numpy.subtract(data, mx) 

28 exp = numpy.exp(sub, out=sub) 

29 mxs = numpy.sum(exp, axis=tax, 

30 keepdims=True, 

31 dtype=data.dtype) 

32 res = numpy.log(mxs) + mx 

33 if not self.keepdims: 

34 res = numpy.squeeze(res, axis=tax) 

35 return (res, ) 

36 

37 

38class ReduceLogSumExp_18(OpRun): 

39 

40 atts = {'keepdims': 1, 'noop_with_empty_axes': 0} 

41 

42 def __init__(self, onnx_node, desc=None, **options): 

43 OpRun.__init__(self, onnx_node, desc=desc, 

44 expected_attributes=ReduceLogSumExp_18.atts, 

45 **options) 

46 

47 def _run(self, data, axes=None, attributes=None, verbose=0, fLOG=None): # pylint: disable=W0221 

48 if ((axes is None or len(axes.shape) == 0 or axes.shape[0] == 0) and 

49 self.noop_with_empty_axes): 

50 return (data, ) 

51 if ((axes is not None and len(axes.shape) > 0 and axes.shape[0] > 0) and 

52 not isinstance(axes, int)): 

53 if isinstance(axes, numpy.ndarray) and len(axes.shape) == 0: 

54 axes = int(axes) 

55 else: 

56 axes = tuple(axes.ravel().tolist()) if len(axes) > 0 else None 

57 try: 

58 tax = tuple(axes) if axes else None 

59 data_max = data.copy() 

60 ind = numpy.isinf(data_max) 

61 data_max[ind] = -numpy.inf 

62 mx = data_max.max(axis=tax, keepdims=True) 

63 sub = numpy.subtract(data, mx) 

64 exp = numpy.exp(sub, out=sub) 

65 mxs = numpy.sum(exp, axis=tax, 

66 keepdims=True, 

67 dtype=data.dtype) 

68 res = numpy.log(mxs) + mx 

69 if not self.keepdims: 

70 res = numpy.squeeze(res, axis=tax) 

71 return (res, ) 

72 except TypeError as e: # pragma: no cover 

73 raise TypeError( 

74 f"Unable to reduce shape {data.shape!r} with axes={axes!r}.") from e 

75 

76 

77if onnx_opset_version() >= 18: 

78 ReduceLogSumExp = ReduceLogSumExp_18 

79else: # pragma: no cover 

80 ReduceLogSumExp = ReduceLogSumExp_1