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

Add Tactivation attribute for QuantizedMatmul and QuantizedMatmulMinMax to explicitly notify which quantizedtype output would be produced by following activation function.

Browse Source 
Change: 142073731
This commit is contained in:
A. Unique TensorFlower 2016-12-14 15:39:14 -08:00 committed by TensorFlower Gardener
parent a7cd5f671e
commit 811629aed4
1 changed files with 3 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

3
tensorflow/core/ops/math_ops.cc
View File

@ -2245,6 +2245,7 @@ REGISTER_OP("QuantizedMatMul")
.Attr("Toutput: quantizedtype = DT_QINT32") .Attr("Toutput: quantizedtype = DT_QINT32")
.Attr("transpose_a: bool = false") .Attr("transpose_a: bool = false")
.Attr("transpose_b: bool = false") .Attr("transpose_b: bool = false")
.Attr("Tactivation: quantizedtype = DT_QUINT8")
.SetShapeFn([](InferenceContext* c) { .SetShapeFn([](InferenceContext* c) {
TF_RETURN_IF_ERROR(shape_inference::MatMulShape(c)); TF_RETURN_IF_ERROR(shape_inference::MatMulShape(c));
ShapeHandle unused; ShapeHandle unused;
@ -2275,6 +2276,8 @@ min_b: The float value that the lowest quantized `b` value represents.
max_b: The float value that the highest quantized `b` value represents. max_b: The float value that the highest quantized `b` value represents.
min_out: The float value that the lowest quantized output value represents. min_out: The float value that the lowest quantized output value represents.
max_out: The float value that the highest quantized output value represents. max_out: The float value that the highest quantized output value represents.
Tactivation: The type of output produced by activation function
following this operation.
)doc"); )doc");