Skip to content

Commit

Permalink
update
Browse files Browse the repository at this point in the history
  • Loading branch information
@jiweibo
jiweibo committed Jul 6, 2022
1 parent 6326314 commit 8830b02
Showing 1 changed file with 0 additions and 124 deletions.
124 changes: 0 additions & 124 deletions paddle/fluid/inference/analysis/ir_passes/tensorrt_subgraph_pass.cc
View file
Expand Up @@ -49,69 +49,6 @@ bool IsFloat(framework::proto::VarType::Type t) {
return false;
}

// // if in mixed model precison and tensorrt_engine's input is bf16 dtype (trt
// not
// // support), we need to add a cast op before engine building phase.
// void InputProcess(
// framework::ir::Graph *graph,
// const std::unordered_set<framework::ir::Node *> &trt_inputs,
// framework::BlockDesc *block_desc,
// phi::Backend backend,
// phi::DataType precision,
// std::set<std::string> *input_names,
// std::set<std::string> *input_names_with_id,
// std::unordered_map<std::string, framework::ir::Node *> *names_changed) {
// // TODO(inference): Support other mixed precision
// std::unordered_set<phi::DataType> trt_not_support_types{
// phi::DataType::BFLOAT16,
// };
// if (!trt_not_support_types.count(precision)) return;

// // framework::BlockDesc *block_desc{nullptr};
// int suffix = 0;
// std::unordered_map<framework::ir::Node *, framework::ir::Node *>
// var_to_cast_op_map;

// std::unordered_set<framework::proto::VarType::Type>
// trt_not_support_var_dtypes{
// framework::proto::VarType::BF16,
// };

// // if trt_engine's input dtype is bf16, we should add cast op.
// for (auto *op_node : framework::ir::TopologySortOperations(*graph)) {
// if (!op_node->IsOp()) continue;
// auto op_type = op_node->Op()->Type();
// // if (op_type == "feed") block_desc = op_node->Op()->Block();
// if (op_type != "tensorrt_engine") continue;

// auto op_inputs = op_node->inputs;

// for (auto *in_node : op_inputs) {
// auto *in_var = in_node->Var();
// if (trt_inputs.count(in_node)) {
// if (trt_not_support_var_dtypes.count(in_var->GetDataType())) {
// AddCastOp(graph,
// in_node,
// op_node,
// in_var->GetDataType(),
// framework::proto::VarType::FP32,
// &suffix,
// block_desc,
// &var_to_cast_op_map);
// input_names->erase(in_node->Name());
// input_names_with_id->erase(in_node->Name() +
// std::to_string(in_node->id()));
// input_names->insert(var_to_cast_op_map[in_node]->Name());
// input_names_with_id->insert(
// var_to_cast_op_map[in_node]->Name() +
// std::to_string(var_to_cast_op_map[in_node]->id()));
// (*names_changed)[in_node->Name()] = var_to_cast_op_map[in_node];
// }
// }
// }
// }
// }

// if in mixed model precision, we should make all tensorrt_engine's output
// floats dtype to float32 dtype.
void OutputProcess(framework::ir::Graph *graph,
Expand Down Expand Up @@ -340,67 +277,6 @@ void TensorRtSubgraphPass::CreateTensorRTOp(
auto mixed_black_list =
Get<std::unordered_set<std::string>>("mixed_black_list");

// // paddle-trt support paddle-inference's mixed precision.
// std::unordered_set<Node *> trt_inputs;
// for (auto *x : node->inputs) {
// if (x->Var()->Persistable()) continue;
// trt_inputs.insert(x);
// }
// std::unordered_map<std::string, framework::ir::Node *> names_changed;
// InputProcess(graph,
// trt_inputs,
// &block_desc,
// phi::Backend::GPU,
// model_precision,
// &input_names,
// &input_names_with_id,
// &names_changed);

// for (size_t index = 0; index < block_desc.OpSize(); ++index) {
// framework::proto::OpDesc *op = block_desc.Op(index)->Proto();
// framework::OpDesc op_desc(*op, nullptr);

// // rename for the input variables of op inside subgraph
// for (int i = 0; i < op->inputs_size(); i++) {
// auto *in_var = op->mutable_inputs(i);
// std::vector<std::string> replaced_names;
// for (int k = 0; k < in_var->arguments_size(); k++) {
// const std::string origin_name = in_var->arguments(k);
// if (names_changed.count(origin_name)) {
// replaced_names.push_back(names_changed[origin_name]->Name());
// } else {
// replaced_names.push_back(origin_name);
// }
// }
// in_var->clear_arguments();
// for (size_t k = 0; k < replaced_names.size(); k++) {
// in_var->add_arguments(replaced_names[k]);
// }
// }
// }

// for (auto *node : subgraph) {
// if (node->NodeType() == Node::Type::kOperation) {
// auto inputs = node->inputs;
// for (auto in : inputs) {
// if (names_changed.count(in->Name())) {
// node->Op()->RenameInput(in->Name(),
// names_changed[in->Name()]->Name());

// auto it = std::find_if(node->inputs.begin(),
// node->inputs.end(),
// [&](const framework::ir::Node *node) {
// return node->Name() == in->Name();
// });
// if (it != node->inputs.end()) {
// node->inputs.erase(it);
// node->inputs.push_back(names_changed[in->Name()]);
// }
// }
// }
// }
// }

std::set<std::string> output_names;
std::set<std::string> output_names_with_id;
std::map<std::string, int> origin_name_output_dims;
Expand Down

1 comment on commit 8830b02

@paddle-bot-old
Copy link

@paddle-bot-old paddle-bot-old bot commented on 8830b02 Jul 6, 2022 •
edited

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

🕵️ CI failures summary

🔍 PR: #44057 Commit ID: 8830b02 contains failed CI.

🔹 Failed: PR-CI-ROCM-Compile

Unknown Failed
Unknown Failed

🔹 Failed: PR-CI-APPROVAL

Unknown Failed
Unknown Failed

🔹 Failed: PR-CI-Static-Check

Unknown Failed
Unknown Failed

🔹 Failed: PR-CI-Inference

Unknown Failed
Unknown Failed

🔹 Failed: PR-CI-Coverage

Unknown Failed
Unknown Failed

🔹 Failed: PR-CI-Windows-Inference

Unknown Failed
Unknown Failed

Please sign in to comment.