[ASCEND] add fc op (#4080)

* [Ascend] add fc op

* [Ascend] fix fc test

* [Ascend] fix code format

* [Ascend] modify the comment

lite/kernels/huawei_ascend_npu/bridges/CMakeLists.txt

@@ -16,6 +16,7 @@ lite_cc_library(subgraph_bridge_elementwise_ops_huawei_ascend_npu SRCS elementwi
 lite_cc_library(subgraph_bridge_batch_norm_op_huawei_ascend_npu SRCS batch_norm_op.cc DEPS ${huawei_ascend_npu_subgraph_bridge_deps})
 lite_cc_library(subgraph_bridge_softmax_op_huawei_ascend_npu SRCS softmax_op.cc DEPS ${huawei_ascend_npu_subgraph_bridge_deps})
 lite_cc_library(subgraph_bridge_dropout_op_huawei_ascend_npu SRCS dropout_op.cc DEPS ${huawei_ascend_npu_subgraph_bridge_deps})
+lite_cc_library(subgraph_bridge_fc_op_huawei_ascend_npu SRCS fc_op.cc DEPS ${huawei_ascend_npu_subgraph_bridge_deps})
 
 set(huawei_ascend_npu_subgraph_bridges
         subgraph_bridge_registry
@@ -30,4 +31,5 @@ set(huawei_ascend_npu_subgraph_bridges
         subgraph_bridge_batch_norm_op_huawei_ascend_npu
         subgraph_bridge_softmax_op_huawei_ascend_npu
         subgraph_bridge_dropout_op_huawei_ascend_npu
+        subgraph_bridge_fc_op_huawei_ascend_npu
         CACHE INTERNAL "huawei_ascend_npu_subgraph_bridges")

lite/kernels/huawei_ascend_npu/bridges/fc_op.cc

+// Copyright (c) 2020 PaddlePaddle Authors. All Rights Reserved.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+#include "lite/kernels/huawei_ascend_npu/bridges/graph.h"
+#include "lite/kernels/huawei_ascend_npu/bridges/utility.h"
+#include "lite/kernels/npu/bridges/registry.h"
+
+namespace paddle {
+namespace lite {
+namespace subgraph {
+namespace huawei_ascend_npu {
+
+int FCConverter(void* ctx, OpLite* op, KernelBase* kernel) {
+  CHECK(ctx != nullptr);
+  CHECK(op != nullptr);
+  auto graph = static_cast<Graph*>(ctx);
+  auto op_info = op->op_info();
+  auto op_type = op_info->Type();
+  auto scope = op->scope();
+  VLOG(3) << "[HUAWEI_ASCEND_NPU] Converting " + op_type + "...";
+
+  // Get input data nodes
+  auto input_name = op_info->Input("Input").front();
+  auto input = scope->FindTensor(input_name);
+  auto input_dims = input->dims();
+
+  auto w_name = op_info->Input("W").front();
+  auto w = scope->FindTensor(w_name);
+  auto w_dims = w->dims();
+  CHECK_EQ(w_dims.size(), 2UL);
+
+  auto out_name = op_info->Output("Out").front();
+  auto out = scope->FindTensor(out_name);
+  auto out_dims = out->dims();
+
+  int in_num_col_dims = op_info->GetAttr<int>("in_num_col_dims");
+  int m = input_dims.Slice(0, in_num_col_dims).production();
+  int k = input_dims.Slice(in_num_col_dims, input_dims.size()).production();
+  int n = w_dims[1];
+  CHECK_EQ(k * n, w_dims.production());
+
+  VLOG(3) << "[HUAWEI_ASCEND_NPU] input_dims = " << input_dims.repr()
+          << ", w_dims = " << w_dims.repr()
+          << ", in_num_col_dims = " << in_num_col_dims << ", m = " << m
+          << ", k = " << k << ", n = " << n;
+
+  // Create input node
+  std::shared_ptr<Node> input_node = nullptr;
+  if (graph->Has(input_name)) {
+    input_node = graph->Get(input_name);
+  } else {
+    input_node = graph->Add(input_name, *input);
+  }
+
+  // w_node, the ascend ddk will transpose the tensor of w
+  // if we set transpose attr to be true
+  auto w_node = graph->Add(w_name, *w);
+
+  // fc node
+  auto fc_node = graph->Add<ge::op::FullyConnection>(out_name);
+  auto fc_op = fc_node->data<ge::op::FullyConnection>();
+  fc_op->set_input_x(*input_node->data());
+  fc_op->set_input_w(*w_node->data());
+  fc_op->set_attr_num_output(n);
+  fc_op->set_attr_transpose(true);
+
+  INPUT_UPDATE(fc_op, x, input_node);
+  INPUT_UPDATE(fc_op, w, w_node);
+  OUTPUT_UPDATE(fc_op, y, fc_node);
+
+  if (HasInputArg(op_info, scope, "Bias")) {
+    std::shared_ptr<Node> bias_node = nullptr;
+    auto bias_name = op_info->Input("Bias").front();
+    if (graph->Has(bias_name)) {
+      bias_node = graph->Get(bias_name);
+    } else {
+      auto bias = scope->FindTensor(bias_name);
+      auto bias_dims = bias->dims();
+      CHECK_EQ(bias_dims.production(), n);
+
+      VLOG(3) << "[HUAWEI_ASCEND_NPU] bias_dims = " << bias_dims.repr();
+      bias_node = graph->Add(bias_name, *bias, {1, n, 1, 1});
+    }
+    fc_op->set_input_b(*bias_node->data());
+    INPUT_UPDATE(fc_op, b, bias_node);
+  }
+
+  return REBUILD_WHEN_SHAPE_CHANGED;
+}
+
+}  // namespace huawei_ascend_npu
+}  // namespace subgraph
+}  // namespace lite
+}  // namespace paddle
+
+REGISTER_SUBGRAPH_BRIDGE(
+    fc,
+    kHuaweiAscendNPU,
+    paddle::lite::subgraph::huawei_ascend_npu::FCConverter);

lite/kernels/huawei_ascend_npu/bridges/paddle_use_bridges.h

@@ -41,3 +41,4 @@ USE_SUBGRAPH_BRIDGE(fusion_elementwise_max_activation, kHuaweiAscendNPU);
 USE_SUBGRAPH_BRIDGE(batch_norm, kHuaweiAscendNPU);
 USE_SUBGRAPH_BRIDGE(softmax, kHuaweiAscendNPU);
 USE_SUBGRAPH_BRIDGE(dropout, kHuaweiAscendNPU);
+USE_SUBGRAPH_BRIDGE(fc, kHuaweiAscendNPU);

lite/tests/kernels/fc_compute_test.cc

@@ -111,6 +111,11 @@ class FcOPTest : public arena::TestCase {
     }
     auto out_data = out->mutable_data<float>();
 
+    // must init out_data to be 0 firstly
+    for (int i = 0; i < out_dim.production(); i++) {
+      out_data[i] = 0;
+    }
+
     int m = x->dims().count(0, in_num_col_dims_);
     CHECK_EQ(wdims_[0], x->dims().count(in_num_col_dims_, x->dims().size()));
     int k = wdims_[0];
@@ -279,6 +284,9 @@ TEST(FcOP, precision) {
   abs_error = 1e-4;
 #elif defined(LITE_WITH_ARM)
   place = TARGET(kARM);
+#elif defined(LITE_WITH_HUAWEI_ASCEND_NPU)
+  place = TARGET(kHuaweiAscendNPU);
+  abs_error = 1e-2;  // precision_mode default is force_fp16
 #else
   return;
 #endif