[arm] add elu activation op (#3978)

* add elu act

* fix elu act not find error, test=develop

* fix format.test=develop

* fix int8 model opt error in conv+conv fusion, test=develop

* fix format. test=develop

* test=develop

lite/api/paddle_place.cc

@@ -55,7 +55,8 @@ const std::string& ActivationTypeToStr(ActivationType act) {
                                            "Tanh",
                                            "Swish",
                                            "Exp",
-                                           "ThresholdedRelu"};
+                                           "ThresholdedRelu",
+                                           "Elu"};
   auto x = static_cast<int>(act);
   CHECK_LT(x, static_cast<int>(ActivationType::NUM));
   return act2string[x];

lite/api/paddle_place.h

@@ -108,7 +108,8 @@ enum class ActivationType : int {
   kHardSwish = 10,
   kReciprocal = 11,
   kThresholdedRelu = 12,
-  NUM = 13,
+  kElu = 13,
+  NUM = 14,
 };
 
 static size_t PrecisionTypeLength(PrecisionType type) {

lite/backends/arm/math/activation.cc

@@ -763,6 +763,91 @@ void act_thresholded_relu<float>(
   }
 }
 
+// elu: out = max(0,x) + min(0, alpha *(exp(x) - 1)
+template <>
+void act_elu<float>(
+    const float* din, float* dout, int size, float alpha, int threads) {
+  int nums_per_thread = size / threads;
+  int thread_remain = size % threads;
+  int neon_loop_cnt_dim16 = nums_per_thread >> 4;
+  int neon_loop_remain_dim16 = nums_per_thread & 15;
+  float32x4_t valpha = vdupq_n_f32(alpha);
+  float32x4_t vzero = vdupq_n_f32(0.f);
+  float32x4_t vone = vdupq_n_f32(1.f);
+  int cnt = neon_loop_remain_dim16 >> 2;
+  int remain = neon_loop_remain_dim16 & 3;
+#pragma omp parallel for
+  for (int i = 0; i < threads; i++) {
+    const float* ptr_in_thread = din + i * nums_per_thread;
+    float* ptr_out_thread = dout + i * nums_per_thread;
+    for (int k = 0; k < neon_loop_cnt_dim16; ++k) {
+      float32x4_t va = vld1q_f32(ptr_in_thread);
+      float32x4_t vb = vld1q_f32(ptr_in_thread + 4);
+      float32x4_t vc = vld1q_f32(ptr_in_thread + 8);
+      float32x4_t vd = vld1q_f32(ptr_in_thread + 12);
+      float32x4_t va_exp = exp_ps(va);
+      float32x4_t va_max = vmaxq_f32(va, vzero);
+      float32x4_t vb_exp = exp_ps(vb);
+      float32x4_t vb_max = vmaxq_f32(vb, vzero);
+      float32x4_t vc_exp = exp_ps(vc);
+      float32x4_t vc_max = vmaxq_f32(vc, vzero);
+      float32x4_t vd_exp = exp_ps(vd);
+      float32x4_t vd_max = vmaxq_f32(vd, vzero);
+      float32x4_t va_sub = vsubq_f32(va_exp, vone);
+      float32x4_t vb_sub = vsubq_f32(vb_exp, vone);
+      float32x4_t vc_sub = vsubq_f32(vc_exp, vone);
+      float32x4_t vd_sub = vsubq_f32(vd_exp, vone);
+      va_sub = vmulq_f32(va_sub, valpha);
+      vb_sub = vmulq_f32(vb_sub, valpha);
+      vc_sub = vmulq_f32(vc_sub, valpha);
+      vd_sub = vmulq_f32(vd_sub, valpha);
+      float32x4_t va_min = vminq_f32(va_sub, vzero);
+      float32x4_t vb_min = vminq_f32(vb_sub, vzero);
+      float32x4_t vc_min = vminq_f32(vc_sub, vzero);
+      float32x4_t vd_min = vminq_f32(vd_sub, vzero);
+      float32x4_t va_rst = vaddq_f32(va_max, va_min);
+      float32x4_t vb_rst = vaddq_f32(vb_max, vb_min);
+      float32x4_t vc_rst = vaddq_f32(vc_max, vc_min);
+      float32x4_t vd_rst = vaddq_f32(vd_max, vd_min);
+      vst1q_f32(ptr_out_thread, va_rst);
+      vst1q_f32(ptr_out_thread + 4, vb_rst);
+      vst1q_f32(ptr_out_thread + 8, vc_rst);
+      vst1q_f32(ptr_out_thread + 12, vd_rst);
+      ptr_out_thread += 16;
+      ptr_in_thread += 16;
+    }
+    for (int j = 0; j < cnt; j++) {
+      float32x4_t va = vld1q_f32(ptr_in_thread);
+      float32x4_t va_exp = exp_ps(va);
+      float32x4_t va_max = vmaxq_f32(va, vzero);
+      float32x4_t va_sub = vsubq_f32(va_exp, vone);
+      va_sub = vmulq_f32(va_sub, valpha);
+      float32x4_t va_min = vminq_f32(va_sub, vzero);
+      float32x4_t va_rst = vaddq_f32(va_max, va_min);
+      vst1q_f32(ptr_out_thread, va_rst);
+      ptr_out_thread += 4;
+      ptr_in_thread += 4;
+    }
+    for (int j = 0; j < remain; j++) {
+      float beta = alpha * (expf(ptr_in_thread[0]) - 1);
+      float max = ptr_in_thread[0] >= 0.f ? ptr_in_thread[0] : 0.f;
+      float min = beta <= 0.f ? beta : 0.f;
+      ptr_out_thread[0] = min + max;
+      ptr_in_thread++;
+      ptr_out_thread++;
+    }
+  }
+  float* ptr_out = dout + threads * nums_per_thread;
+  const float* ptr_in = din + threads * nums_per_thread;
+  for (int j = 0; j < thread_remain; j++) {
+    float beta = alpha * (expf(ptr_in[0]) - 1);
+    float max = ptr_in[0] >= 0.f ? ptr_in[0] : 0.f;
+    float min = beta <= 0.f ? beta : 0.f;
+    ptr_out[0] = max + min;
+    ptr_in++;
+    ptr_out++;
+  }
+}
 }  // namespace math
 }  // namespace arm
 }  // namespace lite

lite/backends/arm/math/activation.h

@@ -90,6 +90,9 @@ template <typename T>
 void act_thresholded_relu(
     const T* din, T* dout, int size, float threshold, int threads);
 
+template <typename T>
+void act_elu(const T* din, T* dout, int size, float alpha, int threads);
+
 }  // namespace math
 }  // namespace arm
 }  // namespace lite

lite/core/mir/fusion/conv_conv_fuse_pass.cc

@@ -27,21 +27,29 @@ void ConvConvFusePass::Apply(const std::unique_ptr<SSAGraph>& graph) {
   // initialze fuser params
   std::vector<bool> conv_has_bias_cases{true, false};
   std::vector<std::string> conv_type_cases{"conv2d", "depthwise_conv2d"};
-  bool has_arm = false;
+  bool has_fp32 = false;
+  bool has_int8 = false;
   for (auto& place : graph->valid_places()) {
-    if (place.target == TARGET(kARM) && place.precision == PRECISION(kFloat)) {
-      has_arm = true;
-      break;
+    if (place.target == TARGET(kARM)) {
+      if (place.precision == PRECISION(kFloat)) {
+        has_fp32 = true;
       }
+      if (place.precision == PRECISION(kInt8)) {
+        has_int8 = true;
       }
-  if (!has_arm) {
+    } else {
+      return;
+    }
+  }
+  // only support arm-fp32
+  if (has_int8 || (has_fp32 && has_int8)) {
     return;
   }
   // only support fp32 fusion
   for (auto conv_has_bias0 : conv_has_bias_cases) {
     for (auto conv_has_bias1 : conv_has_bias_cases) {
       for (auto conv_type0 : conv_type_cases) {
-        for (auto conv_type1 : conv_type_cases) {
+        for (auto conv_type1 : {"conv2d"}) {  // it mustbe 1x1s1p0_conv
           VLOG(4) << "conv_has_bias0:" << conv_has_bias0
                   << " conv_type0:" << conv_type0;
           VLOG(4) << "conv_has_bias1:" << conv_has_bias1

lite/core/mir/fusion/conv_conv_fuser.cc

@@ -106,7 +106,7 @@ void ConvConvFuser::InsertNewNode(SSAGraph* graph, const key2nodes_t& matched) {
   int kw = weight1_t->dims()[2];
   int kh = weight1_t->dims()[3];
   if (!(kw == 1 && kh == 1)) {
-    return;
+    LOG(FATAL) << "The kernel size of the second conv must be 1x1";
   }
   CHECK_EQ(enable0_int8, enable1_int8) << "The Conv compute type must be same";
   CHECK_EQ(groups1, 1) << "The groups of weight1_dim must be 1";
@@ -117,11 +117,11 @@ void ConvConvFuser::InsertNewNode(SSAGraph* graph, const key2nodes_t& matched) {
                              << " must be 1";
   }
   for (int i = 0; i < paddings1.size(); i++) {
-    CHECK_EQ(paddings1[i], 0) << "paddings1[" << i << "]: " << paddings1[i]
-                              << " must be 0";
+    CHECK_EQ(paddings1[i], 1) << "paddings[" << i << "]: " << paddings1[i]
+                              << " must be 1";
   }
   for (int i = 0; i < dilations1.size(); i++) {
-    CHECK_EQ(dilations1[i], 1) << "dilations1[" << i << "]: " << dilations1[i]
+    CHECK_EQ(dilations1[i], 1) << "dilations[" << i << "]: " << dilations1[i]
                                << " must be 1";
   }
   // comupte new_wight and new bias
@@ -140,8 +140,7 @@ void ConvConvFuser::InsertNewNode(SSAGraph* graph, const key2nodes_t& matched) {
   //   new_bias = az + b
   ///////////////////////////////////////////////////////////////////////////////
   if (enable0_int8) {
-    LOG(FATAL) << "it doesn't support";
-    return;
+    LOG(FATAL) << "it doesn't support int8";
   } else {
     // compute new conv_weight
     Tensor weight_tensor;

lite/kernels/arm/activation_compute.cc

@@ -228,6 +228,17 @@ void ThresholdedReluCompute::Run() {
       x_data, output_data, x_dims.production(), threshold, ctx.threads());
 }
 
+void EluCompute::Run() {
+  auto& param = this->Param<param_t>();
+  auto& ctx = this->ctx_->template As<ARMContext>();
+  auto x_dims = param.X->dims();
+  auto x_data = param.X->data<float>();
+  auto output_data = param.Out->mutable_data<float>();
+  float alpha = param.Elu_alpha;
+  lite::arm::math::act_elu<float>(
+      x_data, output_data, x_dims.production(), alpha, ctx.threads());
+}
+
 }  // namespace arm
 }  // namespace kernels
 }  // namespace lite
@@ -356,3 +367,8 @@ REGISTER_LITE_KERNEL(thresholded_relu,
     .BindInput("X", {LiteType::GetTensorTy(TARGET(kARM))})
     .BindOutput("Out", {LiteType::GetTensorTy(TARGET(kARM))})
     .Finalize();
+REGISTER_LITE_KERNEL(
+    elu, kARM, kFloat, kNCHW, paddle::lite::kernels::arm::EluCompute, def)
+    .BindInput("X", {LiteType::GetTensorTy(TARGET(kARM))})
+    .BindOutput("Out", {LiteType::GetTensorTy(TARGET(kARM))})
+    .Finalize();

lite/kernels/arm/activation_compute.h

@@ -185,6 +185,15 @@ class ThresholdedReluCompute
   virtual ~ThresholdedReluCompute() = default;
 };
 
+class EluCompute : public KernelLite<TARGET(kARM), PRECISION(kFloat)> {
+ public:
+  using param_t = operators::ActivationParam;
+
+  void Run() override;
+
+  virtual ~EluCompute() = default;
+};
+
 }  // namespace arm
 }  // namespace kernels
 }  // namespace lite

lite/operators/activation_ops.cc

@@ -85,6 +85,9 @@ bool ActivationOp::AttachImpl(const cpp::OpDesc& opdesc, lite::Scope* scope) {
   } else if (opdesc.Type() == "thresholded_relu") {
     param_.active_type = lite_api::ActivationType::kThresholdedRelu;
     param_.relu_threshold = opdesc.GetAttr<float>("threshold");
+  } else if (opdesc.Type() == "elu") {
+    param_.active_type = lite_api::ActivationType::kElu;
+    param_.Elu_alpha = opdesc.GetAttr<float>("alpha");
   }
 
   VLOG(4) << "opdesc.Type():" << opdesc.Type();
@@ -105,3 +108,4 @@ REGISTER_LITE_OP(leaky_relu, paddle::lite::operators::ActivationOp);
 REGISTER_LITE_OP(relu6, paddle::lite::operators::ActivationOp);
 REGISTER_LITE_OP(prelu, paddle::lite::operators::ActivationOp);
 REGISTER_LITE_OP(thresholded_relu, paddle::lite::operators::ActivationOp);
+REGISTER_LITE_OP(elu, paddle::lite::operators::ActivationOp);

lite/operators/activation_ops.h

@@ -83,6 +83,9 @@ class ActivationOp : public OpLite {
       case lite_api::ActivationType::kThresholdedRelu:
         ch->macs = param_.X->numel();
         break;
+      case lite_api::ActivationType::kElu:
+        ch->macs = param_.X->numel();
+        break;
       default:
         LOG(FATAL) << "This Type of Activation:"
                    << static_cast<int>(param_.active_type)

lite/operators/op_params.h

@@ -359,6 +359,8 @@ struct ActivationParam : ParamBase {
   float hard_swish_offset{3.0};
   // thresholded_relu
   float relu_threshold{1.0f};
+  // elu
+  float Elu_alpha{1.0f};
 };
 
 struct ActivationGradParam : ParamBase {

lite/tests/kernels/activation_compute_test.cc

@@ -39,7 +39,8 @@ enum activation_type_test {
   SQUARE,
   HARD_SWISH,
   RECIPROCAL,
-  THRESHOLDED_RELU
+  THRESHOLDED_RELU,
+  ELU
 };
 
 class ActivationComputeTester : public arena::TestCase {
@@ -56,6 +57,7 @@ class ActivationComputeTester : public arena::TestCase {
   float hard_swish_scale = 6.0;
   float hard_swish_offset = 3.0;
   float relu_threshold_ = 1.0;
+  float elu_alpha_ = 1.0;
   DDim dims_{{1}};
   std::string type_ = "";
   activation_type_test act_type_ = RELU;
@@ -67,6 +69,7 @@ class ActivationComputeTester : public arena::TestCase {
                           float relu_clipped_coef,
                           std::string prelu_mode,
                           float swish_beta,
+                          float elu_alpha,
                           DDim dims,
                           std::string type,
                           activation_type_test act_type)
@@ -75,6 +78,7 @@ class ActivationComputeTester : public arena::TestCase {
         relu_clipped_coef_(relu_clipped_coef),
         prelu_mode_(prelu_mode),
         swish_beta_(swish_beta),
+        elu_alpha_(elu_alpha),
         dims_(dims),
         type_(type),
         act_type_(act_type) {}
@@ -87,6 +91,7 @@ class ActivationComputeTester : public arena::TestCase {
 
     auto* x = scope->FindTensor(input_);
     const auto* x_data = x->data<float>();
+    LOG(INFO) << act_type_;
     switch (act_type_) {
       case RELU: {
         for (int i = 0; i < dims_.production(); i++) {
@@ -226,8 +231,17 @@ class ActivationComputeTester : public arena::TestCase {
         }
         break;
       }
+      case ELU: {
+        for (int i = 0; i < dims_.production(); i++) {
+          float tmp = std::exp(x_data[i]) - 1;
+          float max = x_data[i] > 0.f ? x_data[i] : 0.f;
+          float min = x_data[i] < 0.f ? elu_alpha_ * tmp : 0.f;
+          output_data[i] = min + max;
+        }
+        break;
+      }
       default:
-        LOG(INFO) << "the type of activation is unknow.";
+        LOG(INFO) << "the type of activation " << act_type_ << " is unknow.";
     }
   }
 
@@ -256,6 +270,9 @@ class ActivationComputeTester : public arena::TestCase {
     if (act_type_ == THRESHOLDED_RELU) {
       op_desc->SetAttr("threshold", relu_threshold_);
     }
+    if (act_type_ == ELU) {
+      op_desc->SetAttr("alpha", elu_alpha_);
+    }
   }
 
   void PrepareData() override {
@@ -312,7 +329,7 @@ TEST(Activation_relu, precision) {
   for (auto dims : std::vector<std::vector<int64_t>>{
            {1, 3, 2, 4}, {2, 3, 4}, {5, 4}, {8}}) {
     std::unique_ptr<arena::TestCase> tester(new ActivationComputeTester(
-        place, "def", 0.01, 6., "all", 0., DDim(dims), "relu", RELU));
+        place, "def", 0.01, 6., "all", 0., 1.0, DDim(dims), "relu", RELU));
     arena::Arena arena(std::move(tester), place, abs_error);
     arena.TestPrecision();
   }
@@ -344,6 +361,7 @@ TEST(Activation_leaky_relu, precision) {
                                       6.,
                                       "all",
                                       0.,
+                                      1.0,
                                       DDim(dims),
                                       "leaky_relu",
                                       LEAKY_RELU));
@@ -376,6 +394,7 @@ TEST(Activation_relu_clipped, precision) {
                                       coef,
                                       "all",
                                       0.,
+                                      1.0,
                                       DDim(dims),
                                       "relu_clipped",
                                       RELU_CLIPPED));
@@ -393,7 +412,7 @@ TEST(Activation_prelu, precision) {
   for (auto dims : std::vector<std::vector<int64_t>>{{1, 3, 2, 4}}) {
     for (auto mode : {"all", "channel", "element"}) {
       std::unique_ptr<arena::TestCase> tester(new ActivationComputeTester(
-          place, "def", 0.01, 6, mode, 0., DDim(dims), "prelu", PRELU));
+          place, "def", 0.01, 6, mode, 0., 1.0, DDim(dims), "prelu", PRELU));
       arena::Arena arena(std::move(tester), place, 2e-5);
       arena.TestPrecision();
     }
@@ -419,8 +438,17 @@ TEST(Activation_sigmoid, precision) {
 
   for (auto dims : std::vector<std::vector<int64_t>>{
            {1, 3, 2, 4}, {2, 3, 4}, {5, 4}, {8}}) {
-    std::unique_ptr<arena::TestCase> tester(new ActivationComputeTester(
-        place, "def", 0.01, 6., "all", 0., DDim(dims), "sigmoid", SIGMOID));
+    std::unique_ptr<arena::TestCase> tester(
+        new ActivationComputeTester(place,
+                                    "def",
+                                    0.01,
+                                    6.,
+                                    "all",
+                                    0.,
+                                    1.0,
+                                    DDim(dims),
+                                    "sigmoid",
+                                    SIGMOID));
     arena::Arena arena(std::move(tester), place, abs_error);
     arena.TestPrecision();
   }
@@ -447,7 +475,7 @@ TEST(Activation_tanh, precision) {
   for (auto dims : std::vector<std::vector<int64_t>>{
            {1, 3, 2, 4}, {2, 3, 4}, {5, 4}, {8}}) {
     std::unique_ptr<arena::TestCase> tester(new ActivationComputeTester(
-        place, "def", 0.01, 6., "all", 0., DDim(dims), "tanh", TANH));
+        place, "def", 0.01, 6., "all", 0., 1.0, DDim(dims), "tanh", TANH));
     arena::Arena arena(std::move(tester), place, abs_error);
     arena.TestPrecision();
   }
@@ -462,7 +490,7 @@ TEST(Activation_swish, precision) {
            {1, 3, 2, 4}, {2, 3, 4}, {5, 4}, {8}}) {
     for (auto coef : {0.01, 0.1}) {
       std::unique_ptr<arena::TestCase> tester(new ActivationComputeTester(
-          place, "def", 0.01, 6, "all", coef, DDim(dims), "swish", SWISH));
+          place, "def", 0.01, 6, "all", coef, 1.0, DDim(dims), "swish", SWISH));
       arena::Arena arena(std::move(tester), place, 2e-5);
       arena.TestPrecision();
     }
@@ -489,7 +517,7 @@ TEST(Activation_relu6, precision) {
   for (auto dims : std::vector<std::vector<int64_t>>{
            {1, 3, 2, 4}, {2, 3, 4}, {5, 4}, {8}}) {
     std::unique_ptr<arena::TestCase> tester(new ActivationComputeTester(
-        place, "def", 0.01, 6., "all", 0., DDim(dims), "relu6", RELU6));
+        place, "def", 0.01, 6., "all", 0., 1.0, DDim(dims), "relu6", RELU6));
     arena::Arena arena(std::move(tester), place, abs_error);
     arena.TestPrecision();
   }
@@ -511,7 +539,7 @@ TEST(Activation_log, precision) {
   for (auto dims : std::vector<std::vector<int64_t>>{
            {1, 3, 2, 4}, {2, 3, 4}, {5, 4}, {8}}) {
     std::unique_ptr<arena::TestCase> tester(new ActivationComputeTester(
-        place, "def", 0.01, 6., "all", 0., DDim(dims), "log", LOG));
+        place, "def", 0.01, 6., "all", 0., 1.0, DDim(dims), "log", LOG));
     arena::Arena arena(std::move(tester), place, abs_error);
     arena.TestPrecision();
   }
@@ -525,7 +553,7 @@ TEST(Activation_exp, precision) {
   for (auto dims : std::vector<std::vector<int64_t>>{
            {1, 3, 2, 4}, {2, 3, 4}, {5, 4}, {8}}) {
     std::unique_ptr<arena::TestCase> tester(new ActivationComputeTester(
-        place, "def", 0.01, 6., "all", 0., DDim(dims), "exp", EXP));
+        place, "def", 0.01, 6., "all", 0., 1.0, DDim(dims), "exp", EXP));
     arena::Arena arena(std::move(tester), place, 2e-5);
     arena.TestPrecision();
   }
@@ -539,7 +567,7 @@ TEST(Activation_floor, precision) {
   for (auto dims : std::vector<std::vector<int64_t>>{
            {1, 3, 2, 4}, {2, 3, 4}, {5, 4}, {8}}) {
     std::unique_ptr<arena::TestCase> tester(new ActivationComputeTester(
-        place, "def", 0.01, 6., "all", 0., DDim(dims), "floor", FLOOR));
+        place, "def", 0.01, 6., "all", 0., 1.0, DDim(dims), "floor", FLOOR));
     arena::Arena arena(std::move(tester), place, 2e-5);
     arena.TestPrecision();
   }
@@ -554,7 +582,7 @@ TEST(Activation_rsqrt, precision) {
   for (auto dims : std::vector<std::vector<int64_t>>{
            {1, 3, 2, 4}, {2, 3, 4}, {5, 4}, {8}}) {
     std::unique_ptr<arena::TestCase> tester(new ActivationComputeTester(
-        place, "def", 0.01, 6., "all", 0., DDim(dims), "rsqrt", RSQRT));
+        place, "def", 0.01, 6., "all", 0., 1.0, DDim(dims), "rsqrt", RSQRT));
     arena::Arena arena(std::move(tester), place, 2e-5);
     arena.TestPrecision();
   }
@@ -577,7 +605,7 @@ TEST(Activation_square, precision) {
   for (auto dims : std::vector<std::vector<int64_t>>{
            {1, 3, 2, 4}, {2, 3, 4}, {5, 4}, {8}}) {
     std::unique_ptr<arena::TestCase> tester(new ActivationComputeTester(
-        place, "def", 0.01, 6., "all", 0., DDim(dims), "square", SQUARE));
+        place, "def", 0.01, 6., "all", 0., 1.0, DDim(dims), "square", SQUARE));
     arena::Arena arena(std::move(tester), place, abs_error);
     arena.TestPrecision();
   }
@@ -596,7 +624,7 @@ TEST(Activation_gelu, precision) {
   for (auto dims : std::vector<std::vector<int64_t>>{
            {1, 3, 2, 4}, {2, 3, 4}, {5, 4}, {8}}) {
     std::unique_ptr<arena::TestCase> tester(new ActivationComputeTester(
-        place, "def", 0.01, 6., "all", 0., DDim(dims), "gelu", GELU));
+        place, "def", 0.01, 6., "all", 0., 1.0, DDim(dims), "gelu", GELU));
     arena::Arena arena(std::move(tester), place, abs_error);
     arena.TestPrecision();
   }
@@ -622,6 +650,7 @@ TEST(activation_hard_swish, precision) {
                                     6.,
                                     "all",
                                     0.,
+                                    1.0,
                                     DDim(dims),
                                     "hard_swish",
                                     HARD_SWISH));
@@ -650,6 +679,7 @@ TEST(activation_reciprocal, precision) {
                                     6.,
                                     "all",
                                     0.,
+                                    1.0,
                                     DDim(dims),
                                     "reciprocal",
                                     RECIPROCAL));
@@ -680,6 +710,7 @@ TEST(Activation_thresholded_relu, precision) {
                                     6.,
                                     "all",
                                     0.,
+                                    1.0,
                                     DDim(dims),
                                     "thresholded_relu",
                                     THRESHOLDED_RELU));
@@ -688,5 +719,20 @@ TEST(Activation_thresholded_relu, precision) {
   }
 }
 
+TEST(Activation_elu, precision) {
+  LOG(INFO) << "test elu op";
+#ifdef LITE_WITH_ARM
+  Place place(TARGET(kARM));
+
+  for (auto dims : std::vector<std::vector<int64_t>>{
+           {1, 3, 2, 4}, {2, 3, 4}, {5, 4}, {8}}) {
+    std::unique_ptr<arena::TestCase> tester(new ActivationComputeTester(
+        place, "def", 0.01, 6., "all", 0., 1.0, DDim(dims), "elu", ELU));
+    arena::Arena arena(std::move(tester), place, 2e-5);
+    arena.TestPrecision();
+  }
+#endif
+}
+
 }  // namespace lite
 }  // namespace paddle

lite/tools/build.sh

@@ -268,6 +268,8 @@ function make_all_tests {
   cmake $root_dir \
       ${CMAKE_COMMON_OPTIONS} \
       -DWITH_TESTING=ON \
+      -DLITE_WITH_PROFILE=OFF \
+      -DLITE_WITH_PRECISION_PROFILE=OFF \
       -DLITE_BUILD_EXTRA=$BUILD_EXTRA \
       -DLITE_WITH_CV=$BUILD_CV \
       -DLITE_WITH_NPU=$BUILD_NPU \