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

a bit more

Browse Source
This commit is contained in:
Dmitry Volodin 2020-10-29 17:44:22 +03:00
parent 4df96ca10a
commit 6c5e25786e
1 changed files with 5 additions and 5 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

10
tensorflow/core/common_runtime/mkl_layout_pass.cc
View File

@ -1571,7 +1571,7 @@ class MklLayoutRewritePass : public GraphOptimizationPass {
}
// If the depth_radius of LRN is not 2, then MKL DNN takes unoptimized
// path. The unoptimized path is slow. Thus we dont rewrite the node
// path. The unoptimized path is slow. Thus we don't rewrite the node
// and use default Eigen. But for depth_radius=2, MKL DNN optimized
// path is taken, i.e., eigen node is rewritten by MKl DNN node.
static bool LrnRewrite(const Node* n) {
@ -1969,7 +1969,7 @@ class MklLayoutRewritePass : public GraphOptimizationPass {
// Helper function used by FixMklMetaDataEdges. Fixes the metadata edge
// pointed by 'e_metadata' corresponding to the data edge 'e_data' in graph
// 'g'. Returns true is fixup was done; otherwise, it returns false.
// 'g'. Returns true if fixup was done; otherwise, it returns false.
bool FixMklMetaDataEdgeIfNeeded(std::unique_ptr<Graph>* g, const Edge* e_data,
const Edge* e_metadata);
@ -3460,7 +3460,7 @@ Status MklLayoutRewritePass::MergeConv2DBackpropFilterWithBiasAddGrad(
// This is because BackpropFilterWithBias is going to emit bias output also.
NodeBuilder nb(fltr->name(), csinfo_.conv2d_grad_filter_with_bias);
// Since Conv2DBackpropFilterWithBias has same number of inputs as
// Conv2DBackpropFilter, we can just copy input edges directly. We dont need
// Conv2DBackpropFilter, we can just copy input edges directly. We don't need
// to copy any data input of BiasAddGrad because that input also goes to
// Conv2DBackpropFilter.
const int fltr_ins = fltr->num_inputs();
@ -3795,7 +3795,7 @@ Status MklLayoutRewritePass::RewriteNode(std::unique_ptr<Graph>* g,
return ret_status;
}
// TODO(mdfaijul): Is there any other elegent way to check for quantized ops
// TODO(mdfaijul): Is there any other elegant way to check for quantized ops
// having attributes other than "T"?
// Current implementation reflects only QuantizedConv2D and its fused Ops.
const MklLayoutRewritePass::RewriteInfo*
@ -3983,7 +3983,7 @@ MklLayoutRewritePass::CheckForNodeFusion(Node* a) const {
for (auto fi = finfo_.begin(); fi != finfo_.end(); ++fi) {
//
// Make sure node "a" and its succeding nodes (b, c ...), match the pattern
// Make sure node "a" and its succeeding nodes (b, c ...), match the pattern
// defined in fusion info (ops[0], ops[1], ...),
// a.k.a. "a->b->c" matches "op1->op2->op3"
//