!26903 [feat] [assistant] [I48OB0] maximum with dynamic infer shape

Merge pull request !26903 from 郑鹏飞/maximum

mindspore/core/ops/maximum.cc

@@ -1,5 +1,5 @@
 /**
- * Copyright 2020 Huawei Technologies Co., Ltd
+ * Copyright 2020-2021 Huawei Technologies Co., Ltd
  *
  * Licensed under the Apache License, Version 2.0 (the "License");
  * you may not use this file except in compliance with the License.
@@ -18,33 +18,68 @@
 #include <string>
 #include "ops/maximum.h"
 #include "ops/op_utils.h"
-#include "utils/check_convert_utils.h"
 
 namespace mindspore {
 namespace ops {
 namespace {
-abstract::ShapePtr InferShape(const PrimitivePtr &primitive, const std::vector<AbstractBasePtr> &input_args) {
+abstract::ShapePtr MaximumInferShape(const PrimitivePtr &primitive, const std::vector<AbstractBasePtr> &input_args) {
   MS_EXCEPTION_IF_NULL(primitive);
-  auto op_name = primitive->name();
-  return BroadCastInferShape(op_name, input_args);
-}
-
-TypePtr InferType(const PrimitivePtr &prim, const std::vector<AbstractBasePtr> &input_args) {
+  auto prim_name = primitive->name();
+  const int64_t kInputNum = 2;
+  (void)CheckAndConvertUtils::CheckInteger("input number", SizeToLong(input_args.size()), kGreaterEqual, kInputNum,
+                                           prim_name);
   for (const auto &item : input_args) {
     MS_EXCEPTION_IF_NULL(item);
   }
+  return BroadCastInferShape(prim_name, input_args);
+}
+
+TypePtr MaximumInferType(const PrimitivePtr &prim, const std::vector<AbstractBasePtr> &input_args) {
+  for (const auto &item : input_args) {
+    MS_EXCEPTION_IF_NULL(item);
+  }
+  auto op_name = prim->name();
+  const int64_t kInputNum = 2;
+  (void)CheckAndConvertUtils::CheckInteger("input number", SizeToLong(input_args.size()), kGreaterEqual, kInputNum,
+                                           op_name);
   std::map<std::string, TypePtr> types;
   (void)types.emplace("x", input_args[0]->BuildType());
   (void)types.emplace("y", input_args[1]->BuildType());
-  return CheckAndConvertUtils::CheckTensorTypeSame(types, common_valid_types, prim->name());
+
+  auto type_x = input_args[0]->BuildType();
+  auto type_y = input_args[1]->BuildType();
+  MS_EXCEPTION_IF_NULL(type_x);
+  MS_EXCEPTION_IF_NULL(type_y);
+  if (type_x->isa<Complex>() || type_y->isa<Complex>()) {
+    if (type_x->type_id() == kNumberTypeComplex64 && type_y->type_id() == kNumberTypeComplex64) {
+      return type_x;
+    } else if (type_x->type_id() == kNumberTypeComplex64 && type_y->type_id() == kNumberTypeFloat32) {
+      return type_x;
+    } else if (type_x->type_id() == kNumberTypeComplex128 && type_y->type_id() == kNumberTypeComplex128) {
+      return type_x;
+    } else if (type_x->type_id() == kNumberTypeComplex128 && type_y->type_id() == kNumberTypeFloat64) {
+      return type_x;
+    } else if (type_x->type_id() == kNumberTypeFloat32 && type_y->type_id() == kNumberTypeComplex64) {
+      return type_y;
+    } else if (type_x->type_id() == kNumberTypeFloat64 && type_y->type_id() == kNumberTypeComplex128) {
+      return type_y;
+    } else {
+      MS_EXCEPTION(TypeError)
+        << "Complex math binary op expecting Tensor [complex64, complex64],[complex64, float32], [float32, "
+           "complex64],[complex128, complex128],[complex128, float64], [float64, complex128], but got["
+        << type_x->ToString() << ", " << type_y->ToString() << "].";
+    }
+  }
+  (void)CheckAndConvertUtils::CheckTensorTypeSame(types, common_valid_types_with_complex, prim->name());
+  return type_x;
 }
 }  // namespace
-
 AbstractBasePtr MaximumInfer(const abstract::AnalysisEnginePtr &, const PrimitivePtr &primitive,
                              const std::vector<AbstractBasePtr> &input_args) {
-  return std::make_shared<abstract::AbstractTensor>(InferType(primitive, input_args),
-                                                    InferShape(primitive, input_args)->shape());
+  auto infer_type = MaximumInferType(primitive, input_args);
+  auto infer_shape = MaximumInferShape(primitive, input_args);
+  return abstract::MakeAbstract(infer_shape, infer_type);
 }
-REGISTER_PRIMITIVE_C(kNameMaximum, Maximum);
+REGISTER_PRIMITIVE_EVAL_IMPL(Maximum, prim::kPrimMaximum, MaximumInfer, nullptr, true);
 }  // namespace ops
 }  // namespace mindspore

mindspore/core/ops/maximum.h

@@ -1,5 +1,5 @@
 /**
- * Copyright 2020 Huawei Technologies Co., Ltd
+ * Copyright 2020-2021 Huawei Technologies Co., Ltd
  *
  * Licensed under the Apache License, Version 2.0 (the "License");
  * you may not use this file except in compliance with the License.
@@ -39,6 +39,7 @@ class MS_CORE_API Maximum : public PrimitiveC {
 };
 AbstractBasePtr MaximumInfer(const abstract::AnalysisEnginePtr &, const PrimitivePtr &primitive,
                              const std::vector<AbstractBasePtr> &input_args);
+using kPrimMaximumPtr = std::shared_ptr<Maximum>;
 }  // namespace ops
 }  // namespace mindspore
 #endif  // MINDSPORE_CORE_OPS_MAXIMUM_H_

mindspore/python/mindspore/ops/_op_impl/tbe/__init__.py

@@ -248,6 +248,7 @@ from .select import _select_tbe
 from .pow import _pow_tbe
 from .pow_ds import _pow_ds_tbe
 from .maximum import _maximum_tbe
+from .maximum_ds import _maximum_ds_tbe
 from .minimum import _minimum_tbe
 from .minimum_ds import _minimum_ds_tbe
 from .minimum_grad import _minimum_grad_tbe

mindspore/python/mindspore/ops/_op_impl/tbe/maximum_ds.py

@@ -0,0 +1,40 @@
+# Copyright 2021 Huawei Technologies Co., Ltd
+#
+# 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.
+# ============================================================================
+
+"""Maximum op"""
+from mindspore.ops.op_info_register import op_info_register, TBERegOp, DataType
+
+maximum_op_info = TBERegOp("Maximum") \
+    .fusion_type("ELEMWISE") \
+    .async_flag(False) \
+    .binfile_name("maximum.so") \
+    .compute_cost(10) \
+    .kernel_name("maximum") \
+    .partial_flag(True) \
+    .dynamic_shape(True) \
+    .input(0, "x1", False, "required", "all") \
+    .input(1, "x2", False, "required", "all") \
+    .output(0, "y", False, "required", "all") \
+    .op_pattern("broadcast") \
+    .dtype_format(DataType.I32_None, DataType.I32_None, DataType.I32_None) \
+    .dtype_format(DataType.F16_None, DataType.F16_None, DataType.F16_None) \
+    .dtype_format(DataType.F32_None, DataType.F32_None, DataType.F32_None) \
+    .get_op_info()
+
+
+@op_info_register(maximum_op_info)
+def _maximum_ds_tbe():
+    """Maximum TBE register"""
+    return

mindspore/python/mindspore/ops/operations/math_ops.py

@@ -2611,15 +2611,6 @@ class Maximum(_MathBinaryOp):
         Float32
     """
 
-    def infer_value(self, x, y):
-        if x is not None and y is not None:
-            x = x.asnumpy()
-            y = y.asnumpy()
-            out = np.maximum(x, y)
-            out = np.array(out, x.dtype)
-            return Tensor(out)
-        return None
-
 
 class RealDiv(_MathBinaryOp):
     """