STT-tensorflow/tensorflow/compiler/xla/tests/isolated_convolution.hlo
A. Unique TensorFlower 029109b4e1 [XLA] Make hlo_test_base support running and comparing an hlo module.
Also,
- Add examples that shows how to use the hlo_test_base as to create text/file-based testcases.
- Change the behavior of GetDefaultPlatform: when only one platform is present, returns that one; when two platforms are present and one of them is the interpreter, returns the other one. This is because some tests included both hlo_test_base and client_library_test_base, but now the hlo_test_base is linked with interpreter by default, which makes client_library_test_base fail getting the default platform.

PiperOrigin-RevId: 178309022
2017-12-07 15:57:22 -08:00

9 lines
459 B
Plaintext

HloModule convolution.167:
ENTRY %convolution.167 (parameter.0: f32[16,28,28,128], parameter.1: f32[3,3,128,128]) -> f32[16,28,28,128] {
%parameter.0 = f32[16,28,28,128]{3,0,2,1} parameter(0)
%parameter.1 = f32[3,3,128,128]{3,2,1,0} parameter(1)
ROOT %convolution.167 = f32[16,28,28,128]{3,0,2,1} convolution(f32[16,28,28,128]{3,0,2,1} %parameter.0, f32[3,3,128,128]{3,2,1,0} %parameter.1), window={size=3x3 pad=1_1x1_1}, dim_labels=b01f_01oi->b01f
}