!1129 Add op bounding box encoder

Merge pull request !1129 from zhaozhenlong/op/bounding-box-encoder

mindspore/_extends/parallel_compile/tbe_compiler/compiler.py

@@ -114,6 +114,9 @@ def build_op(build_type, json_str):
             return get_op_pattern()
 
         # call function
+        if kernel_name[0:19] == "bounding_box_encode":
+            return op_func(*inputs_args, *outputs_args, *attrs_args, kernel_name_val=kernel_name)
+
         return op_func(*inputs_args, *outputs_args, *attrs_args, kernel_name=kernel_name)
 
     except Exception as e:

mindspore/ccsrc/kernel/tbe/tbe_adapter.cc

@@ -82,7 +82,11 @@ static std::map<string, string> tbe_func_adapter_map = {
   {"resize_bilinear_grad", "resize_bilinear_v2_grad"},
   {"adam", "apply_adam"},
   {"r_oi_align", "roi_align"},
-  {"r_oi_align_grad", "roi_align_grad"}};
+  {"r_oi_align_grad", "roi_align_grad"},
+  {"i_ou", "iou"},
+  {"s_gd", "sgd"},
+  {"l_ars_update", "lars_v2_update"},
+  {"n_ms_with_mask", "nms_with_mask"}};
 
 void TbeAdapter::NormalizeFuncName(std::string *func_name) {
   if (func_name == nullptr) {

mindspore/ccsrc/kernel/tbe/tbe_kernel_build.cc

@@ -430,6 +430,18 @@ void TbeKernelJsonCreator::ParseAttrValue(const std::string &type, const mindspo
       attr_value = GetValue<std::vector<int>>(value);
     }
     (*attr_obj)["value"] = attr_value;
+  } else if (type == "listFloat") {
+    std::vector<float> attr_value;
+    auto value_type = value->type();
+    MS_EXCEPTION_IF_NULL(value_type);
+    auto value_type_str = value_type->ToString();
+    if (value_type_str == "float") {
+      float data = GetValue<float>(value);
+      attr_value.push_back(data);
+    } else {
+      attr_value = GetValue<std::vector<float>>(value);
+    }
+    (*attr_obj)["value"] = attr_value;
   } else if (type == "listListInt") {
     auto attr_value = GetValue<std::vector<std::vector<int>>>(value);
     (*attr_obj)["value"] = attr_value;

mindspore/ops/_op_impl/tbe/__init__.py

@@ -171,3 +171,11 @@ from .resize_bilinear_grad import _resize_bilinear_grad_tbe
 from .flatten import _flatten_tbe
 from .roi_align import _roi_align_tbe
 from .roi_align_grad import _roi_align_grad_tbe
+from .bounding_box_decode import _bounding_box_decode_tbe
+from .bounding_box_encode import _bounding_box_encode_tbe
+from .check_valid import _check_valid_tbe
+from .iou import _iou_tbe
+from .nms_with_mask import nms_with_mask_op_info
+from .random_choice_with_mask import random_choice_with_mask_op_info
+from .sgd import sgd_op_info
+from .lars_update import lars_update_op_info

mindspore/ops/_op_impl/tbe/bounding_box_decode.py

+# Copyright 2020 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.
+# ============================================================================
+
+"""BoundingBoxDecode op"""
+from mindspore.ops.op_info_register import op_info_register, TBERegOp, DataType
+
+bounding_box_decode_op_info = TBERegOp("BoundingBoxDecode") \
+    .fusion_type("OPAQUE") \
+    .async_flag(False) \
+    .binfile_name("bounding_box_decode.so") \
+    .compute_cost(10) \
+    .kernel_name("bounding_box_decode") \
+    .partial_flag(True) \
+    .attr("means", "optional", "listFloat", "all") \
+    .attr("stds", "optional", "listFloat", "all") \
+    .attr("max_shape", "optional", "listInt", "all") \
+    .attr("wh_ratio_clip", "optional", "float", "all") \
+    .input(0, "rois", False, "required", "all") \
+    .input(1, "deltas", False, "required", "all") \
+    .output(0, "bboxes", False, "required", "all") \
+    .dtype_format(DataType.F16_Default, DataType.F16_Default, DataType.F16_Default) \
+    .dtype_format(DataType.F32_Default, DataType.F32_Default, DataType.F32_Default) \
+    .get_op_info()
+
+
+@op_info_register(bounding_box_decode_op_info)
+def _bounding_box_decode_tbe():
+    """BoundingBoxDecode TBE register"""
+    return

mindspore/ops/_op_impl/tbe/bounding_box_encode.py

+# Copyright 2020 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.
+# ============================================================================
+
+"""BoundingBoxEncode op"""
+from mindspore.ops.op_info_register import op_info_register, TBERegOp, DataType
+
+bounding_box_encode_op_info = TBERegOp("BoundingBoxEncode") \
+    .fusion_type("OPAQUE") \
+    .async_flag(False) \
+    .binfile_name("bounding_box_encode.so") \
+    .compute_cost(10) \
+    .kernel_name("bounding_box_encode") \
+    .partial_flag(True) \
+    .attr("means", "optional", "listFloat", "all") \
+    .attr("stds", "optional", "listFloat", "all") \
+    .input(0, "anchor_box", False, "required", "all") \
+    .input(1, "ground_truth_box", False, "required", "all") \
+    .output(0, "delats", False, "required", "all") \
+    .dtype_format(DataType.F16_Default, DataType.F16_Default, DataType.F16_Default) \
+    .get_op_info()
+
+
+@op_info_register(bounding_box_encode_op_info)
+def _bounding_box_encode_tbe():
+    """BoundingBoxEncode TBE register"""
+    return

mindspore/ops/_op_impl/tbe/check_valid.py

+# Copyright 2020 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.
+# ============================================================================
+
+"""CheckValid op"""
+from mindspore.ops.op_info_register import op_info_register, TBERegOp, DataType
+
+check_valid_op_info = TBERegOp("CheckValid") \
+    .fusion_type("OPAQUE") \
+    .async_flag(False) \
+    .binfile_name("check_valid.so") \
+    .compute_cost(10) \
+    .kernel_name("check_valid") \
+    .partial_flag(True) \
+    .input(0, "bbox_tensor", False, "required", "all") \
+    .input(1, "img_tas", False, "required", "all") \
+    .output(0, "valid_tensor", False, "required", "all") \
+    .dtype_format(DataType.F16_Default, DataType.F16_Default, DataType.I8_Default) \
+    .dtype_format(DataType.F16_Default, DataType.F16_Default, DataType.BOOL_Default) \
+    .get_op_info()
+
+
+@op_info_register(check_valid_op_info)
+def _check_valid_tbe():
+    """CheckValid TBE register"""
+    return

mindspore/ops/_op_impl/tbe/iou.py

+# Copyright 2020 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.
+# ============================================================================
+
+"""Iou op"""
+from mindspore.ops.op_info_register import op_info_register, TBERegOp, DataType
+
+iou_op_info = TBERegOp("IOU") \
+    .fusion_type("OPAQUE") \
+    .async_flag(False) \
+    .binfile_name("iou.so") \
+    .compute_cost(10) \
+    .kernel_name("iou") \
+    .partial_flag(True) \
+    .attr("mode", "required", "str", "all") \
+    .input(0, "bboxes", False, "required", "all") \
+    .input(1, "gtboxes", False, "required", "all") \
+    .output(0, "overlap", False, "required", "all") \
+    .dtype_format(DataType.F16_Default, DataType.F16_Default, DataType.F16_Default) \
+    .get_op_info()
+
+
+@op_info_register(iou_op_info)
+def _iou_tbe():
+    """Iou TBE register"""
+    return

mindspore/ops/_op_impl/tbe/lars_update.py

+# Copyright 2020 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.
+# ============================================================================
+
+"""LarsUpdate op"""
+from mindspore.ops.op_info_register import op_info_register, TBERegOp, DataType
+
+lars_update_op_info = TBERegOp("LARSUpdate") \
+    .fusion_type("OPAQUE") \
+    .async_flag(False) \
+    .binfile_name("lars_v2_update.so") \
+    .compute_cost(10) \
+    .kernel_name("lars_v2_update") \
+    .partial_flag(True) \
+    .attr("hyperpara", "optional", "float", "all") \
+    .attr("epsilon", "optional", "float", "all") \
+    .attr("use_clip", "optional", "bool", "all") \
+    .input(0, "w", False, "required", "all") \
+    .input(1, "g", False, "required", "all") \
+    .input(2, "w_square_sum", False, "required", "all") \
+    .input(3, "g_square_sum", False, "required", "all") \
+    .input(4, "weight_decay", False, "required", "all") \
+    .input(5, "learning_rate", False, "required", "all") \
+    .output(0, "g_new", False, "required", "all") \
+    .dtype_format(DataType.F32_FracZ, DataType.F32_FracZ, DataType.F32_Default, DataType.F32_Default,
+                  DataType.F32_Default, DataType.F32_Default, DataType.F32_FracZ) \
+    .dtype_format(DataType.F32_C1HWNCoC0, DataType.F32_C1HWNCoC0, DataType.F32_Default, DataType.F32_Default,
+                  DataType.F32_Default, DataType.F32_Default, DataType.F32_C1HWNCoC0) \
+    .dtype_format(DataType.F32_5HD, DataType.F32_5HD, DataType.F32_Default, DataType.F32_Default,
+                  DataType.F32_Default, DataType.F32_Default, DataType.F32_5HD) \
+    .dtype_format(DataType.F32_Default, DataType.F32_Default, DataType.F32_Default, DataType.F32_Default,
+                  DataType.F32_Default, DataType.F32_Default, DataType.F32_Default) \
+    .get_op_info()
+
+
+@op_info_register(lars_update_op_info)
+def _lars_update_tbe():
+    """LarsUpdate TBE register"""
+    return

mindspore/ops/_op_impl/tbe/nms_with_mask.py

+# Copyright 2020 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.
+# ============================================================================
+
+"""NMSWithMask op"""
+from mindspore.ops.op_info_register import op_info_register, TBERegOp, DataType
+
+nms_with_mask_op_info = TBERegOp("NMSWithMask") \
+    .fusion_type("OPAQUE") \
+    .async_flag(False) \
+    .binfile_name("nms_with_mask.so") \
+    .compute_cost(10) \
+    .kernel_name("nms_with_mask") \
+    .partial_flag(True) \
+    .attr("iou_threshold", "optional", "float", "all") \
+    .input(0, "box_scores", False, "required", "all") \
+    .output(0, "selected_boxes", False, "required", "all") \
+    .output(0, "selected_idx", False, "required", "all") \
+    .output(0, "selected_mask", False, "required", "all") \
+    .dtype_format(DataType.F16_Default, DataType.F16_Default, DataType.I32_Default, DataType.U8_Default) \
+    .dtype_format(DataType.F16_Default, DataType.F16_Default, DataType.I32_Default, DataType.BOOL_Default) \
+    .get_op_info()
+
+
+@op_info_register(nms_with_mask_op_info)
+def _nms_with_mask_tbe():
+    """NMSWithMask TBE register"""
+    return

mindspore/ops/_op_impl/tbe/random_choice_with_mask.py

+# Copyright 2020 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.
+# ============================================================================
+
+"""RandomChoiceWithMask op"""
+from mindspore.ops.op_info_register import op_info_register, TBERegOp, DataType
+
+random_choice_with_mask_op_info = TBERegOp("RandomChoiceWithMask") \
+    .fusion_type("OPAQUE") \
+    .async_flag(False) \
+    .binfile_name("random_choice_with_mask.so") \
+    .compute_cost(10) \
+    .kernel_name("random_choice_with_mask") \
+    .partial_flag(True) \
+    .attr("max_shape", "optional", "listInt", "all") \
+    .attr("means", "optional", "listFloat", "all") \
+    .attr("stds", "optional", "listFloat", "all") \
+    .attr("wh_ratio_clip", "optional", "float", "all") \
+    .input(0, "rois", False, "required", "all") \
+    .input(1, "deltas", False, "required", "all") \
+    .output(0, "bboxes", False, "required", "all") \
+    .dtype_format(DataType.F16_Default, DataType.F16_Default, DataType.F16_Default) \
+    .dtype_format(DataType.F32_Default, DataType.F32_Default, DataType.F32_Default) \
+    .get_op_info()
+
+
+@op_info_register(random_choice_with_mask_op_info)
+def _random_choice_with_mask_tbe():
+    """RandomChoiceWithMask TBE register"""
+    return

mindspore/ops/_op_impl/tbe/sgd.py

+# Copyright 2020 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.
+# ============================================================================
+
+"""SGD op"""
+from mindspore.ops.op_info_register import op_info_register, TBERegOp, DataType
+
+sgd_op_info = TBERegOp("SGD") \
+    .fusion_type("OPAQUE") \
+    .async_flag(False) \
+    .binfile_name("sgd.so") \
+    .compute_cost(10) \
+    .kernel_name("sgd") \
+    .partial_flag(True) \
+    .attr("dampening", "optional", "float", "all") \
+    .attr("weight_decay", "optional", "float", "all") \
+    .attr("nesterov", "optional", "bool", "all") \
+    .input(0, "parameters", False, "required", "all") \
+    .input(1, "gradient", False, "required", "all") \
+    .input(2, "learning_rate", False, "required", "all") \
+    .input(3, "accum", False, "required", "all") \
+    .input(4, "momentum", False, "required", "all") \
+    .input(5, "stat", False, "required", "all") \
+    .output(0, "parameters", False, "required", "all") \
+    .dtype_format(DataType.F16_5HD, DataType.F16_5HD, DataType.F16_Default, DataType.F16_5HD,
+                  DataType.F16_Default, DataType.F16_5HD, DataType.F16_5HD) \
+    .dtype_format(DataType.F16_Default, DataType.F16_Default, DataType.F16_Default, DataType.F16_Default,
+                  DataType.F16_Default, DataType.F16_Default, DataType.F16_Default) \
+    .dtype_format(DataType.F16_FracZ, DataType.F16_FracZ, DataType.F16_Default, DataType.F16_FracZ,
+                  DataType.F16_Default, DataType.F16_FracZ, DataType.F16_FracZ) \
+    .dtype_format(DataType.F32_5HD, DataType.F32_5HD, DataType.F32_Default, DataType.F32_5HD,
+                  DataType.F32_Default, DataType.F32_5HD, DataType.F32_5HD) \
+    .dtype_format(DataType.F32_Default, DataType.F32_Default, DataType.F32_Default, DataType.F32_Default,
+                  DataType.F32_Default, DataType.F32_Default, DataType.F32_Default) \
+    .dtype_format(DataType.F32_FracZ, DataType.F32_FracZ, DataType.F32_Default, DataType.F32_FracZ,
+                  DataType.F32_Default, DataType.F32_FracZ, DataType.F32_FracZ) \
+    .get_op_info()
+
+
+@op_info_register(sgd_op_info)
+def _sgd_tbe():
+    """SGD TBE register"""
+    return

mindspore/ops/operations/math_ops.py

@@ -16,6 +16,7 @@
 """Operators for math."""
 
 import numpy as np
+from ... import context
 from ..._c_expression import signature_rw as sig_rw
 from ..._c_expression import signature_kind as sig_kind
 from ..._c_expression import signature_dtype as sig_dtype
@@ -1950,12 +1951,16 @@ class NMSWithMask(PrimitiveWithInfer):
         """Init NMSWithMask"""
         validator.check_value_type("iou_threshold", iou_threshold, [float], self.name)
         self.init_prim_io_names(inputs=['bboxes'], outputs=['selected_boxes', 'selected_idx', 'selected_mask'])
+        self.is_ge = context.get_context("enable_ge")
 
     def infer_shape(self, bboxes_shape):
         cls_name = self.name
         validator.check_integer("bboxes rank", len(bboxes_shape), 2, Rel.EQ, cls_name)
         validator.check_integer("bboxes.shape()[0]", bboxes_shape[0], 0, Rel.GT, cls_name)
-        validator.check_integer("bboxes.shape()[1]", bboxes_shape[1], 5, Rel.EQ, cls_name)
+        if not self.is_ge:
+            validator.check_integer("bboxes.shape()[1]", bboxes_shape[1], 8, Rel.EQ, cls_name)
+        else:
+            validator.check_integer("bboxes.shape()[1]", bboxes_shape[1], 5, Rel.EQ, cls_name)
         num = bboxes_shape[0]
         return (bboxes_shape, (num,), (num,))
 

mindspore/ops/operations/other_ops.py

@@ -175,10 +175,10 @@ class CheckValid(PrimitiveWithInfer):
         self.init_prim_io_names(inputs=['bboxes', 'img_metas'], outputs=['output'])
 
     def infer_shape(self, bboxes_shape, metas_shape):
-        validator.check_integer("bboxes rank", len(bboxes_shape), 2, Rel.EQ, self.name)
-        validator.check_integer("bboxes_shape[-1]", bboxes_shape[-1], 4, Rel.EQ, self.name)
-        validator.check_integer("img_metas rank", len(metas_shape), 1, Rel.EQ, self.name)
-        validator.check_integer("img_metas shape[0]", metas_shape[0], 3, Rel.EQ, self.name)
+        validator.check("bboxes rank", len(bboxes_shape), "", 2, Rel.EQ, self.name)
+        validator.check("bboxes_shape[-1]", bboxes_shape[-1], "", 4, Rel.EQ, self.name)
+        validator.check("img_metas rank", len(metas_shape), "", 1, Rel.EQ, self.name)
+        validator.check("img_metas shape[0]", metas_shape[0], "", 3, Rel.EQ, self.name)
         return bboxes_shape[:-1]
 
     def infer_dtype(self, bboxes_type, metas_type):

tests/mindspore_test_framework/utils/block_util.py

@@ -188,6 +188,10 @@ class InputOpNet(nn.Cell):
         x = self.op(x1, x2, x3, x4, self.c1)
         return x
 
+    def construct4_c2(self, x1, x2, x3, x4):
+        x = self.op(x1, x2, x3, x4, self.c1, self.c2)
+        return x
+
     def construct4_c4(self, x1, x2, x3, x4):
         x = self.op(x1, x2, x3, x4, self.c1, self.c2, self.c3, self.c4)
         return x

tests/ut/python/ops/test_ops.py

@@ -950,6 +950,17 @@ test_case_nn_ops = [
         'desc_inputs': [[1, 1, 2, 2], [1, 5]],
         'desc_bprop': [[1, 1, 2, 2]],
         'skip': ['backward']}),
+    ('LARSUpdate', {
+        'block': P.LARSUpdate(1e-05, 0.001, False),
+        'desc_const': [0.0, 0.001],
+        'desc_inputs': [[3, 3], [3, 3], [3, 3], [3, 3]],
+        'desc_bprop': [3, 3],
+        'skip': ['backward']}),
+    ('SGD', {
+       'block': P.SGD(0.0, 0.0, False),
+        'desc_inputs': [[3, 3], [3, 3], Tensor(0.001, mstype.float32), [3, 3], Tensor(0.1, mstype.float32), [3, 3]],
+        'desc_bprop': [3, 3],
+        'skip': ['backward']}),
 ]
 
 test_case_array_ops = [