add support for tf.math.is_finite #936
Conversation
guschmue
commented
May 26, 2020
guschmue
commented
- add support for tf.math.is_finite
- expanddims() does not need to depend on shape in opset-11
- remove Inverse() from onnx_opset/math.py since this is in ms-domain
| @@ -3224,6 +3224,14 @@ def func(x, y): | |||
| tf.math.greater_equal(x, y, name=_TFOUTPUT1) | |||
| self._run_test_case(func, [_OUTPUT, _OUTPUT1], {_INPUT: x_val, _INPUT1: y_val}) | |||
|
|
|||
| @check_opset_min_version(10) | |||
| def test_is_finite(self): | |||
| x_val = np.array([5.0, 4.8, 6.8, np.inf, np.nan], dtype=np.float32) | |||
There was a problem hiding this comment.
Sometimes data has negative-infinity, so might be something to check.
| node.set_attr("axes", [dim]) | ||
| ctx.remove_input(node, node.input[1]) | ||
| return | ||
| raise ValueError("non-const dim is not supported") | ||
|
|
There was a problem hiding this comment.
Is ver_11 same as ver_7, but without the negative axis check? ... if yes, the should we just call ver_7 (less code)? If no, then ignore this comment :)
There was a problem hiding this comment.
opset-11 added support for the negative axes.
|
This pull request introduces 1 alert when merging a5373c0 into f4f2f04 - view on LGTM.com new alerts:
|
|
This pull request introduces 1 alert when merging f31ff36 into f4f2f04 - view on LGTM.com new alerts:
|
|
This pull request introduces 1 alert when merging 2d5760d into f4f2f04 - view on LGTM.com new alerts:
|
|
This pull request introduces 1 alert when merging f2bbe8d into f4f2f04 - view on LGTM.com new alerts:
|
|
This pull request introduces 1 alert when merging c891780 into f4f2f04 - view on LGTM.com new alerts:
|
