!19806 add graph kernel erf, erfc, floor, floordiv, mod, floormod, div

Merge pull request !19806 from 杨林枫/basic_graph_kernels

mindspore/_extends/graph_kernel/expanders/__init__.py

@@ -24,6 +24,8 @@ from .clip_by_norm_no_div_sum import ClipByNormNoDivSum
 from .conv2d import Conv2D
 from .complex import CAbs, CAdd, CDiv, CMul, CSub
 from .dropout_grad import DropoutGrad
+from .equal_count import EqualCount
+from .erfc import Erfc
 from .expand_dims import ExpandDims
 from .fused_adam import FusedAdam
 from .fused_adam_weight_decay import FusedAdamWeightDecay
@@ -57,4 +59,3 @@ from .squared_difference import SquaredDifference
 from .squeeze import Squeeze
 from .tanh_grad import TanhGrad
 from .tile import Tile
-from .equal_count import EqualCount

mindspore/_extends/graph_kernel/expanders/erfc.py

@@ -0,0 +1,35 @@
+# 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.
+# ===========================================================================
+"""generate json desc for erfc"""
+from ._utils import Expander
+
+
+class Erfc(Expander):
+    """Erfc expander"""
+
+    def _expand(self, graph_builder):
+        input_x = self.inputs[0]
+        result = None
+        if input_x.dtype == "float16":
+            const_one = graph_builder.value("float32", 1)
+            input_x = graph_builder.emit('Cast', [input_x], attrs={'dst_type': "float32"})
+            erf_result = graph_builder.emit('Erf', [input_x])
+            result = graph_builder.emit('Sub', [const_one, erf_result])
+            result = graph_builder.emit('Cast', [result], attrs={'dst_type': "float16"})
+            return result
+        const_one = graph_builder.value(input_x.dtype, 1)
+        erf_result = graph_builder.emit('Erf', [input_x])
+        result = graph_builder.emit('Sub', [const_one, erf_result])
+        return result

mindspore/_extends/graph_kernel/model/model.py

@@ -158,7 +158,14 @@ class PrimLib:
         'Exp': Prim(ELEMWISE),
         'Rsqrt': Prim(ELEMWISE),
         'Sqrt': Prim(ELEMWISE),
+        'Div': Prim(ELEMWISE),
+        'FloorDiv': Prim(ELEMWISE),
         'RealDiv': Prim(ELEMWISE),
+        'Mod': Prim(ELEMWISE),
+        'Floor': Prim(ELEMWISE),
+        'FloorMod': Prim(ELEMWISE),
+        'Erf': Prim(ELEMWISE),
+        'Erfc': Prim(ELEMWISE),
         'Cast': Prim(ELEMWISE),
         'Pow': Prim(ELEMWISE),
         'Minimum': Prim(ELEMWISE),

mindspore/ccsrc/backend/optimizer/graph_kernel/graph_kernel_cluster.cc

@@ -60,17 +60,8 @@ std::vector<PrimitivePtr> GetClusterableOpList() {
     prim::kPrimRealDiv,
     prim::kPrimReduceSum,
     prim::kPrimEqual,
-    prim::kPrimNotEqual,
-    prim::kPrimLogicalAnd,
-    prim::kPrimLogicalOr,
-    prim::kPrimLogicalNot,
     prim::kPrimAssign,
     prim::kPrimInplaceAssign,
-    prim::kPrimAtan,
-    prim::kPrimAtan2,
-    prim::kPrimExpm1,
-    prim::kPrimAsinh,
-    prim::kPrimAcosh,
 #if ENABLE_D
     prim::kPrimMatMul,
     prim::KPrimTransData,
@@ -88,6 +79,21 @@ std::vector<PrimitivePtr> GetClusterableOpList() {
     prim::kPrimGreaterEqual,
     prim::kPrimLessEqual,
     prim::kPrimSelect,
+    prim::kPrimAtan,
+    prim::kPrimAtan2,
+    prim::kPrimExpm1,
+    prim::kPrimAsinh,
+    prim::kPrimAcosh,
+    prim::kPrimDiv,
+    prim::kPrimFloorDiv,
+    prim::kPrimMod,
+    prim::kPrimFloor,
+    prim::kPrimFloorMod,
+    prim::kPrimErf,
+    prim::kPrimNotEqual,
+    prim::kPrimLogicalAnd,
+    prim::kPrimLogicalOr,
+    prim::kPrimLogicalNot,
 #endif
   };
   const auto &flags = context::GraphKernelFlags::GetInstance();

mindspore/ccsrc/backend/optimizer/graph_kernel/graph_kernel_expander.cc

@@ -60,6 +60,7 @@ std::vector<PrimitivePtr> GetExpandOps() {
     prim::kPrimTile,
     prim::kPrimMatMul,
     prim::kPrimBatchMatMul,
+    prim::kPrimErfc,
 #if ENABLE_D
     prim::kPrimSqrtGrad,
     prim::kPrimClipByNormNoDivSum,

mindspore/core/base/core_ops.h

@@ -622,6 +622,7 @@ inline const PrimitivePtr kPrimLayerNormFusion = std::make_shared<Primitive>("La
 inline const PrimitivePtr kPrimDType = std::make_shared<Primitive>("DType");
 inline const PrimitivePtr kPrimDivFusion = std::make_shared<Primitive>("DivFusion");
 inline const PrimitivePtr kPrimErf = std::make_shared<Primitive>("Erf");
+inline const PrimitivePtr kPrimErfc = std::make_shared<Primitive>("Erfc");
 inline const PrimitivePtr kPrimSplice = std::make_shared<Primitive>("Splice");
 inline const PrimitivePtr kPrimAffine = std::make_shared<Primitive>("Affine");
 

tests/st/ops/graph_kernel/test_erf_erfc.py

@@ -0,0 +1,73 @@
+# 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.
+# ============================================================================
+import numpy as np
+import pytest
+import mindspore.context as context
+import mindspore.nn as nn
+from mindspore import Tensor
+from mindspore.ops import operations as P
+
+
+class ErfNet(nn.Cell):
+    def __init__(self):
+        super(ErfNet, self).__init__()
+        self.erf = P.Erf()
+
+    def construct(self, x):
+        return self.erf(x)
+
+class ErfcNet(nn.Cell):
+    def __init__(self):
+        super(ErfcNet, self).__init__()
+        self.erfc = P.Erfc()
+
+    def construct(self, x):
+        return self.erfc(x)
+
+def get_output(net, inp, enable_graph_kernel=False):
+    context.set_context(enable_graph_kernel=enable_graph_kernel)
+    output = net()(inp)
+    return output
+
+def basic_test(net, datatype):
+    inp = Tensor(np.random.random((2, 3)).astype(datatype))
+    expect = get_output(net, inp, False)
+    output = get_output(net, inp, True)
+    expect_np = expect.asnumpy().copy()
+    output_np = output.asnumpy().copy()
+    assert np.allclose(expect_np, output_np, 1.e-4, 1.e-7)
+
+    inp = Tensor(np.random.random((2, 3, 3, 4, 5)).astype(datatype))
+    expect = get_output(net, inp, False)
+    output = get_output(net, inp, True)
+    expect_np = expect.asnumpy().copy()
+    output_np = output.asnumpy().copy()
+    assert np.allclose(expect_np, output_np, 1.e-4, 1.e-7)
+
+@pytest.mark.level0
+@pytest.mark.platform_x86_gpu_training
+@pytest.mark.env_onecard
+def test_gpu_fp16():
+    context.set_context(mode=context.GRAPH_MODE, device_target="GPU")
+    basic_test(ErfNet, np.float16)
+    basic_test(ErfcNet, np.float16)
+
+@pytest.mark.level0
+@pytest.mark.platform_x86_gpu_training
+@pytest.mark.env_onecard
+def test_gpu_fp32():
+    context.set_context(mode=context.GRAPH_MODE, device_target="GPU")
+    basic_test(ErfNet, np.float32)
+    basic_test(ErfcNet, np.float32)