experiments/STT-tensorflow
1
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

Update operation summary for TF _FusedConv2D.

Browse Source 
PiperOrigin-RevId: 315733508
Change-Id: I55221ee946117e6bdf68c2ad22158e12f7eba077
This commit is contained in:
Lucy Fox 2020-06-10 11:39:17 -07:00 committed by TensorFlower Gardener
parent 4660a50f07
commit 0fecf6f89f
2 changed files with 37 additions and 4 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

20
tensorflow/compiler/mlir/tensorflow/ir/tf_generated_ops.td
View File

@ -10841,11 +10841,27 @@ def TF_ZerosLikeOp : TF_Op<"ZerosLike", [NoSideEffect, SameOperandsAndResultType
def TF__FusedConv2DOp : TF_Op<"_FusedConv2D", [NoSideEffect]> {
let summary = [{
*NOTE*: Do not invoke this operator directly in Python. Grappler is
Performs a convolution followed by a specified series of operations.
}];
let description = [{
expected to create these operators.
The inputs to the convolution are `input` and `filter`. The series of operations
that follows is specified by the `fused_ops` attribute, which is a list of TF op
names specified as strings (e.g. "Relu"). They are performed in order, where the
(first) input to each op is the output of the preceding op. The first input and
the output of each fused_op must be of type T.
Currently supported fused_op combinations are: [X] and [X,A], where X is one of
{"BiasAdd","FusedBatchNorm"} and A is one of {"Elu","Relu","Relu6"}.
* The first input to op X is the Conv2D result, and the additional input(s) to X
are specified by `args`.
* If there is an op A specified, the output of op X is the input to op A, and op
A produces the _FusedConv2D output. Otherwise, op X produces the _FusedConv2D
output.
*NOTE*: Do not invoke this operator directly in Python. Grappler is expected to
create these operators.
}];
let arguments = (ins

21
tensorflow/core/ops/nn_ops.cc
View File

@ -405,8 +405,25 @@ REGISTER_OP("_FusedConv2D")
// ---------------------------------------------------------------------- //
.SetShapeFn(shape_inference::Conv2DShapeWithExplicitPadding)
.Doc(R"doc(
*NOTE*: Do not invoke this operator directly in Python. Grappler is
expected to create these operators.
Performs a convolution followed by a specified series of operations.
The inputs to the convolution are `input` and `filter`. The series of operations
that follows is specified by the `fused_ops` attribute, which is a list of TF op
names specified as strings (e.g. "Relu"). They are performed in order, where the
(first) input to each op is the output of the preceding op. The first input and
the output of each fused_op must be of type T.
Currently supported fused_op combinations are: [X] and [X,A], where X is one of
{"BiasAdd","FusedBatchNorm"} and A is one of {"Elu","Relu","Relu6"}.
* The first input to op X is the Conv2D result, and the additional input(s) to X
are specified by `args`.
* If there is an op A specified, the output of op X is the input to op A, and op
A produces the _FusedConv2D output. Otherwise, op X produces the _FusedConv2D
output.
*NOTE*: Do not invoke this operator directly in Python. Grappler is expected to
create these operators.
)doc");
namespace {