[original author: mrnikwaws] Neuron glu operator support#5466
[original author: mrnikwaws] Neuron glu operator support#5466aws-kingrj wants to merge 11 commits intopytorch:masterfrom
Conversation
aws-kingrj
commented
Aug 18, 2023
- Adding glu operator support and cpp test
- original author mrnikaws
- tested internally with AWS Neuron
| return xla::ShapeUtil::MakeShape(input_shape.element_type(), output_sizes); | ||
| } | ||
|
|
||
| xla::Shape GtScalarOutputShape(const torch::lazy::Value& self, |
There was a problem hiding this comment.
can you add a cpp test similar to
xla/test/cpp/test_aten_xla_tensor_5.cpp
Line 209 in 028df4d
There was a problem hiding this comment.
There should be a cpp test already added or another one?
There was a problem hiding this comment.
you should add another one similar to Gelu.
There was a problem hiding this comment.
Hey @JackCaoG can you help me understand what is missing?
The test in https://github.com/pytorch/xla/pull/5466/files#diff-0c11cc2fbeb0cfbc8fe223f9ac3df0904dcbb1277946a55a56cb102642f4faefR1322-R1342 was intended to be similar to the Gelu one, it tests a variety of input permutations across both input variables and uses the same checks.
There was a problem hiding this comment.
You can build it with BUILD_CPP_TESTS=1 and run CPP test with ./test/cpp/run_test.sh
|
@aws-kingrj can you rebase this pr? I should have fixed the build timeout. |