[cherry-pick]BatchNorm use inplace (#49529)

att, cherry-pick#48254, and resolve conflict

paddle/phi/api/yaml/generator/generate_sparse_op.py

@@ -84,6 +84,8 @@ def main(
     backward_api_dict = to_named_dict(backward_apis)
 
     for api in apis:
+        if api['name'][-1] == '_':
+            api['name'] = api['name'][:-1]
         api['op_name'] = SPARSE_OP_PREFIX + api['name']
         api['name'] = api['op_name']
         if api["backward"] is not None:

paddle/phi/api/yaml/sparse_backward.yaml

@@ -101,7 +101,7 @@
            atanh_csr_grad {sparse_csr, sparse_csr -> sparse_csr}
 
 - backward_op : batch_norm_grad
-  forward : batch_norm (Tensor x, Tensor scale, Tensor bias, Tensor mean, Tensor variance, float momentum, float epsilon, str data_layout, bool is_test, bool use_global_stats, bool trainable_statistics, bool fuse_with_relu) -> Tensor(out), Tensor(mean_out), Tensor(variance_out), Tensor(saved_mean), Tensor(saved_variance), Tensor(reserve_space)
+  forward : batch_norm_ (Tensor x, Tensor scale, Tensor bias, Tensor mean, Tensor variance, float momentum, float epsilon, str data_layout, bool is_test, bool use_global_stats, bool trainable_statistics, bool fuse_with_relu) -> Tensor(out), Tensor(mean_out), Tensor(variance_out), Tensor(saved_mean), Tensor(saved_variance), Tensor(reserve_space)
   args : (Tensor x, Tensor scale, Tensor bias, Tensor mean_out, Tensor variance_out, Tensor saved_mean, Tensor saved_variance, Tensor reserve_space, Tensor out_grad, float momentum, float epsilon, str data_layout, bool is_test, bool use_global_stats, bool trainable_statistics, bool fuse_with_relu)
   output : Tensor(x_grad), Tensor(scale_grad), Tensor(bias_grad)
   infer_meta :

paddle/phi/api/yaml/sparse_ops.yaml

@@ -87,7 +87,7 @@
     layout : x
   backward : atanh_grad
 
-- op : batch_norm
+- op : batch_norm_
   args : (Tensor x, Tensor scale, Tensor bias, Tensor mean, Tensor variance, float momentum, float epsilon, str data_layout, bool is_test, bool use_global_stats, bool trainable_statistics, bool fuse_with_relu)
   output : Tensor(out), Tensor(mean_out), Tensor(variance_out), Tensor(saved_mean), Tensor(saved_variance), Tensor(reserve_space)
   infer_meta :
@@ -95,7 +95,7 @@
   kernel :
     func : batch_norm_coo {sparse_coo, dense, dense, dense, dense -> sparse_coo, dense, dense, dense, dense, dense}
     data_type : x
-  view : (mean -> mean_out), (variance -> variance_out)
+  inplace : (mean -> mean_out), (variance -> variance_out)
   backward : batch_norm_grad
 
 - op : cast

paddle/phi/kernels/sparse/batch_norm_kernel.h

@@ -23,7 +23,7 @@ namespace phi {
 namespace sparse {
 
 template <typename T, typename Context>
-void BatchNormKernel(const Context& dev_ctx,
+void BatchNormCooKernel(const Context& dev_ctx,
                         const SparseCooTensor& x,
                         const DenseTensor& scale,
                         const DenseTensor& bias,

python/paddle/sparse/nn/layer/norm.py

@@ -138,7 +138,7 @@ class BatchNorm(paddle.nn.BatchNorm1D):
         data_format = 'NCHW' if self._data_format[1] == 'C' else 'NHWC'
 
         if in_dynamic_mode():
-            batch_norm_out, _, _, _, _, _ = _C_ops.sparse_batch_norm(
+            batch_norm_out, _, _, _, _, _ = _C_ops.sparse_batch_norm_(
                 input,
                 self.weight,
                 self.bias,