!12914 [GraphKernel]expander lamb_apply_weight_assign

From: @wenfangpei
Reviewed-by: @anyrenwei,@gaoxiong1,@gaoxiong1
Signed-off-by: @anyrenwei

mindspore/_extends/graph_kernel/expanders/__init__.py

@@ -46,3 +46,4 @@ from .square import Square
 from .tanh_grad import TanhGrad
 from .tile import Tile
 from .lamb_apply_optimizer_assign import LambApplyOptimizerAssign
+from .lamb_apply_weight_assign import LambApplyWeightAssign

mindspore/_extends/graph_kernel/expanders/lamb_apply_weight_assign.py

@@ -0,0 +1,56 @@
+# 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 LambApplyWeightAssign"""
+from ._utils import Expander, ExpanderInfoValidator as VLD
+
+@VLD.check_all_formats_same
+class LambApplyWeightAssign(Expander):
+    """LambApplyWeightAssign expander"""
+
+    def _expand(self, graph_builder):
+
+        w_norm, g_norm, input_lr, update, input_param = self.inputs
+        # ratio
+        const_zero = graph_builder.value(g_norm.dtype, 0)
+        const_one = graph_builder.value(g_norm.dtype, 1)
+        dtype = update.dtype
+
+        g_norm_greater_res = graph_builder.emit('Greater', [g_norm, const_zero])
+        g_norm_greater_res_float = graph_builder.emit('Cast', [g_norm_greater_res], attrs={'dst_type': dtype})
+
+        w_norm_g_norm = graph_builder.emit('RealDiv', [w_norm, g_norm])
+        # select
+        g_norm_greater_res_neg = graph_builder.emit('Neg', [g_norm_greater_res_float])
+        g_norm_greater_res_f = graph_builder.emit('Add', [g_norm_greater_res_neg, const_one])
+        g_norm_value_1 = graph_builder.emit('Mul', [g_norm_greater_res_float, w_norm_g_norm])
+        g_norm_value = graph_builder.emit('Add', [g_norm_value_1, g_norm_greater_res_f])
+
+        w_norm_greater_res = graph_builder.emit('Greater', [w_norm, const_zero])
+        w_norm_greater_res_float = graph_builder.emit('Cast', [w_norm_greater_res], attrs={'dst_type': dtype})
+
+        # select
+        w_norm_greater_res_neg = graph_builder.emit('Neg', [w_norm_greater_res_float])
+        w_norm_greater_res_f = graph_builder.emit('Add', [w_norm_greater_res_neg, const_one])
+        w_norm_value_1 = graph_builder.emit('Mul', [w_norm_greater_res_float, g_norm_value])
+        ratio = graph_builder.emit('Add', [w_norm_value_1, w_norm_greater_res_f])
+
+        # ratio * input_lr * update
+        update_with_ir = graph_builder.emit('Mul', [update, input_lr])
+        ratio_update_with_ir = graph_builder.emit('Mul', [update_with_ir, ratio])
+
+        # input_param - ratio_update_with_ir
+        next_param = graph_builder.emit('Sub', [input_param, ratio_update_with_ir])
+
+        return [next_param]

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

@@ -39,7 +39,8 @@ namespace mindspore {
 namespace opt {
 namespace {
 constexpr size_t kAssignInputIdx = 1;
-constexpr size_t kLambInputIdx = 12;
+constexpr size_t kLambOptimizerInputIdx = 12;
+constexpr size_t kLambWeightInputIdx = 4;
 
 std::vector<PrimitivePtr> GetExpandOps() {
   std::vector<PrimitivePtr> expand_ops = {
@@ -51,6 +52,7 @@ std::vector<PrimitivePtr> GetExpandOps() {
     prim::kPrimSqrtGrad,
     prim::kPrimClipByNormNoDivSum,
     prim::kLambApplyOptimizerAssign,
+    prim::kLambApplyWeightAssign,
 #elif ENABLE_GPU
     prim::kPrimBiasAdd,
     prim::kPrimBiasAddGrad,
@@ -176,7 +178,8 @@ ExpanderPtr GraphKernelExpander::GetExpander(const AnfNodePtr &node) {
     {prim::kPrimDropout, std::make_shared<DropoutExpander>()},
     {prim::kPrimAssignAdd, std::make_shared<OpUMonadExpander>(kAssignInputIdx)},
     {prim::kPrimAssignSub, std::make_shared<OpUMonadExpander>(kAssignInputIdx)},
-    {prim::kLambApplyOptimizerAssign, std::make_shared<OpUMonadExpander>(kLambInputIdx)},
+    {prim::kLambApplyOptimizerAssign, std::make_shared<OpUMonadExpander>(kLambOptimizerInputIdx)},
+    {prim::kLambApplyWeightAssign, std::make_shared<OpUMonadExpander>(kLambWeightInputIdx)},
   };
 
   for (auto &e : expanders) {

mindspore/core/base/core_ops.h

@@ -305,6 +305,8 @@ inline const PrimitivePtr kPrimTensorMove = std::make_shared<Primitive>("TensorM
 inline const PrimitivePtr kPrimL2Normalize = std::make_shared<Primitive>("L2Normalize");
 inline const PrimitivePtr kPrimCustomExtractFeatures = std::make_shared<Primitive>("CustomExtractFeatures");
 inline const PrimitivePtr kLambApplyOptimizerAssign = std::make_shared<Primitive>("LambApplyOptimizerAssign");
+inline const PrimitivePtr kLambApplyWeightAssign = std::make_shared<Primitive>("LambApplyWeightAssign");
+
 // Comm ops
 inline const PrimitivePtr kPrimMirror = std::make_shared<Primitive>("_MirrorOperator");
 inline const PrimitivePtr kPrimMirrorMiniStep = std::make_shared<Primitive>("_MirrorMiniStepOperator");

tests/st/ops/graph_kernel/test_lamb_apply_weight_assign.py

@@ -0,0 +1,58 @@
+# 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 Net(nn.Cell):
+    def __init__(self):
+        super(Net, self).__init__()
+        self.lamb_apply_weight_assign = P.LambApplyWeightAssign()
+
+    def construct(self, w_norm, g_norm, lr, update, param):
+        return self.lamb_apply_weight_assign(w_norm, g_norm, lr, update, param)
+
+def get_output(w_norm, g_norm, lr, update, param, enable_graph_kernel=False):
+    context.set_context(enable_graph_kernel=enable_graph_kernel)
+    opt = Net()
+    output = opt(Tensor(w_norm), Tensor(g_norm), Tensor(lr), Tensor(update), Tensor(param))
+    return output
+
+def lamb_apply_weight_assign():
+
+    w_norm = np.array([0.11]).astype(np.float32)
+    g_norm = np.array([1.2]).astype(np.float32)
+    lr = np.array([0.012]).astype(np.float32)
+    update = np.array([0.01, 0.03, 0.05]).astype(np.float32)
+    param = np.array([1, 3, 5]).astype(np.float32)
+
+    expect = get_output(w_norm, g_norm, lr, update, param, False)
+    output = get_output(w_norm, g_norm, lr, update, param, True)
+
+    assert np.allclose(output.asnumpy(), expect.asnumpy())
+
+@pytest.mark.level0
+@pytest.mark.platform_arm_ascend_training
+@pytest.mark.platform_x86_ascend_training
+@pytest.mark.env_onecard
+def test_lamb_apply_weight_assign_ascend():
+    context.set_context(mode=context.GRAPH_MODE, device_target="Ascend")
+    lamb_apply_weight_assign()