Skip to content
This repository was archived by the owner on Jul 1, 2024. It is now read-only.

Fix const op translation #332

Merged
cavusmustafa merged 3 commits into from
Jul 11, 2022
Merged

Fix const op translation #332

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
fa86b0b
Fix copying values from const nodes, bug caught in Windows
ck-intel Jul 11, 2022
dac6d4d
Update ocm submodule, test cases list updated for Myriad
ck-intel Jul 11, 2022
3680db9
Remove unused commented code
ck-intel Jul 11, 2022
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
2 changes: 1 addition & 1 deletion ocm
Open in desktop
Submodule ocm updated 1 files
+76 −1 test/python/invalid_tests_list_MYRIAD.txt
9 changes: 0 additions & 9 deletions openvino_tensorflow/mark_for_clustering.cc
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -105,18 +105,11 @@ const std::map<std::string, SetAttributesFunction>& GetAttributeSetters() {
set_attributes_map["Conv2DBackpropInput"] = SetStaticInputs({0});
set_attributes_map["CropAndResize"] = SetStaticInputs({1, 2, 3});
set_attributes_map["ExpandDims"] = SetStaticInputs({1});
// set_attributes_map["Fill"] = SetStaticInputs({0});
set_attributes_map["GatherV2"] = SetStaticInputs({2});
set_attributes_map["Max"] = SetStaticInputs({1});
set_attributes_map["Mean"] = SetStaticInputs({1});
set_attributes_map["Min"] = SetStaticInputs({1});
set_attributes_map["MirrorPad"] = SetStaticInputs({1});
// set_attributes_map["NonMaxSuppression"] = SetStaticInputs({2});
// set_attributes_map["NonMaxSuppressionV2"] = SetStaticInputs({2});
// set_attributes_map["NonMaxSuppressionV3"] = SetStaticInputs({2});
// set_attributes_map["NonMaxSuppressionV4"] = SetStaticInputs({2});
// set_attributes_map["NonMaxSuppressionV5"] = SetStaticInputs({2});
// set_attributes_map["OneHot"] = SetStaticInputs({1});
set_attributes_map["Pad"] = SetStaticInputs({1});
set_attributes_map["PadV2"] = SetStaticInputs({1});
set_attributes_map["Prod"] = SetStaticInputs({1});
Expand All @@ -128,9 +121,7 @@ const std::map<std::string, SetAttributesFunction>& GetAttributeSetters() {
set_attributes_map["SplitV"] = SetStaticInputs({1, 2});
set_attributes_map["StridedSlice"] = SetStaticInputs({1, 2, 3});
set_attributes_map["Sum"] = SetStaticInputs({1});
// set_attributes_map["TopKV2"] = SetStaticInputs({1});
set_attributes_map["Tile"] = SetStaticInputs({1});
// set_attributes_map["Range"] = SetStaticInputs({0, 1, 2});
initialized = true;
}
return set_attributes_map;
Expand Down
111 changes: 70 additions & 41 deletions openvino_tensorflow/ovtf_builder.cc
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -398,54 +398,84 @@ static Status ValuesFromConstNode(const NodeDef& node,
}
values->resize(n_elements);

auto val_lastsaved = (T)0; // cast
// Extract dtype and size of the values provided in proto
auto& tensor = node.attr().at("value").tensor();
auto dt = node.attr().at("dtype").type();
int64 val_size = 0;
switch (dt) {
// TODO(amprocte/NGRAPH-2502): there are more element types to support
// here
case DT_INT32:
val_size = tensor.int_val_size();
break;
case DT_INT64:
val_size = tensor.int64_val_size();
break;
case DT_FLOAT:
val_size = tensor.float_val_size();
break;
case DT_BOOL:
val_size = tensor.bool_val_size();
break;
case DT_DOUBLE:
val_size = tensor.double_val_size();
break;
default:
OVTF_VLOG(0) << "Const node has empty tensor and we don't know how to "
"handle this element type";
OVTF_VLOG(0) << node.DebugString();
OVTF_VLOG(0) << shape.DebugString();
return errors::Unimplemented("Encountered unknown element type ",
DataType_Name(dt), " on an empty tensor");
}

auto val_lastsaved = (T)0; // cast
for (auto i = 0; i < n_elements; i++) {
auto& tensor = node.attr().at("value").tensor();
auto dt = node.attr().at("dtype").type();
int64 val_size = 0;
auto val_i = (T)0; // cast
switch (dt) {
// TODO(amprocte/NGRAPH-2502): there are more element types to support
// here
case DT_INT32:
val_size = tensor.int_val_size();
if (val_size > 0) val_i = tensor.int_val()[i];
break;
case DT_INT64:
val_size = tensor.int64_val_size();
if (val_size > 0) val_i = tensor.int64_val()[i];
break;
case DT_FLOAT:
val_size = tensor.float_val_size();
if (val_size > 0) val_i = tensor.float_val()[i];
break;
case DT_BOOL:
val_size = tensor.bool_val_size();
if (val_size > 0) val_i = tensor.bool_val()[i];
break;
case DT_DOUBLE:
val_size = tensor.double_val_size();
if (val_size > 0) val_i = tensor.double_val()[i];
break;
default:
OVTF_VLOG(0)
<< "Const node has empty tensor and we don't know how to "
"handle this element type";
OVTF_VLOG(0) << node.DebugString();
OVTF_VLOG(0) << shape.DebugString();
return errors::Unimplemented("Encountered unknown element type ",
DataType_Name(dt),
" on an empty tensor");
}
// Default value set to 0 and typecasted to the dtype of the const op
auto val_i = (T)0;
// If no values are specified for the const node, fill all the values with
// default value or return error based on the TF version
if (val_size == 0) {
#if (TF_MAJOR_VERSION > 1 && TF_MINOR_VERSION >= 7)
val_i = 0;
(*values)[i] = val_i;
#else
return errors::InvalidArgument("Empty values vector");
#endif
continue;
}

} else if (i < val_size) {
// If the values are same for all the elements or repeating after certain
// index, then copy the single value or the last occured value for all the
// remaining indices
if (i < val_size) {
switch (dt) {
// TODO(amprocte/NGRAPH-2502): there are more element types to support
// here
case DT_INT32:
val_i = tensor.int_val()[i];
break;
case DT_INT64:
val_i = tensor.int64_val()[i];
break;
case DT_FLOAT:
val_i = tensor.float_val()[i];
break;
case DT_BOOL:
val_i = tensor.bool_val()[i];
break;
case DT_DOUBLE:
val_i = tensor.double_val()[i];
break;
default:
OVTF_VLOG(0)
<< "Const node has empty tensor and we don't know how to "
"handle this element type";
OVTF_VLOG(0) << node.DebugString();
OVTF_VLOG(0) << shape.DebugString();
return errors::Unimplemented("Encountered unknown element type ",
DataType_Name(dt),
" on an empty tensor");
}
(*values)[i] = val_i;
val_lastsaved = val_i;
} else {
Expand Down Expand Up @@ -4412,7 +4442,6 @@ Status Builder::TranslateGraph(
OVTF_VLOG(5) << "Not using Fused translation for "
<< "CTCGreedyDecoder and SparseToDense Ops";
}
//

const function<Status(const Node*, const std::vector<const Tensor*>&,
Builder::OpMap&)>* op_fun;
Expand Down