Adapt to batch_norm_grad op and add align function in roi_align op for kunlun (#39685)

* Adapt to batch_norm_grad op and add align function in
roi_align op for kunlun, *test=kunlun

* Adapt to batch_norm, batch_norm_grad op api for kunlun, and add unit-tests of batch_norm, roi_align. *test=kunlun

paddle/fluid/operators/batch_norm_op_xpu.cc

@@ -38,23 +38,25 @@ class BatchNormXPUKernel : public framework::OpKernel<T> {
     bool global_stats = test_mode || use_global_stats;
     const auto &data_layout_str = ctx.Attr<std::string>("data_layout");
     const auto data_layout = framework::StringToDataLayout(data_layout_str);
-    PADDLE_ENFORCE_EQ(data_layout, DataLayout::kNCHW,
+    PADDLE_ENFORCE_EQ(data_layout_str == "NCHW" || data_layout_str == "NHWC",
+                      true,
                       platform::errors::InvalidArgument(
-                          "The 'data_layout' attribute must be NCHW. But "
-                          "recevived 'data_layout' is [%s].",
+                          "The 'data_layout' attribute must be NCHW or NHWC. "
+                          "But recevived 'data_layout' is [%s].",
                           data_layout_str));
 
     const auto *x = ctx.Input<Tensor>("X");
     const auto &x_dims = x->dims();
-    PADDLE_ENFORCE_EQ(x_dims.size(), 4,
+    PADDLE_ENFORCE_EQ(
+        x_dims.size() >= 2 && x_dims.size() <= 5, true,
         platform::errors::InvalidArgument(
-                          "The input tensor X's dimension must equal to 4. But "
-                          "received X's shape = [%s], X's dimension = [%d].",
-                          x_dims, x_dims.size()));
-    const int N = x_dims[0];
-    const int C = x_dims[1];
-    const int H = x_dims[2];
-    const int W = x_dims[3];
+            "The size of input's dimensions should be between 2 and 5"
+            "But received: the size of input's dimensions is [%d]",
+            x_dims.size()));
+
+    int N, C, H, W, D;
+    ExtractNCWHD(x_dims, data_layout, &N, &C, &H, &W, &D);
+
     const auto *scale = ctx.Input<Tensor>("Scale");
     const auto *bias = ctx.Input<Tensor>("Bias");
     const auto *x_data = x->data<T>();
@@ -75,6 +77,7 @@ class BatchNormXPUKernel : public framework::OpKernel<T> {
     saved_variance->mutable_data<float>(ctx.GetPlace());
 
     auto &dev_ctx = ctx.template device_context<DeviceContext>();
+    bool is_nchw = data_layout_str == "NCHW";
 
     if (!global_stats) {
       auto *mean_out_data = mean_out->data<float>();
@@ -95,7 +98,7 @@ class BatchNormXPUKernel : public framework::OpKernel<T> {
       int r = xpu::batch_norm<T>(dev_ctx.x_context(), x_data, y_data, N, C, H,
                                  W, epsilon, momentum, scale_data, bias_data,
                                  saved_mean_data, saved_variance_data,
-                                 mean_out_data, variance_out_data, true);
+                                 mean_out_data, variance_out_data, is_nchw);
       PADDLE_ENFORCE_EQ(r, xpu::Error_t::SUCCESS,
                         platform::errors::External(
                             "The batch_norm XPU API return wrong value[%d %s]",
@@ -107,7 +110,7 @@ class BatchNormXPUKernel : public framework::OpKernel<T> {
       const auto *variance_data = variance->data<float>();
       int r = xpu::batch_norm_infer(dev_ctx.x_context(), x_data, y_data, N, C,
                                     H, W, epsilon, scale_data, bias_data,
-                                    mean_data, variance_data, true);
+                                    mean_data, variance_data, is_nchw);
       PADDLE_ENFORCE_EQ(
           r, xpu::Error_t::SUCCESS,
           platform::errors::External(
@@ -168,11 +171,11 @@ class BatchNormGradXPUKernel : public framework::OpKernel<T> {
     const float epsilon = ctx.Attr<float>("epsilon");
     const auto data_layout = framework::StringToDataLayout(data_layout_str);
 
-    // TODO(guozbin): Transform input tensor from NHWC to NCHW
-    PADDLE_ENFORCE_EQ(data_layout, DataLayout::kNCHW,
+    PADDLE_ENFORCE_EQ(data_layout_str == "NCHW" || data_layout_str == "NHWC",
+                      true,
                       platform::errors::InvalidArgument(
-                          "The 'data_layout' attribute must be NCHW. But "
-                          "recevived 'data_layout' is [%s].",
+                          "The 'data_layout' attribute must be NCHW or NHWC. "
+                          "But recevived 'data_layout' is [%s].",
                           data_layout_str));
 
     auto *d_x = ctx.Output<Tensor>(framework::GradVarName("X"));
@@ -207,15 +210,15 @@ class BatchNormGradXPUKernel : public framework::OpKernel<T> {
     }
 
     const auto &x_dims = x->dims();
-    PADDLE_ENFORCE_EQ(x_dims.size(), 4,
+    PADDLE_ENFORCE_EQ(
+        x_dims.size() >= 2 && x_dims.size() <= 5, true,
         platform::errors::InvalidArgument(
-                          "The input tensor X's dimension must equal to 4. But "
-                          "received X's shape = [%s], X's dimension = [%d].",
-                          x_dims, x_dims.size()));
-    const int N = x_dims[0];
-    const int C = x_dims[1];
-    const int H = x_dims[2];
-    const int W = x_dims[3];
+            "The size of input's dimensions should be between 2 and 5"
+            "But received: the size of input's dimensions is [%d]",
+            x_dims.size()));
+
+    int N, C, H, W, D;
+    ExtractNCWHD(x_dims, data_layout, &N, &C, &H, &W, &D);
 
     const auto *x_data = x->data<T>();
     const auto *d_y_data = d_y->data<T>();
@@ -250,38 +253,35 @@ class BatchNormGradXPUKernel : public framework::OpKernel<T> {
     auto &dev_ctx = ctx.template device_context<DeviceContext>();
     xpu::ctx_guard RAII_GUARD(dev_ctx.x_context());
 
-    const T *mean_data = nullptr;
-    const T *inv_var_data = nullptr;
+    const auto *batch_mean = ctx.Input<Tensor>("SavedMean");
+    const auto *batch_inv_std = ctx.Input<Tensor>("SavedVariance");
+    const auto *global_mean = ctx.Input<Tensor>("Mean");
+    const auto *global_var = ctx.Input<Tensor>("Variance");
 
     // TODO(guozibin): hadle the situation case of N * H * W = 1
-    if (!use_global_stats) {
-      const auto *saved_mean = ctx.Input<Tensor>("SavedMean");
-      // SavedVariance have been reverted in forward operator
-      const auto *saved_inv_variance = ctx.Input<Tensor>("SavedVariance");
-      mean_data = saved_mean->data<float>();
-      inv_var_data = saved_inv_variance->data<float>();
-    } else {
-      const auto *running_mean = ctx.Input<Tensor>("Mean");
-      const auto *running_variance = ctx.Input<Tensor>("Variance");
-      mean_data = running_mean->data<float>();
-      inv_var_data = running_variance->data<float>();
-      float *running_inv_var_data =
-          RAII_GUARD.alloc_l3_or_gm<float>(running_variance->numel());
+    if (is_inplace) {
+      float *global_inv_std_data;
+      if (use_global_stats) {
+        global_inv_std_data =
+            RAII_GUARD.alloc_l3_or_gm<float>(global_var->numel());
         float *epsilon_data = RAII_GUARD.alloc_l3_or_gm<float>(1);
-      int r1 = calculate_inv_var(dev_ctx.x_context(), inv_var_data, epsilon, C,
-                                 epsilon_data, running_inv_var_data);
+        int r1 =
+            calculate_inv_var(dev_ctx.x_context(), global_var->data<float>(),
+                              epsilon, C, epsilon_data, global_inv_std_data);
         PADDLE_ENFORCE_EQ(r1, XPU_SUCCESS, platform::errors::External(
                                                "XPU API(batch_norm_grad "
                                                "calculate_inv_var function) "
                                                "return wrong value[%d %s]",
                                                r1, XPUAPIErrorMsg[r1]));
-      inv_var_data = running_inv_var_data;
       }
-    if (is_inplace) {
       auto px = *x;
+      auto *inv_std_data =
+          use_global_stats ? global_inv_std_data : batch_inv_std->data<float>();
+      auto mean_data = use_global_stats ? global_mean->data<float>()
+                                        : batch_mean->data<float>();
       int r2 = calculate_inv_BN_Y(
           dev_ctx.x_context(), px.mutable_data<T>(ctx.GetPlace()),
-          scale->data<float>(), bias->data<float>(), mean_data, inv_var_data, N,
+          scale->data<float>(), bias->data<float>(), mean_data, inv_std_data, N,
           C, H * W, x->data<T>());
       PADDLE_ENFORCE_EQ(r2, XPU_SUCCESS, platform::errors::External(
                                              "XPU API(batch_norm_grad "
@@ -289,6 +289,15 @@ class BatchNormGradXPUKernel : public framework::OpKernel<T> {
                                              "return wrong value[%d %s]",
                                              r2, XPUAPIErrorMsg[r2]));
     }
+
+    int r3;
+    bool is_nchw = data_layout_str == "NCHW";
+    if (use_global_stats) {
+      r3 = xpu::batch_norm_grad<T>(
+          dev_ctx.x_context(), x_data, d_y_data, d_x_data, N, C, H, W,
+          scale_data, nullptr, nullptr, d_scale_data, d_bias_data, is_nchw,
+          global_mean->data<float>(), global_var->data<float>(), epsilon);
+    } else {
       if (!d_x) {
         d_x_data = RAII_GUARD.alloc_l3_or_gm<T>(x->numel());
       }
@@ -298,10 +307,11 @@ class BatchNormGradXPUKernel : public framework::OpKernel<T> {
       if (!d_bias_data) {
         d_bias_data = RAII_GUARD.alloc_l3_or_gm<float>(C);
       }
-
-    int r3 = xpu::batch_norm_grad<T>(
-        dev_ctx.x_context(), x_data, d_y_data, d_x_data, N, C, H, W, scale_data,
-        mean_data, inv_var_data, d_scale_data, d_bias_data, true);
+      r3 = xpu::batch_norm_grad<T>(
+          dev_ctx.x_context(), x_data, d_y_data, d_x_data, N, C, H, W,
+          scale_data, batch_mean->data<float>(), batch_inv_std->data<float>(),
+          d_scale_data, d_bias_data, is_nchw);
+    }
     PADDLE_ENFORCE_EQ(r3, XPU_SUCCESS, platform::errors::External(
                                            "XPU API(batch_norm_grad) return "
                                            "wrong value[%d %s]",

paddle/fluid/operators/roi_align_op_xpu.cc

@@ -32,6 +32,7 @@ class XPUROIAlignOpKernel : public framework::OpKernel<T> {
     auto pooled_width = ctx.Attr<int>("pooled_width");
     auto spatial_scale = ctx.Attr<float>("spatial_scale");
     auto sampling_ratio = ctx.Attr<int>("sampling_ratio");
+    auto aligned = ctx.Attr<bool>("aligned");
 
     auto in_dims = in->dims();
     int batch_size = in_dims[0];
@@ -117,7 +118,7 @@ class XPUROIAlignOpKernel : public framework::OpKernel<T> {
         dev_ctx.x_context(), in->data<T>(),
         out->mutable_data<T>(ctx.GetPlace()), rois->data<T>(), roi_id_data,
         batch_size, channels, height, width, out->dims()[0], pooled_height,
-        pooled_width, spatial_scale, sampling_ratio, true);
+        pooled_width, spatial_scale, sampling_ratio, true, aligned);
     PADDLE_ENFORCE_EQ(r, xpu::Error_t::SUCCESS,
                       platform::errors::External(
                           "The roi_align XPU OP return wrong value[%d %s]", r,
@@ -143,6 +144,7 @@ class XPUROIAlignGradOpKernel : public framework::OpKernel<T> {
     auto pooled_width = ctx.Attr<int>("pooled_width");
     auto spatial_scale = ctx.Attr<float>("spatial_scale");
     auto sampling_ratio = ctx.Attr<int>("sampling_ratio");
+    auto aligned = ctx.Attr<bool>("aligned");
 
     int rois_num = rois->dims()[0];
     int channels = in->dims()[1];
@@ -197,7 +199,7 @@ class XPUROIAlignGradOpKernel : public framework::OpKernel<T> {
           dev_ctx.x_context(), out_grad->data<T>(), in_grad->data<T>(),
           rois->data<T>(), roi_id_data, in->dims()[0], channels, height, width,
           out_grad->dims()[0], pooled_height, pooled_width, spatial_scale,
-          sampling_ratio, true);
+          sampling_ratio, true, aligned);
       PADDLE_ENFORCE_EQ(
           r, xpu::Error_t::SUCCESS,
           platform::errors::External(

python/paddle/fluid/tests/unittests/xpu/test_batch_norm_op_xpu.py

@@ -296,7 +296,9 @@ class TestXPUBatchNormOpUseGlobalStats(unittest.TestCase):
                     net2.training = False
                 y1 = net1(x)
                 y2 = net2(x)
-                self.assertEqual(np.allclose(y1.numpy(), y2.numpy()), True)
+                self.assertEqual(
+                    np.allclose(
+                        y1.numpy(), y2.numpy(), atol=1e-4), True)
 
 
 class TestXPUBatchNormUseGlobalStatsCase1(TestXPUBatchNormOpUseGlobalStats):
@@ -320,5 +322,12 @@ class TestXPUBatchNormUseGlobalStatsCase3(TestXPUBatchNormOpUseGlobalStats):
         self.trainable_statistics = True
 
 
+class TestXPUBatchNormUseGlobalStatsCase4(TestXPUBatchNormOpUseGlobalStats):
+    ### train mode
+    def init_test(self):
+        self.use_global_stats = True
+        self.trainable_statistics = False
+
+
 if __name__ == "__main__":
     unittest.main()

python/paddle/fluid/tests/unittests/xpu/test_roi_align_op_xpu.py

@@ -40,7 +40,8 @@ class TestROIAlignOp(XPUOpTest):
             'spatial_scale': self.spatial_scale,
             'pooled_height': self.pooled_height,
             'pooled_width': self.pooled_width,
-            'sampling_ratio': self.sampling_ratio
+            'sampling_ratio': self.sampling_ratio,
+            'aligned': self.continuous_coordinate
         }
 
         self.outputs = {'Out': self.out_data}
@@ -51,6 +52,8 @@ class TestROIAlignOp(XPUOpTest):
         self.height = 8
         self.width = 6
 
+        self.xpu_version = core.get_xpu_device_version(0)
+
         # n, c, h, w
         self.x_dim = (self.batch_size, self.channels, self.height, self.width)
 
@@ -58,7 +61,10 @@ class TestROIAlignOp(XPUOpTest):
         self.pooled_height = 2
         self.pooled_width = 2
         self.sampling_ratio = -1
-
+        if self.xpu_version == core.XPUVersion.XPU1:
+            self.continuous_coordinate = False
+        else:
+            self.continuous_coordinate = bool(np.random.randint(2))
         self.x = np.random.random(self.x_dim).astype('float32')
 
     def pre_calc(self, x_i, roi_xmin, roi_ymin, roi_bin_grid_h, roi_bin_grid_w,
@@ -124,12 +130,16 @@ class TestROIAlignOp(XPUOpTest):
             roi = self.rois[i]
             roi_batch_id = int(roi[0])
             x_i = self.x[roi_batch_id]
-            roi_xmin = roi[1] * self.spatial_scale
-            roi_ymin = roi[2] * self.spatial_scale
-            roi_xmax = roi[3] * self.spatial_scale
-            roi_ymax = roi[4] * self.spatial_scale
-            roi_width = max(roi_xmax - roi_xmin, 1)
-            roi_height = max(roi_ymax - roi_ymin, 1)
+            roi_offset = 0.5 if self.continuous_coordinate else 0
+            roi_xmin = roi[1] * self.spatial_scale - roi_offset
+            roi_ymin = roi[2] * self.spatial_scale - roi_offset
+            roi_xmax = roi[3] * self.spatial_scale - roi_offset
+            roi_ymax = roi[4] * self.spatial_scale - roi_offset
+            roi_width = roi_xmax - roi_xmin
+            roi_height = roi_ymax - roi_ymin
+            if not self.continuous_coordinate:
+                roi_width = max(roi_width, 1)
+                roi_height = max(roi_height, 1)
             bin_size_h = float(roi_height) / float(self.pooled_height)
             bin_size_w = float(roi_width) / float(self.pooled_width)
             roi_bin_grid_h = self.sampling_ratio if self.sampling_ratio > 0 else \
@@ -203,7 +213,8 @@ class TestROIAlignInLodOp(TestROIAlignOp):
             'spatial_scale': self.spatial_scale,
             'pooled_height': self.pooled_height,
             'pooled_width': self.pooled_width,
-            'sampling_ratio': self.sampling_ratio
+            'sampling_ratio': self.sampling_ratio,
+            'aligned': self.continuous_coordinate
         }
 
         self.outputs = {'Out': self.out_data}