Support bool inputs for MatrixBandPart #2269
Conversation
Signed-off-by: Alexander Gerstenberger <agerstenberger@apptek.com>
Signed-off-by: Alexander Gerstenberger <agerstenberger@apptek.com>
Signed-off-by: Alexander Gerstenberger <agerstenberger@apptek.com>
Signed-off-by: Alexander Gerstenberger <agerstenberger@apptek.com>
Signed-off-by: Alexander Gerstenberger <agerstenberger@apptek.com>
Signed-off-by: Alexander Gerstenberger <agerstenberger@apptek.com>
Signed-off-by: Alexander Gerstenberger <agerstenberger@apptek.com>
|
Added a test case |
tests/test_backend.py
Outdated
| @@ -3926,6 +3926,15 @@ def func(input_x, low, high): | |||
| return tf.identity(res, name=_TFOUTPUT) | |||
| self._run_test_case(func, [_OUTPUT], {_INPUT: input_val, _INPUT1: low_val, _INPUT2: high_val}) | |||
|
|
|||
| @check_opset_min_version(11, "CumSum") | |||
There was a problem hiding this comment.
Why do we need to add a limitation for CumSum op for this test?
There was a problem hiding this comment.
i copied the function from above and then made edits (the other test don't need this too then? check opset 7 would be sufficient for MatrixBandPart op?). I removed the limitation for the test now.
There was a problem hiding this comment.
It depends on if ONNX op CumSum is required during the conversion of a tf op. If yes, then we need to limit min opset version to 11 because ONNX supports CumSum op since 11. By adding this, the test case will be ignored if the test suite is running with an ONNX opset lower than 11.
There was a problem hiding this comment.
MatrixBandPart is a tf op. It doesn't have a requirement for ONNX opset version. The ONNX opset version requirement depends on all ONNX ops we use during the conversion of a TF op.
Signed-off-by: Alexander Gerstenberger <agerstenberger@apptek.com>
Fixes #2268
Please let me know if this is something you want to support.
Should there also be a test case?