Skip to content

X
X2Paddle

PaddlePaddle / X2Paddle
大约 1 年 前同步成功

通知 328
Star 698
Fork 167
X
X2Paddle
未验证 提交 4a2df7e1 编写于 作者: W WJJ1995 提交者： GitHub
浏览文件
操作

Add MMdetection FCOS && Yolov3 support (#595) 

* add BatchToSpaceND and SpaceToBatchND op convert

* add Less and fixed Resize RoiAlign Greater Tile op

* add nms custom layer

* fix topk int64 bug

* update model_zoo

* deal with comments
Co-authored-by: Nchanningss <chen_lingchi@163.com>
上级 3903bade
隐藏空白更改
内联 并排
Showing with 868 addition and 662 deletion
docs/introduction/x2paddle_model_zoo.md
浏览文件 @ 4a2df7e1
...@@ -18,12 +18,12 @@ ...@@ -18,12 +18,12 @@
| ResNet_V2_101 | [code](https://github.com/tensorflow/models/tree/master/research/slim/nets) | | ResNet_V2_101 | [code](https://github.com/tensorflow/models/tree/master/research/slim/nets) |
| UNet | [code1](https://github.com/jakeret/tf_unet)/[code2](https://github.com/lyatdawn/Unet-Tensorflow) | | UNet | [code1](https://github.com/jakeret/tf_unet)/[code2](https://github.com/lyatdawn/Unet-Tensorflow) |
| MTCNN | [code](https://github.com/AITTSMD/MTCNN-Tensorflow) | | MTCNN | [code](https://github.com/AITTSMD/MTCNN-Tensorflow) |
| YOLO-V3| [code](https://github.com/YunYang1994/tensorflow-yolov3) | | YOLO-V3| [code](https://github.com/YunYang1994/tensorflow-yolov3) |
| FALSR | [code](https://github.com/xiaomi-automl/FALSR) | | FALSR | [code](https://github.com/xiaomi-automl/FALSR) |
| DCSCN | [code](https://modelzoo.co/model/dcscn-super-resolution) | | DCSCN | [code](https://modelzoo.co/model/dcscn-super-resolution) |
| Bert（albert） | [code](https://github.com/google-research/albert#pre-trained-models) | | Bert（albert） | [code](https://github.com/google-research/albert#pre-trained-models) |
| Bert（chinese_L-12_H-768_A-12） | [code](https://github.com/google-research/bert#pre-trained-models) | | Bert（chinese_L-12_H-768_A-12） | [code](https://github.com/google-research/bert#pre-trained-models) |
| Bert（multi_cased_L-12_H-768_A-12） | [code](https://github.com/google-research/bert#pre-trained-models) | | Bert（multi_cased_L-12_H-768_A-12） | [code](https://github.com/google-research/bert#pre-trained-models) |
## Caffe预测模型 ## Caffe预测模型
...@@ -71,7 +71,9 @@ ...@@ -71,7 +71,9 @@
|Ultra-Light-Fast-Generic-Face-Detector-1MB| [onnx_model](https://github.com/Linzaer/Ultra-Light-Fast-Generic-Face-Detector-1MB/tree/master/models/onnx)|9 | |Ultra-Light-Fast-Generic-Face-Detector-1MB| [onnx_model](https://github.com/Linzaer/Ultra-Light-Fast-Generic-Face-Detector-1MB/tree/master/models/onnx)|9 |
|BERT| [pytorch(huggingface)](https://github.com/huggingface/transformers/blob/master/notebooks/04-onnx-export.ipynb)|11|转换时需指定input shape，见[文档Q3](../inference_model_convertor/FAQ.md)| |BERT| [pytorch(huggingface)](https://github.com/huggingface/transformers/blob/master/notebooks/04-onnx-export.ipynb)|11|转换时需指定input shape，见[文档Q3](../inference_model_convertor/FAQ.md)|
|GPT2| [pytorch(huggingface)](https://github.com/huggingface/transformers/blob/master/notebooks/04-onnx-export.ipynb)|11|转换时需指定input shape，见[文档Q3](../inference_model_convertor/FAQ.md)| |GPT2| [pytorch(huggingface)](https://github.com/huggingface/transformers/blob/master/notebooks/04-onnx-export.ipynb)|11|转换时需指定input shape，见[文档Q3](../inference_model_convertor/FAQ.md)|
|CifarNet | [tensorflow](https://github.com/tensorflow/models/blob/master/research/slim/nets/cifarnet.py)|9|| |CifarNet | [tensorflow](https://github.com/tensorflow/models/blob/master/research/slim/nets/cifarnet.py)|9|
|Fcos | [pytorch(mmdetection)](https://github.com/open-mmlab/mmdetection/blob/master/configs/fcos/fcos_r50_caffe_fpn_gn-head_1x_coco.py)|11|
|Yolov3 | [pytorch(mmdetection)](https://github.com/open-mmlab/mmdetection/blob/master/configs/yolo/yolov3_d53_mstrain-608_273e_coco.py)|11||
## PyTorch预测模型 ## PyTorch预测模型
......
x2paddle/op_mapper/dygraph/onnx2paddle/onnx_custom_layer/__init__.py
浏览文件 @ 4a2df7e1
...@@ -12,10 +12,10 @@ ...@@ -12,10 +12,10 @@
# See the License for the specific language governing permissions and # See the License for the specific language governing permissions and
# limitations under the License. # limitations under the License.
from .one_hot import OneHot from .one_hot import OneHot
from .pad_two_input import PadWithTwoInput from .pad_two_input import PadWithTwoInput
from .pad_all_dim2 import PadAllDim2 from .pad_all_dim2 import PadAllDim2
from .pad_all_dim4 import PadAllDim4 from .pad_all_dim4 import PadAllDim4
from .pad_all_dim4_one_input import PadAllDim4WithOneInput from .pad_all_dim4_one_input import PadAllDim4WithOneInput
from .lrn import LocalResponseNorm from .lrn import LocalResponseNorm
from .nms import NMS
x2paddle/op_mapper/dygraph/onnx2paddle/onnx_custom_layer/nms.py 0 → 100644
浏览文件 @ 4a2df7e1
# Copyright (c) 2020 PaddlePaddle Authors. All Rights Reserved.
#
# 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 paddle
from paddle.fluid import core
from paddle.fluid.framework import Variable, in_dygraph_mode
from paddle.fluid.layer_helper import LayerHelper
def multiclass_nms(bboxes,
scores,
score_threshold,
nms_top_k,
keep_top_k,
nms_threshold=0.3,
normalized=True,
nms_eta=1.,
background_label=-1,
return_index=False,
return_rois_num=True,
rois_num=None,
name=None):
helper = LayerHelper('multiclass_nms3', **locals())
if in_dygraph_mode():
attrs = ('background_label', background_label, 'score_threshold',
score_threshold, 'nms_top_k', nms_top_k, 'nms_threshold',
nms_threshold, 'keep_top_k', keep_top_k, 'nms_eta', nms_eta,
'normalized', normalized)
output, index, nms_rois_num = core.ops.multiclass_nms3(bboxes, scores,
rois_num, *attrs)
if not return_index:
index = None
return output, nms_rois_num, index
else:
output = helper.create_variable_for_type_inference(dtype=bboxes.dtype)
index = helper.create_variable_for_type_inference(dtype='int')
inputs = {'BBoxes': bboxes, 'Scores': scores}
outputs = {'Out': output, 'Index': index}
if rois_num is not None:
inputs['RoisNum'] = rois_num
if return_rois_num:
nms_rois_num = helper.create_variable_for_type_inference(
dtype='int32')
outputs['NmsRoisNum'] = nms_rois_num
helper.append_op(
type="multiclass_nms3",
inputs=inputs,
attrs={
'background_label': background_label,
'score_threshold': score_threshold,
'nms_top_k': nms_top_k,
'nms_threshold': nms_threshold,
'keep_top_k': keep_top_k,
'nms_eta': nms_eta,
'normalized': normalized
},
outputs=outputs)
output.stop_gradient = True
index.stop_gradient = True
if not return_index:
index = None
if not return_rois_num:
nms_rois_num = None
return output, nms_rois_num, index
class NMS(object):
def __init__(self, score_threshold, nms_top_k, nms_threshold):
self.score_threshold = score_threshold
self.nms_top_k = nms_top_k
self.nms_threshold = nms_threshold
def __call__(self, bboxes, scores):
attrs = {
'background_label': -1,
'score_threshold': self.score_threshold,
'nms_top_k': self.nms_top_k,
'nms_threshold': self.nms_threshold,
'keep_top_k': -1,
'nms_eta': 1.0,
'normalized': False,
'return_index': True
}
output, nms_rois_num, index = multiclass_nms(bboxes, scores, **attrs)
clas = paddle.slice(output, axes=[1], starts=[0], ends=[1])
clas = paddle.cast(clas, dtype="int64")
index = paddle.cast(index, dtype="int64")
if bboxes.shape[0] == 1:
batch = paddle.zeros_like(clas, dtype="int64")
else:
bboxes_count = bboxes.shape[1]
batch = paddle.divide(index, bboxes_count)
index = paddle.mod(index, bboxes_count)
res = paddle.concat([batch, clas, index], axis=1)
return res
x2paddle/op_mapper/dygraph/onnx2paddle/opset9/opset.py
浏览文件 @ 4a2df7e1
...@@ -42,20 +42,23 @@ def _const_weight_or_none(node, necessary=False): ...@@ -42,20 +42,23 @@ def _const_weight_or_none(node, necessary=False):
return None return None
def _rename_or_remove_weight(weights, origin_name, target_name=None, is_remove=True): def _rename_or_remove_weight(weights,
''' origin_name,
target_name=None,
is_remove=True):
'''
Rename parameters by Paddle's naming rule of parameters. Rename parameters by Paddle's naming rule of parameters.
Args: Args:
weights(dict[String:np.ndarray]): Dict stored paramters, the key in weights is name of parameter. weights(dict[String:np.ndarray]): Dict stored paramters, the key in weights is name of parameter.
origin_name(String): Name of parameter to rename or remove. origin_name(String): Name of parameter to rename or remove.
target_name(String, optional): if target_name is not None, add new key-value pair target_name(String, optional): if target_name is not None, add new key-value pair
{target_name:weights[origin_name]} to weights, and target_name must follow paddle's {target_name:weights[origin_name]} to weights, and target_name must follow paddle's
naming rule of parameters. Default: None. naming rule of parameters. Default: None.
is_remove: if is_remove is True, remove origin key-value pair. Default: True. is_remove: if is_remove is True, remove origin key-value pair. Default: True.
Returns: Returns:
None None
''' '''
if origin_name not in weights: if origin_name not in weights:
raise KeyError('{} not a key in {}'.format(origin_name, weights)) raise KeyError('{} not a key in {}'.format(origin_name, weights))
if is_remove: if is_remove:
...@@ -67,6 +70,7 @@ def _rename_or_remove_weight(weights, origin_name, target_name=None, is_remove=T ...@@ -67,6 +70,7 @@ def _rename_or_remove_weight(weights, origin_name, target_name=None, is_remove=T
# rename weight # rename weight
weights[target_name] = data weights[target_name] = data
def _is_static_shape(shape): def _is_static_shape(shape):
negtive_dims = 0 negtive_dims = 0
error_dims = 0 error_dims = 0
...@@ -96,9 +100,8 @@ def print_mapping_info(func): ...@@ -96,9 +100,8 @@ def print_mapping_info(func):
try: try:
res = func(*args, **kwargs) res = func(*args, **kwargs)
except: except:
print("convert failed node:{}, op_type is {}".format( raise Exception("convert failed node:{}, op_type is {}".format(
node.name[9:], node.layer_type)) node.name[9:], node.layer_type))
raise
else: else:
return res return res
...@@ -112,51 +115,61 @@ class OpSet9(): ...@@ -112,51 +115,61 @@ class OpSet9():
'Sub': 'paddle.subtract', 'Sub': 'paddle.subtract',
'Mul': 'paddle.multiply', 'Mul': 'paddle.multiply',
'Pow': 'paddle.pow', 'Pow': 'paddle.pow',
'Less': 'paddle.less_than',
} }
directly_map_ops = { directly_map_ops = {
'Ceil': ['paddle.ceil'], 'Ceil': ['paddle.ceil'],
# reduce function # reduce function
'ReduceMean': ['paddle.mean', 'ReduceMean': [
dict(axes='axis', keepdims='keepdim'), 'paddle.mean', dict(
dict(axes=None, keepdims=1)], axes='axis', keepdims='keepdim'), dict(
'ReduceSum': ['paddle.sum', axes=None, keepdims=1)
dict(axes='axis', keepdims='keepdim'), ],
dict(axes=None, keepdims=1)], 'ReduceSum': [
'ReduceMin': ['paddle.min', 'paddle.sum', dict(
dict(axes='axis', keepdims='keepdim'), axes='axis', keepdims='keepdim'), dict(
dict(axes=None, keepdim=1)], axes=None, keepdims=1)
'ReduceMax': ['paddle.max', ],
dict(axes='axis', keepdims='keepdim'), 'ReduceMin': [
dict(axes=None, keepdim=1)], 'paddle.min', dict(
'ReduceProd': ['paddle.prod', axes='axis', keepdims='keepdim'), dict(
dict(axes='axis', keepdims='keepdim'), axes=None, keepdim=1)
dict(axes=None, keepdim=1)], ],
'ReduceMax': [
'paddle.max', dict(
axes='axis', keepdims='keepdim'), dict(
axes=None, keepdim=1)
],
'ReduceProd': [
'paddle.prod', dict(
axes='axis', keepdims='keepdim'), dict(
axes=None, keepdim=1)
],
# active function # active function
'Relu': ['paddle.nn.ReLU'], 'Relu': ['paddle.nn.ReLU'],
'LeakyRelu': ['paddle.nn.LeakyReLU', 'LeakyRelu': [
dict(alpha='negative_slope'), 'paddle.nn.LeakyReLU', dict(alpha='negative_slope'),
dict(negative_slope=.01)], dict(negative_slope=.01)
'Elu': ['paddle.nn.functional.elu', ],
dict(alpha='alpha'), 'Elu':
dict(alpha=1.)], ['paddle.nn.functional.elu', dict(alpha='alpha'), dict(alpha=1.)],
'ThresholdedRelu': ['paddle.nn.functional.thresholded_relu', 'ThresholdedRelu': [
dict(alpha='threshold'), 'paddle.nn.functional.thresholded_relu', dict(alpha='threshold'),
dict(alpha=1.)], dict(alpha=1.)
],
'Tanh': ['paddle.nn.Tanh'], 'Tanh': ['paddle.nn.Tanh'],
'Sigmoid': ['paddle.nn.Sigmoid'], 'Sigmoid': ['paddle.nn.Sigmoid'],
'Softsign': ['paddle.nn.Softsign'], 'Softsign': ['paddle.nn.Softsign'],
'Softplus': ['paddle.nn.Softplus', 'Softplus': [
dict(threshold='threshold'), 'paddle.nn.Softplus', dict(threshold='threshold'),
dict(threshold=float(sys.maxsize))], dict(threshold=float(sys.maxsize))
],
'Exp': ['paddle.exp'], 'Exp': ['paddle.exp'],
'Log': ['paddle.log'], 'Log': ['paddle.log'],
'LogSoftmax': ['paddle.nn.functional.log_softmax', 'LogSoftmax':
dict(axis='axis'), ['paddle.nn.functional.log_softmax', dict(axis='axis'), dict(axis=1)],
dict(axis=1)], 'Softmax': ['paddle.nn.Softmax', dict(axis='axis'), dict(axis=1)],
'Softmax': ['paddle.nn.Softmax',
dict(axis='axis'),
dict(axis=1)],
'Sqrt': ['paddle.sqrt'], 'Sqrt': ['paddle.sqrt'],
'Floor': ['paddle.floor'], 'Floor': ['paddle.floor'],
'Abs': ['paddle.abs'], 'Abs': ['paddle.abs'],
...@@ -209,20 +222,16 @@ class OpSet9(): ...@@ -209,20 +222,16 @@ class OpSet9():
kernel=paddle_op, kernel=paddle_op,
inputs={"x": input.name}, inputs={"x": input.name},
outputs=[node.name], outputs=[node.name],
**layer_attrs) **layer_attrs)
@print_mapping_info @print_mapping_info
def elementwise_map(self, node): def elementwise_map(self, node):
op_type = self.elementwise_ops[node.layer_type] op_type = self.elementwise_ops[node.layer_type]
val_x = self.graph.get_input_node(node, idx=0, copy=True) val_x = self.graph.get_input_node(node, idx=0, copy=True)
val_y = self.graph.get_input_node(node, idx=1, copy=True) val_y = self.graph.get_input_node(node, idx=1, copy=True)
inputs_dict = {'x': val_x.name, inputs_dict = {'x': val_x.name, 'y': val_y.name}
'y': val_y.name}
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
op_type, op_type, inputs=inputs_dict, outputs=[node.name])
inputs=inputs_dict,
outputs=[node.name])
@print_mapping_info @print_mapping_info
def place_holder(self, node): def place_holder(self, node):
...@@ -248,8 +257,8 @@ class OpSet9(): ...@@ -248,8 +257,8 @@ class OpSet9():
shape = node.out_shapes[0] shape = node.out_shapes[0]
if hasattr(node.weight, "shape") and len(node.weight.shape) == 0: if hasattr(node.weight, "shape") and len(node.weight.shape) == 0:
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
"paddle.full", "paddle.full",
inputs={}, inputs={},
outputs=[node.name], outputs=[node.name],
dtype=string(dtype), dtype=string(dtype),
shape=[1], shape=[1],
...@@ -263,7 +272,7 @@ class OpSet9(): ...@@ -263,7 +272,7 @@ class OpSet9():
shape=shape, shape=shape,
attr=string(node.name), attr=string(node.name),
dtype=string(dtype), dtype=string(dtype),
default_initializer="paddle.nn.initializer.Constant(value=0.0)") default_initializer="paddle.nn.initializer.Constant(value=0.0)")
def _pad_if_asymmetric(self, node, pads, val_name): # pads: SSEE def _pad_if_asymmetric(self, node, pads, val_name): # pads: SSEE
assert len(pads) & 1 == 0 assert len(pads) & 1 == 0
...@@ -286,15 +295,17 @@ class OpSet9(): ...@@ -286,15 +295,17 @@ class OpSet9():
if len(node.layer.input) == 2: if len(node.layer.input) == 2:
# opset 10 # opset 10
val_scales = self.graph.get_input_node(node, idx=1, copy=True) val_scales = self.graph.get_input_node(node, idx=1, copy=True)
# TODO(syf): paddle.nn.functional.interpolate will support the length # TODO(syf): paddle.nn.functional.interpolate will support the length
# which is the same as the rank of input. # which is the same as the rank of input.
attrs['scale_factor'] = self.weights[val_scales.name].tolist()[2:] attrs['scale_factor'] = self.weights[val_scales.name].tolist()[
2:]
elif len(node.layer.input) == 3: elif len(node.layer.input) == 3:
# opset 11 # opset 11
val_scales = self.graph.get_input_node(node, idx=2, copy=True) val_scales = self.graph.get_input_node(node, idx=2, copy=True)
# TODO(syf): paddle.nn.functional.interpolate will support the length # TODO(syf): paddle.nn.functional.interpolate will support the length
# which is the same as the rank of input. # which is the same as the rank of input.
attrs['scale_factor'] = self.weights[val_scales.name].tolist()[2:] attrs['scale_factor'] = self.weights[val_scales.name].tolist()[
2:]
elif len(node.layer.input) == 4: elif len(node.layer.input) == 4:
# opset 11 # opset 11
val_sizes = self.graph.get_input_node(node, idx=3, copy=True) val_sizes = self.graph.get_input_node(node, idx=3, copy=True)
...@@ -311,8 +322,17 @@ class OpSet9(): ...@@ -311,8 +322,17 @@ class OpSet9():
outputs=[var_hw], outputs=[var_hw],
dtype=string('int32')) dtype=string('int32'))
inputs['size'] = var_hw inputs['size'] = var_hw
attrs = {"align_corners": False, attrs = {
"mode": string(node.get_attr('mode', 'nearest'))} "align_corners": False,
"mode": string(node.get_attr('mode', 'nearest'))
}
mode = node.get_attr('mode', 'nearest')
if mode == "linear":
attrs["mode"] = string("bilinear")
if node.get_attr('coordinate_transformation_mode',
'half_pixel') == 'pytorch_half_pixel':
attrs["align_corners"] = False
attrs["align_mode"] = 0
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
kernel="paddle.nn.functional.interpolate", kernel="paddle.nn.functional.interpolate",
inputs=inputs, inputs=inputs,
...@@ -329,21 +349,28 @@ class OpSet9(): ...@@ -329,21 +349,28 @@ class OpSet9():
starts=[2], starts=[2],
ends=[4]) ends=[4])
inputs['scale_factor'] = val_scales.name inputs['scale_factor'] = val_scales.name
mode = node.get_attr('mode', 'nearest') mode = node.get_attr('mode', 'nearest')
attrs.update({"align_corners": False, attrs.update({
"mode": string(mode), "align_corners": False,
"align_mode": 1}) "mode": string(mode),
"align_mode": 1
})
val_x_shape = val_x.out_shapes[0] val_x_shape = val_x.out_shapes[0]
if mode == "linear" and len(val_x_shape) == 4: if mode == "linear" and len(val_x_shape) == 4:
attrs["mode"] = string("bilinear") attrs["mode"] = string("bilinear")
attrs["align_corners"] = True if node.get_attr('coordinate_transformation_mode',
'half_pixel') == 'pytorch_half_pixel':
attrs["align_corners"] = False
attrs["align_mode"] = 0
else:
attrs["align_corners"] = True
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
kernel="paddle.nn.functional.interpolate", kernel="paddle.nn.functional.interpolate",
inputs=inputs, inputs=inputs,
outputs=[node.name], outputs=[node.name],
**attrs) **attrs)
@print_mapping_info @print_mapping_info
def HardSigmoid(self, node): def HardSigmoid(self, node):
val_x = self.graph.get_input_node(node, idx=0, copy=True) val_x = self.graph.get_input_node(node, idx=0, copy=True)
...@@ -360,8 +387,8 @@ class OpSet9(): ...@@ -360,8 +387,8 @@ class OpSet9():
inputs={"x": node.name + "_val"}, inputs={"x": node.name + "_val"},
outputs=[node.name], outputs=[node.name],
min=0.0, min=0.0,
max=1.0) max=1.0)
@print_mapping_info @print_mapping_info
def Shape(self, node): def Shape(self, node):
val_x = self.graph.get_input_node(node, idx=0, copy=True) val_x = self.graph.get_input_node(node, idx=0, copy=True)
...@@ -370,10 +397,10 @@ class OpSet9(): ...@@ -370,10 +397,10 @@ class OpSet9():
inputs={"input": val_x.name}, inputs={"input": val_x.name},
outputs=[node.name]) outputs=[node.name])
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
'paddle.cast', 'paddle.cast',
inputs={"x": node.name}, inputs={"x": node.name},
outputs=[node.name], outputs=[node.name],
dtype=string('int64')) dtype=string('int64'))
@print_mapping_info @print_mapping_info
def RoiAlign(self, node): def RoiAlign(self, node):
...@@ -384,11 +411,25 @@ class OpSet9(): ...@@ -384,11 +411,25 @@ class OpSet9():
pooled_width = node.get_attr('output_width') pooled_width = node.get_attr('output_width')
spatial_scale = node.get_attr('spatial_scale') spatial_scale = node.get_attr('spatial_scale')
sampling_ratio = node.get_attr('sampling_ratio') sampling_ratio = node.get_attr('sampling_ratio')
#dygraph rois_num is necessary
val_rois_shape = val_rois.name + '_shape'
self.paddle_graph.add_layer(
kernel="paddle.shape",
inputs={"input": val_rois.name},
outputs=[val_rois_shape])
val_rois_num = val_rois.name + '_num'
self.paddle_graph.add_layer(
'paddle.split',
inputs={"x": val_rois_shape},
outputs=[val_rois_num, '_', '_', '_'],
num_or_sections=[1, 1, 1, 1],
axis=0)
layer_attrs = { layer_attrs = {
'pooled_height': pooled_height, 'pooled_height': pooled_height,
'pooled_width': pooled_width, 'pooled_width': pooled_width,
'spatial_scale': spatial_scale, 'spatial_scale': spatial_scale,
'sampling_ratio': sampling_ratio, 'sampling_ratio': sampling_ratio,
'rois_num': val_rois_num,
} }
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
'paddle.fluid.layers.roi_align', 'paddle.fluid.layers.roi_align',
...@@ -396,7 +437,6 @@ class OpSet9(): ...@@ -396,7 +437,6 @@ class OpSet9():
'rois': val_rois.name}, 'rois': val_rois.name},
outputs=[node.name], outputs=[node.name],
**layer_attrs) **layer_attrs)
@print_mapping_info @print_mapping_info
def MaxRoiPool(self, node): def MaxRoiPool(self, node):
...@@ -446,12 +486,14 @@ class OpSet9(): ...@@ -446,12 +486,14 @@ class OpSet9():
if is_pads_attr: if is_pads_attr:
paddings = [] paddings = []
if len(pads) == 10 and sum(pads) == 0: if len(pads) == 10 and sum(pads) == 0:
pads = pads[0: 6] pads = pads[0:6]
if len(pads) in [2, 4, 6]: if len(pads) in [2, 4, 6]:
if data_shape: if data_shape:
assume_pad |= data_shape and 2 * (len(data_shape) - 2) == len(pads) # NCHW assume_pad |= data_shape and 2 * (len(data_shape) - 2
) == len(pads) # NCHW
if output_shape: if output_shape:
assume_pad |= output_shape and 2 * (len(output_shape) - 2) == len(pads) # NCHW assume_pad |= output_shape and 2 * (len(output_shape) - 2
) == len(pads) # NCHW
if assume_pad: if assume_pad:
paddle_op = 'paddle.nn.Pad{}D'.format(len(output_shape) - 2) paddle_op = 'paddle.nn.Pad{}D'.format(len(output_shape) - 2)
paddings = np.array(pads).reshape( paddings = np.array(pads).reshape(
...@@ -460,21 +502,27 @@ class OpSet9(): ...@@ -460,21 +502,27 @@ class OpSet9():
layer_attrs['padding'] = paddings layer_attrs['padding'] = paddings
else: else:
if data_shape: if data_shape:
assume_pad |= data_shape and 2 * len(data_shape) == len(pads) # NCHW assume_pad |= data_shape and 2 * len(data_shape) == len(
pads) # NCHW
if output_shape: if output_shape:
assume_pad |= output_shape and 2 * len(output_shape) == len(pads) # NCHW assume_pad |= output_shape and 2 * len(
output_shape) == len(pads) # NCHW
if assume_pad: if assume_pad:
paddle_op = 'paddle.nn.functional.pad' paddle_op = 'paddle.nn.functional.pad'
paddings = np.array(pads).reshape( paddings = np.array(pads).reshape(
(2, -1)).transpose().astype("int32").flatten().tolist() (2,
-1)).transpose().astype("int32").flatten().tolist()
layer_attrs['pad'] = paddings layer_attrs['pad'] = paddings
else: else:
raise Exception("The padding value {} is wrong!".format(pads)) raise Exception("The padding value {} is wrong!".format(
pads))
elif len(pads) == 8: elif len(pads) == 8:
if data_shape: if data_shape:
assume_pad |= data_shape and 2 * len(data_shape) == len(pads) # NCHW assume_pad |= data_shape and 2 * len(data_shape) == len(
pads) # NCHW
if output_shape: if output_shape:
assume_pad |= output_shape and 2 * len(output_shape) == len(pads) # NCHW assume_pad |= output_shape and 2 * len(output_shape) == len(
pads) # NCHW
if assume_pad: if assume_pad:
paddle_op = 'paddle.nn.Pad2D' paddle_op = 'paddle.nn.Pad2D'
paddings = np.array(pads).reshape( paddings = np.array(pads).reshape(
...@@ -487,11 +535,12 @@ class OpSet9(): ...@@ -487,11 +535,12 @@ class OpSet9():
layer_attrs["pad"] = paddings layer_attrs["pad"] = paddings
paddle_op = "custom_layer:PadAllDim4WithOneInput" paddle_op = "custom_layer:PadAllDim4WithOneInput"
else: else:
raise Exception("The padding value {} is wrong!".format(pads)) raise Exception("The padding value {} is wrong!".format(pads))
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
paddle_op, paddle_op,
inputs={'x': val_x.name}, inputs={'x': val_x.name},
outputs=layer_outputs[1:] if paddle_op == 'paddle.nn.functional.pad' else layer_outputs, outputs=layer_outputs[1:]
if paddle_op == 'paddle.nn.functional.pad' else layer_outputs,
**layer_attrs) **layer_attrs)
if not op_independent: if not op_independent:
return node.name + '_paded' return node.name + '_paded'
...@@ -499,9 +548,11 @@ class OpSet9(): ...@@ -499,9 +548,11 @@ class OpSet9():
pads_len = val_pad.out_shapes[0][0] pads_len = val_pad.out_shapes[0][0]
if pads_len in [2, 4, 6]: if pads_len in [2, 4, 6]:
if data_shape: if data_shape:
assume_pad |= data_shape and 2 * (len(data_shape) - 2) == pads_len # NCHW assume_pad |= data_shape and 2 * (len(data_shape) - 2
) == pads_len # NCHW
if output_shape: if output_shape:
assume_pad |= output_shape and 2 * (len(output_shape) - 2) == pads_len # NCHW assume_pad |= output_shape and 2 * (len(output_shape) - 2
) == pads_len # NCHW
if assume_pad: if assume_pad:
if pads_len == 2: if pads_len == 2:
data_format = "NCL" data_format = "NCL"
...@@ -510,41 +561,48 @@ class OpSet9(): ...@@ -510,41 +561,48 @@ class OpSet9():
else: else:
data_format = "NCDHW" data_format = "NCDHW"
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
"custom_layer:PadWithTwoInput", "custom_layer:PadWithTwoInput",
inputs={'x': val_x.name, 'pad': val_pad.name}, inputs={'x': val_x.name,
'pad': val_pad.name},
outputs=layer_outputs, outputs=layer_outputs,
value=value, value=value,
mode=string(mode), mode=string(mode),
data_format=string(data_format)) data_format=string(data_format))
else: else:
if data_shape: if data_shape:
assume_pad |= data_shape and 2 * len(data_shape) == pads_len # NCHW assume_pad |= data_shape and 2 * len(
data_shape) == pads_len # NCHW
if output_shape: if output_shape:
assume_pad |= output_shape and 2 * len(output_shape) == pads_len # NCHW assume_pad |= output_shape and 2 * len(
output_shape) == pads_len # NCHW
if assume_pad: if assume_pad:
if pads_len == 4: if pads_len == 4:
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
"custom_layer:PadAllDim2", "custom_layer:PadAllDim2",
inputs={'x': val_x.name, 'pad': val_pad.name}, inputs={'x': val_x.name,
outputs=layer_outputs, 'pad': val_pad.name},
outputs=layer_outputs,
value=value, value=value,
mode=string(mode)) mode=string(mode))
else: else:
raise Exception("The padding value is wrong!") raise Exception("The padding value is wrong!")
elif pads_len == 8: elif pads_len == 8:
if data_shape: if data_shape:
assume_pad |= data_shape and 2 * len(data_shape) == pads_len # NCHW assume_pad |= data_shape and 2 * len(
data_shape) == pads_len # NCHW
if output_shape: if output_shape:
assume_pad |= output_shape and 2 * len(output_shape) == pads_len # NCHW assume_pad |= output_shape and 2 * len(
output_shape) == pads_len # NCHW
if assume_pad: if assume_pad:
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
"custom_layer:PadAllDim4", "custom_layer:PadAllDim4",
inputs={'x': val_x.name, 'pad': val_pad.name}, inputs={'x': val_x.name,
outputs=layer_outputs, 'pad': val_pad.name},
outputs=layer_outputs,
value=value, value=value,
mode=string(mode)) mode=string(mode))
else: else:
raise Exception("The padding value is wrong!") raise Exception("The padding value is wrong!")
if not op_independent: if not op_independent:
return node.name + '_paded' return node.name + '_paded'
...@@ -562,8 +620,8 @@ class OpSet9(): ...@@ -562,8 +620,8 @@ class OpSet9():
shape=[1]) shape=[1])
else: else:
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
'paddle.unsqueeze', 'paddle.unsqueeze',
inputs={"x": val_x.name}, inputs={"x": val_x.name},
outputs=[node.name], outputs=[node.name],
**layer_attrs) **layer_attrs)
...@@ -574,9 +632,9 @@ class OpSet9(): ...@@ -574,9 +632,9 @@ class OpSet9():
lambd = node.get_attr('lambd') lambd = node.get_attr('lambd')
assert bias == 0.0, 'not support bias!=0' assert bias == 0.0, 'not support bias!=0'
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
'paddle.nn.functional.hardshrink', 'paddle.nn.functional.hardshrink',
inputs={"x": val_x.name}, inputs={"x": val_x.name},
outputs=[node.name], outputs=[node.name],
threshold=lambd) threshold=lambd)
@print_mapping_info @print_mapping_info
...@@ -603,8 +661,8 @@ class OpSet9(): ...@@ -603,8 +661,8 @@ class OpSet9():
value = value.tolist() value = value.tolist()
value = value[0] value = value[0]
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
"paddle.full", "paddle.full",
inputs={}, inputs={},
outputs=[node.name], outputs=[node.name],
dtype=string(dtype), dtype=string(dtype),
shape=[1], shape=[1],
...@@ -638,8 +696,8 @@ class OpSet9(): ...@@ -638,8 +696,8 @@ class OpSet9():
val_scale = self.graph.get_input_node(node, idx=1, copy=True) val_scale = self.graph.get_input_node(node, idx=1, copy=True)
val_b = self.graph.get_input_node(node, idx=2, copy=True) val_b = self.graph.get_input_node(node, idx=2, copy=True)
epsilon = node.get_attr('epsilon', 1e-5) epsilon = node.get_attr('epsilon', 1e-5)
self.weights[op_name+'.scale'] = self.weights[val_scale.name] self.weights[op_name + '.scale'] = self.weights[val_scale.name]
self.weights[op_name+'.bias'] = self.weights[val_b.name] self.weights[op_name + '.bias'] = self.weights[val_b.name]
layer_attrs = { layer_attrs = {
'num_features': node.out_shapes[0][1], 'num_features': node.out_shapes[0][1],
'epsilon': epsilon, 'epsilon': epsilon,
...@@ -652,11 +710,13 @@ class OpSet9(): ...@@ -652,11 +710,13 @@ class OpSet9():
elif dim == 5: elif dim == 5:
paddle_op = "paddle.nn.InstanceNorm3D" paddle_op = "paddle.nn.InstanceNorm3D"
else: else:
raise Exception("The paddle only support 2D, 3D, 4D or 5D input in InstanceNormalization.") raise Exception(
"The paddle only support 2D, 3D, 4D or 5D input in InstanceNormalization."
)
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
paddle_op, paddle_op,
inputs={"x": val_x.name}, inputs={"x": val_x.name},
outputs=layer_outputs, outputs=layer_outputs,
**layer_attrs) **layer_attrs)
@print_mapping_info @print_mapping_info
...@@ -671,16 +731,10 @@ class OpSet9(): ...@@ -671,16 +731,10 @@ class OpSet9():
'fill_value': 1 'fill_value': 1
} }
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
'paddle.full', 'paddle.full', inputs={}, outputs=[name_ones], **attr_ones)
inputs={}, inputs_dict = {'x': name_ones, 'y': val_x.name}
outputs=[name_ones],
**attr_ones)
inputs_dict = {'x': name_ones,
'y': val_x.name}
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
'paddle.multiply', 'paddle.multiply', inputs=inputs_dict, outputs=[node.name])
inputs=inputs_dict,
outputs=[node.name])
@print_mapping_info @print_mapping_info
def Gather(self, node): def Gather(self, node):
...@@ -734,9 +788,9 @@ class OpSet9(): ...@@ -734,9 +788,9 @@ class OpSet9():
for i in range(len(perm)): for i in range(len(perm)):
new_perm[perm[i]] = i new_perm[perm[i]] = i
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
'paddle.transpose', 'paddle.transpose',
inputs={"x": node.name}, inputs={"x": node.name},
outputs=[node.name], outputs=[node.name],
perm=new_perm) perm=new_perm)
if len(indices_shape) < 1: if len(indices_shape) < 1:
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
...@@ -843,9 +897,11 @@ class OpSet9(): ...@@ -843,9 +897,11 @@ class OpSet9():
if len(indices.out_shapes[0]) == 1: if len(indices.out_shapes[0]) == 1:
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
'paddle.scatter', 'paddle.scatter',
inputs={'x': val_x.name, inputs={
'index': indices.name, 'x': val_x.name,
'updates': updates.name}, 'index': indices.name,
'updates': updates.name
},
outputs=[node.name]) outputs=[node.name])
else: else:
input_inner_indices = node.name + '_input_inner_indices' input_inner_indices = node.name + '_input_inner_indices'
...@@ -920,9 +976,11 @@ class OpSet9(): ...@@ -920,9 +976,11 @@ class OpSet9():
val_limit = self.graph.get_input_node(node, idx=1, copy=True) val_limit = self.graph.get_input_node(node, idx=1, copy=True)
val_delta = self.graph.get_input_node(node, idx=2, copy=True) val_delta = self.graph.get_input_node(node, idx=2, copy=True)
dtype = val_start.dtype dtype = val_start.dtype
inputs = {'start': val_start.name, inputs = {
'end': val_limit.name, 'start': val_start.name,
'step': val_delta.name} 'end': val_limit.name,
'step': val_delta.name
}
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
'paddle.arange', 'paddle.arange',
inputs=inputs, inputs=inputs,
...@@ -952,7 +1010,7 @@ class OpSet9(): ...@@ -952,7 +1010,7 @@ class OpSet9():
if len(node.inputs) > 4: if len(node.inputs) > 4:
steps = self.graph.get_input_node(node, idx=4, copy=True) steps = self.graph.get_input_node(node, idx=4, copy=True)
steps = _const_weight_or_none(steps).tolist() steps = _const_weight_or_none(steps).tolist()
layer_attrs = { layer_attrs = {
"axes": axes, "axes": axes,
"starts": starts.name, "starts": starts.name,
...@@ -962,12 +1020,13 @@ class OpSet9(): ...@@ -962,12 +1020,13 @@ class OpSet9():
starts_value = starts_value.copy() starts_value = starts_value.copy()
ends_value = ends_value.copy() ends_value = ends_value.copy()
for idx in range(len(ends_value)): for idx in range(len(ends_value)):
if starts_value[idx] >= val_x.out_shapes[0][axes[idx]] and val_x.out_shapes[0][axes[idx]] > 0: if starts_value[idx] >= val_x.out_shapes[0][axes[
idx]] and val_x.out_shapes[0][axes[idx]] > 0:
starts_value[idx] = val_x.out_shapes[0][axes[idx]] - 1 starts_value[idx] = val_x.out_shapes[0][axes[idx]] - 1
ends_value[idx] = val_x.out_shapes[0][axes[idx]] ends_value[idx] = val_x.out_shapes[0][axes[idx]]
elif ends_value[idx] > 2**31 - 1: elif ends_value[idx] > 2**31 - 1:
ends_value[idx] = 2**31 - 1 ends_value[idx] = 2**31 - 1
layer_attrs = { layer_attrs = {
"axes": axes, "axes": axes,
"starts": starts_value, "starts": starts_value,
...@@ -1001,19 +1060,18 @@ class OpSet9(): ...@@ -1001,19 +1060,18 @@ class OpSet9():
ends[idx] = 2**31 - 1 ends[idx] = 2**31 - 1
layer_attrs = {"axes": axes, "starts": starts, "ends": ends} layer_attrs = {"axes": axes, "starts": starts, "ends": ends}
if steps is not None: if steps is not None:
layer_attrs['strides'] = steps layer_attrs['strides'] = steps
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
'paddle.strided_slice', 'paddle.strided_slice',
inputs={"x": val_x.name}, inputs={"x": val_x.name},
outputs=[node.name], outputs=[node.name],
**layer_attrs) **layer_attrs)
else: else:
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
'paddle.slice', 'paddle.slice',
inputs={"input": val_x.name}, inputs={"input": val_x.name},
outputs=[node.name], outputs=[node.name],
**layer_attrs) **layer_attrs)
@print_mapping_info @print_mapping_info
...@@ -1028,13 +1086,10 @@ class OpSet9(): ...@@ -1028,13 +1086,10 @@ class OpSet9():
'this is not supported') 'this is not supported')
if len(value) == 1: if len(value) == 1:
value = value[0] value = value[0]
layer_attrs = { layer_attrs = {'dtype': string(dtype), 'fill_value': value}
'dtype': string(dtype),
'fill_value': value
}
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
"paddle.full", "paddle.full",
inputs={'shape': val_shape.name}, inputs={'shape': val_shape.name},
outputs=[node.name], outputs=[node.name],
**layer_attrs) **layer_attrs)
...@@ -1050,11 +1105,11 @@ class OpSet9(): ...@@ -1050,11 +1105,11 @@ class OpSet9():
'max': max_value, 'max': max_value,
'min': min_value, 'min': min_value,
} }
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
'paddle.clip', 'paddle.clip',
inputs={"x": val_x.name}, inputs={"x": val_x.name},
outputs=[node.name], outputs=[node.name],
**layer_attrs) **layer_attrs)
else: else:
min_ipt = self.graph.get_input_node(node, idx=1, copy=True) min_ipt = self.graph.get_input_node(node, idx=1, copy=True)
...@@ -1068,9 +1123,9 @@ class OpSet9(): ...@@ -1068,9 +1123,9 @@ class OpSet9():
if max_value is not None and min_value is not None: if max_value is not None and min_value is not None:
layer_attrs = {'max': max_value, 'min': min_value} layer_attrs = {'max': max_value, 'min': min_value}
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
'paddle.clip', 'paddle.clip',
inputs={"x": val_x.name}, inputs={"x": val_x.name},
outputs=[node.name], outputs=[node.name],
**layer_attrs) **layer_attrs)
else: else:
raise raise
...@@ -1095,9 +1150,9 @@ class OpSet9(): ...@@ -1095,9 +1150,9 @@ class OpSet9():
else: else:
outputs_list.append(node.name) outputs_list.append(node.name)
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
'paddle.split', 'paddle.split',
inputs={"x": val_x.name}, inputs={"x": val_x.name},
outputs=outputs_list, outputs=outputs_list,
**layer_attrs) **layer_attrs)
@print_mapping_info @print_mapping_info
...@@ -1154,17 +1209,18 @@ class OpSet9(): ...@@ -1154,17 +1209,18 @@ class OpSet9():
if output_dtype: if output_dtype:
assert dtype == output_dtype, 'dtype of to unmatches output' assert dtype == output_dtype, 'dtype of to unmatches output'
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
'paddle.cast', 'paddle.cast',
inputs={'x': val_input.name}, inputs={'x': val_input.name},
outputs=[node.name], outputs=[node.name],
dtype=string(dtype)) dtype=string(dtype))
@print_mapping_info @print_mapping_info
def Not(self, node): def Not(self, node):
val_input = self.graph.get_input_node(node, idx=0, copy=True) val_input = self.graph.get_input_node(node, idx=0, copy=True)
self.paddle_graph.add_layer('paddle.logical_not', self.paddle_graph.add_layer(
inputs={'x': val_input.name}, 'paddle.logical_not',
outputs=[node.name]) inputs={'x': val_input.name},
outputs=[node.name])
@print_mapping_info @print_mapping_info
def AveragePool(self, node): def AveragePool(self, node):
...@@ -1201,9 +1257,9 @@ class OpSet9(): ...@@ -1201,9 +1257,9 @@ class OpSet9():
"exclusive": 'True', "exclusive": 'True',
} }
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
paddle_op, paddle_op,
inputs={'x': val_x if isinstance(val_x, str) else val_x.name}, inputs={'x': val_x if isinstance(val_x, str) else val_x.name},
outputs=layer_outputs, outputs=layer_outputs,
**layer_attrs) **layer_attrs)
@print_mapping_info @print_mapping_info
...@@ -1218,9 +1274,9 @@ class OpSet9(): ...@@ -1218,9 +1274,9 @@ class OpSet9():
assert 'Unspported situation happened, please create issue on https://github.com/PaddlePaddle/X2Paddle/issues.' assert 'Unspported situation happened, please create issue on https://github.com/PaddlePaddle/X2Paddle/issues.'
axis = node.get_attr('axis') axis = node.get_attr('axis')
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
'paddle.concat', 'paddle.concat',
inputs={"x": inputs_list}, inputs={"x": inputs_list},
outputs=[node.name], outputs=[node.name],
axis=axis) axis=axis)
@print_mapping_info @print_mapping_info
...@@ -1238,8 +1294,8 @@ class OpSet9(): ...@@ -1238,8 +1294,8 @@ class OpSet9():
for s in output_shape[axis:]: for s in output_shape[axis:]:
shape_list[1] *= s shape_list[1] *= s
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
'paddle.reshape', 'paddle.reshape',
inputs={"x": val_x.name}, inputs={"x": val_x.name},
outputs=[node.name], outputs=[node.name],
shape=shape_list) shape=shape_list)
...@@ -1254,8 +1310,7 @@ class OpSet9(): ...@@ -1254,8 +1310,7 @@ class OpSet9():
trans_a = bool(node.get_attr('transA', 0)) # optional trans_a = bool(node.get_attr('transA', 0)) # optional
trans_b = bool(node.get_attr('transB', 0)) # optional trans_b = bool(node.get_attr('transB', 0)) # optional
val_mm = node.name + '_mm' val_mm = node.name + '_mm'
matmul_inputs = {"x": val_a.name, matmul_inputs = {"x": val_a.name, "y": val_b.name}
"y": val_b.name}
attr_matmul = { attr_matmul = {
"transpose_x": trans_a, "transpose_x": trans_a,
"transpose_y": trans_b, "transpose_y": trans_b,
...@@ -1266,19 +1321,13 @@ class OpSet9(): ...@@ -1266,19 +1321,13 @@ class OpSet9():
outputs=[val_mm], outputs=[val_mm],
**attr_matmul) **attr_matmul)
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
"paddle.scale", "paddle.scale", inputs={"x": val_mm}, outputs=[val_mm], scale=alpha)
inputs={"x": val_mm},
outputs=[val_mm],
scale=alpha)
if beta != 0: if beta != 0:
if beta == 1.: if beta == 1.:
add_inputs = {"x": val_mm, add_inputs = {"x": val_mm, "y": val_c.name}
"y": val_c.name}
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
"paddle.add", "paddle.add", inputs=add_inputs, outputs=[node.name])
inputs=add_inputs,
outputs=[node.name])
else: else:
var_beta = node.name + '_beta' var_beta = node.name + '_beta'
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
...@@ -1288,9 +1337,7 @@ class OpSet9(): ...@@ -1288,9 +1337,7 @@ class OpSet9():
scale=beta) scale=beta)
add_inputs = {"x": val_mm, "y": var_beta} add_inputs = {"x": val_mm, "y": var_beta}
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
"paddle.add", "paddle.add", inputs=add_inputs, outputs=[node.name])
inputs=add_inputs,
outputs=[node.name])
@print_mapping_info @print_mapping_info
def Sum(self, node): def Sum(self, node):
...@@ -1301,9 +1348,8 @@ class OpSet9(): ...@@ -1301,9 +1348,8 @@ class OpSet9():
"y": self.graph.get_input_node( "y": self.graph.get_input_node(
node, idx=1, copy=True).name, node, idx=1, copy=True).name,
} }
self.paddle_graph.add_layer("paddle.add", self.paddle_graph.add_layer(
inputs=inputs_dict, "paddle.add", inputs=inputs_dict, outputs=[node.name])
outputs=[node.name])
for idx, ipt in enumerate(val_inps[2:]): for idx, ipt in enumerate(val_inps[2:]):
y = self.graph.get_input_node(node, idx=idx, copy=True) y = self.graph.get_input_node(node, idx=idx, copy=True)
...@@ -1312,9 +1358,7 @@ class OpSet9(): ...@@ -1312,9 +1358,7 @@ class OpSet9():
"y": y.name, "y": y.name,
} }
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
"paddle.add", "paddle.add", inputs=inputs_dict, outputs=[node.name])
inputs=inputs_dict,
outputs=[node.name])
@print_mapping_info @print_mapping_info
def MatMul(self, node): def MatMul(self, node):
...@@ -1322,8 +1366,7 @@ class OpSet9(): ...@@ -1322,8 +1366,7 @@ class OpSet9():
val_y = self.graph.get_input_node(node, idx=1, copy=True) val_y = self.graph.get_input_node(node, idx=1, copy=True)
x_shape = val_x.out_shapes[0] x_shape = val_x.out_shapes[0]
y_shape = val_y.out_shapes[0] y_shape = val_y.out_shapes[0]
inputs_dict = {"x": val_x.name, inputs_dict = {"x": val_x.name, "y": val_y.name}
"y": val_y.name}
if y_shape[0] == 1 and x_shape[-1] != 1 and x_shape[0] != 1: if y_shape[0] == 1 and x_shape[-1] != 1 and x_shape[0] != 1:
y_squeeze = val_y.name + '_squeeze' y_squeeze = val_y.name + '_squeeze'
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
...@@ -1333,14 +1376,10 @@ class OpSet9(): ...@@ -1333,14 +1376,10 @@ class OpSet9():
axis=[0]) axis=[0])
inputs_dict['y'] = y_squeeze inputs_dict['y'] = y_squeeze
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
"paddle.matmul", "paddle.matmul", inputs=inputs_dict, outputs=[node.name])
inputs=inputs_dict,
outputs=[node.name])
else: else:
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
"paddle.matmul", "paddle.matmul", inputs=inputs_dict, outputs=[node.name])
inputs=inputs_dict,
outputs=[node.name])
@print_mapping_info @print_mapping_info
def BatchNormalization(self, node): def BatchNormalization(self, node):
...@@ -1357,10 +1396,13 @@ class OpSet9(): ...@@ -1357,10 +1396,13 @@ class OpSet9():
epsilon = node.get_attr('epsilon', 1e-5) epsilon = node.get_attr('epsilon', 1e-5)
c = val_x.out_shapes[0][1] c = val_x.out_shapes[0][1]
_rename_or_remove_weight(self.weights, val_scale.name, op_name+'.weight') _rename_or_remove_weight(self.weights, val_scale.name,
_rename_or_remove_weight(self.weights, val_b.name, op_name+'.bias') op_name + '.weight')
_rename_or_remove_weight(self.weights, val_var.name, op_name+'._variance') _rename_or_remove_weight(self.weights, val_b.name, op_name + '.bias')
_rename_or_remove_weight(self.weights, val_mean.name, op_name+'._mean') _rename_or_remove_weight(self.weights, val_var.name,
op_name + '._variance')
_rename_or_remove_weight(self.weights, val_mean.name,
op_name + '._mean')
# Attribute: spatial is used in BatchNormalization-1,6,7 # Attribute: spatial is used in BatchNormalization-1,6,7
spatial = bool(node.get_attr('spatial')) spatial = bool(node.get_attr('spatial'))
...@@ -1372,9 +1414,9 @@ class OpSet9(): ...@@ -1372,9 +1414,9 @@ class OpSet9():
"use_global_stats": False, "use_global_stats": False,
} }
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
"paddle.nn.BatchNorm", "paddle.nn.BatchNorm",
inputs={"x": val_x.name}, inputs={"x": val_x.name},
outputs=layer_outputs, outputs=layer_outputs,
**layer_attrs) **layer_attrs)
@print_mapping_info @print_mapping_info
...@@ -1385,9 +1427,9 @@ class OpSet9(): ...@@ -1385,9 +1427,9 @@ class OpSet9():
perm_default.reverse() perm_default.reverse()
perm = node.get_attr('perm', perm_default) perm = node.get_attr('perm', perm_default)
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
"paddle.transpose", "paddle.transpose",
inputs={"x": val_x.name}, inputs={"x": val_x.name},
outputs=[node.name], outputs=[node.name],
perm=perm) perm=perm)
@print_mapping_info @print_mapping_info
...@@ -1406,57 +1448,61 @@ class OpSet9(): ...@@ -1406,57 +1448,61 @@ class OpSet9():
if mode == "element": if mode == "element":
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
"paddle.zeros", "paddle.zeros",
inputs={}, inputs={},
outputs=[output_name + "__zeros"], outputs=[output_name + "__zeros"],
shape=shape_slope, shape=shape_slope,
dtype=string(node.dtype)) dtype=string(node.dtype))
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
"paddle.maximum", "paddle.maximum",
inputs={"x": val_x.name, inputs={"x": val_x.name,
"y": output_name + "__zeros"}, "y": output_name + "__zeros"},
outputs=[output_name + "__max"]) outputs=[output_name + "__max"])
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
"paddle.minimum", "paddle.minimum",
inputs={"x": val_x.name, inputs={"x": val_x.name,
"y": output_name + "__zeros"}, "y": output_name + "__zeros"},
outputs=[output_name + "__max"]) outputs=[output_name + "__max"])
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
"paddle.multiply", "paddle.multiply",
inputs={"x": val_slope.name, inputs={"x": val_slope.name,
"y": output_name + "__min"}, "y": output_name + "__min"},
outputs=[output_name + "__mul"]) outputs=[output_name + "__mul"])
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
"paddle.add", "paddle.add",
inputs={"x": output_name + "__max", inputs={
"y": output_name + "__mul"}, "x": output_name + "__max",
"y": output_name + "__mul"
},
outputs=[output_name]) outputs=[output_name])
else: else:
if mode == 'channel': if mode == 'channel':
slope_data = _const_weight_or_none(val_slope) slope_data = _const_weight_or_none(val_slope)
if slope_data is None: if slope_data is None:
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
"paddle.reshape", "paddle.reshape",
inputs={"x": val_slope.name}, inputs={"x": val_slope.name},
outputs=[val_slope.name], outputs=[val_slope.name],
shape=[shape_slope[0]]) shape=[shape_slope[0]])
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
"paddle.nn.functional.prelu", "paddle.nn.functional.prelu",
inputs={"x": val_x.name, inputs={"x": val_x.name,
"weight": val_slope.name}, "weight": val_slope.name},
outputs=[node.name]) outputs=[node.name])
return return
_rename_or_remove_weight(self.weights, val_slope.name) _rename_or_remove_weight(self.weights, val_slope.name)
if len(shape_slope) > 1: if len(shape_slope) > 1:
self.weights[op_name+'._weight'] = np.reshape(slope_data, shape_slope[0]) self.weights[op_name + '._weight'] = np.reshape(
slope_data, shape_slope[0])
num_parameters = val_x.out_shapes[0][1] num_parameters = val_x.out_shapes[0][1]
else: else:
num_parameters = 1 num_parameters = 1
_rename_or_remove_weight(self.weights, val_slope.name) _rename_or_remove_weight(self.weights, val_slope.name)
self.weights[op_name+'._weight'] = np.reshape(self.weights[val_slope.name], [1]) self.weights[op_name + '._weight'] = np.reshape(
self.weights[val_slope.name], [1])
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
"paddle.nn.PReLU", "paddle.nn.PReLU",
inputs={"x": val_x.name}, inputs={"x": val_x.name},
outputs=layer_outputs, outputs=layer_outputs,
num_parameters=num_parameters) num_parameters=num_parameters)
@print_mapping_info @print_mapping_info
...@@ -1471,9 +1517,9 @@ class OpSet9(): ...@@ -1471,9 +1517,9 @@ class OpSet9():
dtype=string(val_x.dtype)) dtype=string(val_x.dtype))
else: else:
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
"paddle.squeeze", "paddle.squeeze",
inputs={"x": val_x.name}, inputs={"x": val_x.name},
outputs=[node.name], outputs=[node.name],
axis=axes) axis=axes)
@print_mapping_info @print_mapping_info
...@@ -1494,8 +1540,7 @@ class OpSet9(): ...@@ -1494,8 +1540,7 @@ class OpSet9():
"paddle.greater_than", "paddle.greater_than",
inputs={'x': val_x.name, inputs={'x': val_x.name,
'y': val_y.name}, 'y': val_y.name},
outputs=[node.name], outputs=[node.name])
param_attr=None)
@print_mapping_info @print_mapping_info
def Where(self, node): def Where(self, node):
...@@ -1545,8 +1590,8 @@ class OpSet9(): ...@@ -1545,8 +1590,8 @@ class OpSet9():
val_x_dim = len(val_x.out_shapes[0]) val_x_dim = len(val_x.out_shapes[0])
if val_x_dim == 1: if val_x_dim == 1:
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
"paddle.nonzero", "paddle.nonzero",
inputs={"x": val_x.name}, inputs={"x": val_x.name},
outputs=[val_x.name]) outputs=[val_x.name])
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
"paddle.transpose", "paddle.transpose",
...@@ -1555,27 +1600,23 @@ class OpSet9(): ...@@ -1555,27 +1600,23 @@ class OpSet9():
perm=[1, 0]) perm=[1, 0])
if val_x_dim > 1: if val_x_dim > 1:
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
"paddle.nonzero", "paddle.nonzero",
inputs={"x": val_x.name}, inputs={"x": val_x.name},
outputs=[val_x.name]) outputs=[val_x.name])
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
"paddle.split", "paddle.split",
inputs={"x": val_x.name}, inputs={"x": val_x.name},
outputs=[val_x.name], outputs=[val_x.name],
num_or_sections=1, num_or_sections=1,
axis=val_x_dim) axis=val_x_dim)
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
"paddle.concat", "paddle.concat", inputs={"x": val_x.name}, outputs=[node.name])
inputs={"x": val_x.name},
outputs=[node.name])
@print_mapping_info @print_mapping_info
def Identity(self, node): def Identity(self, node):
val_x = self.graph.get_input_node(node, idx=0, copy=True) val_x = self.graph.get_input_node(node, idx=0, copy=True)
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
"paddle.assign", "paddle.assign", inputs={"x": val_x.name}, outputs=[node.name])
inputs={"x": val_x.name},
outputs=[node.name])
@print_mapping_info @print_mapping_info
def Tile(self, node): def Tile(self, node):
...@@ -1589,22 +1630,25 @@ class OpSet9(): ...@@ -1589,22 +1630,25 @@ class OpSet9():
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
"paddle.cast", "paddle.cast",
inputs={"x": repeats}, inputs={"x": repeats},
outputs=["{}.tmp".format(repeats)], outputs=["{}_tmp".format(repeats)],
dtype=string("int32")) dtype=string("int32"))
repeats = "{}.tmp".format(repeats) repeats = "{}_tmp".format(repeats)
elif isinstance(repeats, int): elif isinstance(repeats, int):
repeats = [repeats] repeats = [repeats]
elif type(repeats) is np.ndarray:
repeats = repeats.tolist()
attr = { attr = {
'expand_times': repeats, 'expand_times': repeats,
"name": string(node.name), "name": string(node.name),
} }
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
"paddle.tile", "paddle.tile",
inputs={"x": val_x.name}, inputs={"x": val_x.name},
outputs=[node.name], outputs=[node.name],
repeat_times=repeats) repeat_times=repeats)
@print_mapping_info @print_mapping_info
def MaxPool(self, node): def MaxPool(self, node):
...@@ -1634,7 +1678,7 @@ class OpSet9(): ...@@ -1634,7 +1678,7 @@ class OpSet9():
pad_w = _get_same_padding(input_shape[3], kernel_shape[1], pad_w = _get_same_padding(input_shape[3], kernel_shape[1],
strides[1]) strides[1])
paddings = pad_h + pad_w paddings = pad_h + pad_w
layer_attrs = { layer_attrs = {
"kernel_size": kernel_shape, "kernel_size": kernel_shape,
"stride": strides, "stride": strides,
...@@ -1642,9 +1686,9 @@ class OpSet9(): ...@@ -1642,9 +1686,9 @@ class OpSet9():
"ceil_mode": ceil_mode, "ceil_mode": ceil_mode,
} }
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
paddle_op, paddle_op,
inputs={'x': val_x if isinstance(val_x, str) else val_x.name}, inputs={'x': val_x if isinstance(val_x, str) else val_x.name},
outputs=layer_outputs, outputs=layer_outputs,
**layer_attrs) **layer_attrs)
@print_mapping_info @print_mapping_info
...@@ -1664,9 +1708,9 @@ class OpSet9(): ...@@ -1664,9 +1708,9 @@ class OpSet9():
assert 1 <= poolnd <= 3, 'only Pool1D, Pool2D and Pool3D are supported' assert 1 <= poolnd <= 3, 'only Pool1D, Pool2D and Pool3D are supported'
output_shape = node.out_shapes[0] output_shape = node.out_shapes[0]
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
paddle_op, paddle_op,
inputs={'x': val_x.name}, inputs={'x': val_x.name},
outputs=layer_outputs, outputs=layer_outputs,
output_size=output_shape[2:]) output_size=output_shape[2:])
@print_mapping_info @print_mapping_info
...@@ -1686,9 +1730,9 @@ class OpSet9(): ...@@ -1686,9 +1730,9 @@ class OpSet9():
assert 1 <= poolnd <= 3, 'only Pool1D, Pool2D and Pool3D are supported' assert 1 <= poolnd <= 3, 'only Pool1D, Pool2D and Pool3D are supported'
output_shape = node.out_shapes[0] output_shape = node.out_shapes[0]
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
paddle_op, paddle_op,
inputs={'x': val_x.name}, inputs={'x': val_x.name},
outputs=layer_outputs, outputs=layer_outputs,
output_size=output_shape[2:]) output_size=output_shape[2:])
@print_mapping_info @print_mapping_info
...@@ -1735,30 +1779,33 @@ class OpSet9(): ...@@ -1735,30 +1779,33 @@ class OpSet9():
"dilation": dilations, "dilation": dilations,
"groups": num_groups, "groups": num_groups,
} }
remove_weight = True if val_w.name in self.done_weight_list else False remove_weight = True if val_w.name in self.done_weight_list else False
if remove_weight: if remove_weight:
self.done_weight_list.append(val_w.name) self.done_weight_list.append(val_w.name)
_rename_or_remove_weight(self.weights, val_w.name, op_name+'.weight', remove_weight) _rename_or_remove_weight(self.weights, val_w.name, op_name + '.weight',
remove_weight)
if has_bias: if has_bias:
remove_bias = True if val_b.name in self.done_weight_list else False remove_bias = True if val_b.name in self.done_weight_list else False
if remove_bias: if remove_bias:
self.done_weight_list.append(val_b_name) self.done_weight_list.append(val_b_name)
_rename_or_remove_weight(self.weights, val_b.name, op_name+'.bias', remove_bias) _rename_or_remove_weight(self.weights, val_b.name,
op_name + '.bias', remove_bias)
else: else:
layer_attrs["bias_attr"] = False layer_attrs["bias_attr"] = False
if reduce(lambda x,y:x*y, input_shape) in [1, -1] and 1 not in input_shape: if reduce(lambda x, y: x * y,
input_shape) in [1, -1] and 1 not in input_shape:
input_shape[1] = num_in_channels * num_groups input_shape[1] = num_in_channels * num_groups
input_shape[0] = 0 input_shape[0] = 0
input_shape[2] = 0 input_shape[2] = 0
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
"paddle.reshape", "paddle.reshape",
inputs=layer_inputs, inputs=layer_inputs,
outputs=[layer_inputs["x"]], outputs=[layer_inputs["x"]],
shape=input_shape) shape=input_shape)
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
paddle_op, paddle_op,
inputs=layer_inputs, inputs=layer_inputs,
outputs=layer_outputs, outputs=layer_outputs,
**layer_attrs) **layer_attrs)
@print_mapping_info @print_mapping_info
...@@ -1808,68 +1855,65 @@ class OpSet9(): ...@@ -1808,68 +1855,65 @@ class OpSet9():
"dilation": dilations, "dilation": dilations,
"padding": paddings, "padding": paddings,
"groups": num_groups, "groups": num_groups,
"output_padding":out_padding} "output_padding": out_padding
}
_rename_or_remove_weight(self.weights, val_w.name, op_name+'.weight',)
_rename_or_remove_weight(
self.weights,
val_w.name,
op_name + '.weight', )
if val_b is not None: if val_b is not None:
_rename_or_remove_weight(self.weights, val_b.name, op_name+'.bias') _rename_or_remove_weight(self.weights, val_b.name,
op_name + '.bias')
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
kernel=paddle_op, kernel=paddle_op,
inputs=inputs_dict, inputs=inputs_dict,
outputs=layer_outputs, outputs=layer_outputs,
**layer_attrs) **layer_attrs)
@print_mapping_info @print_mapping_info
def ArgMax(self, node): def ArgMax(self, node):
val_x = self.graph.get_input_node(node, idx=0, copy=True) val_x = self.graph.get_input_node(node, idx=0, copy=True)
axis = node.get_attr('axis') axis = node.get_attr('axis')
keepdims = False if node.get_attr('keepdims') == 0 else True keepdims = False if node.get_attr('keepdims') == 0 else True
layer_attrs = {'axis': axis, layer_attrs = {'axis': axis, 'keepdim': keepdims}
'keepdim': keepdims}
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
'paddle.argmax', 'paddle.argmax',
inputs={"x": val_x.name}, inputs={"x": val_x.name},
outputs=[node.name], outputs=[node.name],
**layer_attrs) **layer_attrs)
@print_mapping_info @print_mapping_info
def Size(self, node): def Size(self, node):
val_x = self.graph.get_input_node(node, idx=0, copy=True) val_x = self.graph.get_input_node(node, idx=0, copy=True)
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
"paddle.shape", "paddle.shape", inputs={"input": val_x.name}, outputs=[node.name])
inputs={"input": val_x.name},
outputs=[node.name])
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
'paddle.cast', 'paddle.cast',
inputs={"x": node.name}, inputs={"x": node.name},
outputs=[node.name], outputs=[node.name],
dtype=string('int64')) dtype=string('int64'))
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
"paddle.prod", "paddle.prod", inputs={"x": node.name}, outputs=[node.name])
inputs={"x": node.name},
outputs=[node.name])
@print_mapping_info @print_mapping_info
def Sign(self, node): def Sign(self, node):
val_x = self.graph.get_input_node(node, idx=0, copy=True) val_x = self.graph.get_input_node(node, idx=0, copy=True)
if node.dtype not in ["float16", "float32", "float64"]: if node.dtype not in ["float16", "float32", "float64"]:
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
"paddle.cast", "paddle.cast",
inputs={"x": val_x.name}, inputs={"x": val_x.name},
outputs=[val_x.name], outputs=[val_x.name],
dtype=string("float32")) dtype=string("float32"))
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
"paddle.sign", "paddle.sign", inputs={"x": val_x.name}, outputs=[node.name])
inputs={"x": val_x.name},
outputs=[node.name])
if node.dtype not in ["float16", "float32", "float64"]: if node.dtype not in ["float16", "float32", "float64"]:
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
"paddle.cast", "paddle.cast",
inputs={"x": node.name}, inputs={"x": node.name},
outputs=[node.name], outputs=[node.name],
dtype=string(node.dtype)) dtype=string(node.dtype))
@print_mapping_info @print_mapping_info
def OneHot(self, node): def OneHot(self, node):
nn_op_name = name_generator("onehot", self.nn_name2id) nn_op_name = name_generator("onehot", self.nn_name2id)
...@@ -1880,20 +1924,20 @@ class OpSet9(): ...@@ -1880,20 +1924,20 @@ class OpSet9():
values = self.graph.get_input_node(node, idx=2, copy=True) values = self.graph.get_input_node(node, idx=2, copy=True)
axis = node.get_attr('axis', -1) axis = node.get_attr('axis', -1)
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
"custom_layer:OneHot", "custom_layer:OneHot",
inputs={"indices": indices.name, inputs={
"depth": depth.name, "indices": indices.name,
"values": values.name}, "depth": depth.name,
"values": values.name
},
outputs=layer_outputs, outputs=layer_outputs,
axis=axis) axis=axis)
@print_mapping_info @print_mapping_info
def Reciprocal(self, node): def Reciprocal(self, node):
val_x = self.graph.get_input_node(node, idx=0, copy=True) val_x = self.graph.get_input_node(node, idx=0, copy=True)
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
"paddle.reciprocal", "paddle.reciprocal", inputs={"x": val_x.name}, outputs=[node.name])
inputs={"x": val_x.name},
outputs=[node.name])
@print_mapping_info @print_mapping_info
def LSTM(self, node): def LSTM(self, node):
...@@ -1905,47 +1949,49 @@ class OpSet9(): ...@@ -1905,47 +1949,49 @@ class OpSet9():
exist_input_nums = 3 exist_input_nums = 3
have_bias = False have_bias = False
if input_nums > 3 and node.layer.input[3] != '': if input_nums > 3 and node.layer.input[3] != '':
bias = self.graph.get_input_node(node, idx=exist_input_nums, copy=True) bias = self.graph.get_input_node(
node, idx=exist_input_nums, copy=True)
have_bias = True have_bias = True
exist_input_nums += 1 exist_input_nums += 1
if input_nums > 4 and node.layer.input[4] != '': if input_nums > 4 and node.layer.input[4] != '':
sequence_lens = self.graph.get_input_node(node, idx=exist_input_nums, copy=True) sequence_lens = self.graph.get_input_node(
node, idx=exist_input_nums, copy=True)
exist_input_nums += 1 exist_input_nums += 1
if input_nums > 5 and node.layer.input[5] != '': if input_nums > 5 and node.layer.input[5] != '':
init_h = self.graph.get_input_node(node, idx=exist_input_nums, copy=True) init_h = self.graph.get_input_node(
node, idx=exist_input_nums, copy=True)
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
'paddle.reshape', 'paddle.reshape',
inputs={"x": init_h.name}, inputs={"x": init_h.name},
outputs=[init_h.name], outputs=[init_h.name],
shape=init_h.out_shapes[0] shape=init_h.out_shapes[0])
)
exist_input_nums += 1 exist_input_nums += 1
if input_nums > 6 and node.layer.input[6] != '': if input_nums > 6 and node.layer.input[6] != '':
init_c = self.graph.get_input_node(node, idx=exist_input_nums, copy=True) init_c = self.graph.get_input_node(
node, idx=exist_input_nums, copy=True)
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
'paddle.reshape', 'paddle.reshape',
inputs={"x": init_c.name}, inputs={"x": init_c.name},
outputs=[init_c.name], outputs=[init_c.name],
shape=init_c.out_shapes[0] shape=init_c.out_shapes[0])
)
input_weight_np = _const_weight_or_none(input_weight) input_weight_np = _const_weight_or_none(input_weight)
_rename_or_remove_weight(self.weights, input_weight.name) _rename_or_remove_weight(self.weights, input_weight.name)
hidden_size = node.get_attr('hidden_size', input_weight_np.shape[1]/4) hidden_size = node.get_attr('hidden_size', input_weight_np.shape[1] / 4)
input_size = input_weight_np.shape[2] input_size = input_weight_np.shape[2]
hidden_weight_np = _const_weight_or_none(hidden_weight) hidden_weight_np = _const_weight_or_none(hidden_weight)
_rename_or_remove_weight(self.weights, hidden_weight.name) _rename_or_remove_weight(self.weights, hidden_weight.name)
bias_np = _const_weight_or_none(bias) bias_np = _const_weight_or_none(bias)
_rename_or_remove_weight(self.weights, bias.name) _rename_or_remove_weight(self.weights, bias.name)
input_bias_np = bias_np[:, :4*hidden_size] input_bias_np = bias_np[:, :4 * hidden_size]
hidden_bias_np = bias_np[:, 4*hidden_size:] hidden_bias_np = bias_np[:, 4 * hidden_size:]
# parameters order in paddle:lstm: # parameters order in paddle:lstm:
# 1. gate order in paddle is: input, forget, cell, output. # 1. gate order in paddle is: input, forget, cell, output.
# 2. gate orfer in onnx is: input, output, forget, cell. # 2. gate orfer in onnx is: input, output, forget, cell.
def reform_weights(w, n, intervals): def reform_weights(w, n, intervals):
slices = [w[:,x * n: y * n] for x, y in intervals] slices = [w[:, x * n:y * n] for x, y in intervals]
return np.concatenate(slices, axis=1) return np.concatenate(slices, axis=1)
def transform_weight_with_bias(weights, n, intervals): def transform_weight_with_bias(weights, n, intervals):
...@@ -1959,7 +2005,7 @@ class OpSet9(): ...@@ -1959,7 +2005,7 @@ class OpSet9():
op_name = name_generator("lstm", self.nn_name2id) op_name = name_generator("lstm", self.nn_name2id)
y_out = node.output(0) y_out = node.output(0)
yh_out = node.output(1) yh_out = node.output(1)
yc_out = node.output(2) yc_out = node.output(2)
direction = node.get_attr('direction', 'forward') direction = node.get_attr('direction', 'forward')
...@@ -1973,20 +2019,24 @@ class OpSet9(): ...@@ -1973,20 +2019,24 @@ class OpSet9():
def assign_params(op_name, weights, weight_idx=0, suffix=''): def assign_params(op_name, weights, weight_idx=0, suffix=''):
param_names = generate_paddle_param_names(op_name, suffix) param_names = generate_paddle_param_names(op_name, suffix)
print(param_names)
for param_name, weight in zip(param_names, weights): for param_name, weight in zip(param_names, weights):
self.weights[param_name] = weight[weight_idx] self.weights[param_name] = weight[weight_idx]
if direction == 'backward': if direction == 'backward':
raise Exception("LSTM support 'forward' or 'bidirectional', except '{}'.".format(direction)) raise Exception(
"LSTM support 'forward' or 'bidirectional', except '{}'.".
format(direction))
else: else:
assign_params(op_name, weights) assign_params(op_name, weights)
if direction == 'bidirectional': if direction == 'bidirectional':
assign_params(op_name, weights, 1, '_reverse') assign_params(op_name, weights, 1, '_reverse')
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
'paddle.nn.LSTM', 'paddle.nn.LSTM',
inputs={'input': x.name, 'initial_states': (init_h.name, init_c.name)}, inputs={
'input': x.name,
'initial_states': (init_h.name, init_c.name)
},
outputs=[op_name, y_out, yh_out, yc_out], outputs=[op_name, y_out, yh_out, yc_out],
input_size=input_size, input_size=input_size,
hidden_size=hidden_size, hidden_size=hidden_size,
...@@ -1998,30 +2048,39 @@ class OpSet9(): ...@@ -1998,30 +2048,39 @@ class OpSet9():
'paddle.reshape', 'paddle.reshape',
inputs={"x": y_out}, inputs={"x": y_out},
outputs=[y_out], outputs=[y_out],
shape=[0, 0, -1, hidden_size] shape=[0, 0, -1, hidden_size])
)
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
'paddle.transpose', 'paddle.transpose',
inputs={"x": y_out}, inputs={"x": y_out},
outputs=[y_out], outputs=[y_out],
perm=[0,2,1,3] perm=[0, 2, 1, 3])
)
@print_mapping_info @print_mapping_info
def TopK(self, node): def TopK(self, node):
val_x = self.graph.get_input_node(node, idx=0, copy=True) val_x = self.graph.get_input_node(node, idx=0, copy=True)
val_k = self.graph.get_input_node(node, idx=1, copy=True) val_k = self.graph.get_input_node(node, idx=1, copy=True)
if val_k.dtype != "int32":
self.paddle_graph.add_layer(
"paddle.cast",
inputs={"x": val_k.name},
outputs=[val_k.name],
dtype=string('int32'))
layer_attrs = dict() layer_attrs = dict()
layer_attrs["axis"] = node.get_attr('axis', -1) layer_attrs["axis"] = node.get_attr('axis', -1)
layer_attrs["largest"] = True if node.get_attr('largest', 1) == 1 else False layer_attrs["largest"] = True if node.get_attr('largest',
layer_attrs["sorted"] = True if node.get_attr('sorted', 1) == 1 else False 1) == 1 else False
layer_attrs["sorted"] = True if node.get_attr('sorted',
1) == 1 else False
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
"paddle.topk", "paddle.topk",
inputs={"x": val_x.name, inputs={"x": val_x.name,
"k": val_k.name}, "k": val_k.name},
outputs=["{}_p{}".format(node.layer_name, 0), "{}_p{}".format(node.layer_name, 1)], outputs=[
"{}_p{}".format(node.layer_name, 0),
"{}_p{}".format(node.layer_name, 1)
],
**layer_attrs) **layer_attrs)
@print_mapping_info @print_mapping_info
def LRN(self, node): def LRN(self, node):
op_name = name_generator("lrn", self.nn_name2id) op_name = name_generator("lrn", self.nn_name2id)
...@@ -2032,18 +2091,13 @@ class OpSet9(): ...@@ -2032,18 +2091,13 @@ class OpSet9():
beta = node.get_attr('beta', 0.75) beta = node.get_attr('beta', 0.75)
bias = node.get_attr('bias', 1.0) bias = node.get_attr('bias', 1.0)
size = node.get_attr('size') size = node.get_attr('size')
layer_attrs = { layer_attrs = {'size': size, 'alpha': alpha, 'beta': beta, 'k': bias}
'size': size,
'alpha': alpha,
'beta': beta,
'k': bias
}
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
"custom_layer:LocalResponseNorm", "custom_layer:LocalResponseNorm",
inputs={"x": val_x.name}, inputs={"x": val_x.name},
outputs=layer_outputs, outputs=layer_outputs,
**layer_attrs) **layer_attrs)
@print_mapping_info @print_mapping_info
def DepthToSpace(self, node): def DepthToSpace(self, node):
val_x = self.graph.get_input_node(node, idx=0, copy=True) val_x = self.graph.get_input_node(node, idx=0, copy=True)
...@@ -2056,37 +2110,65 @@ class OpSet9(): ...@@ -2056,37 +2110,65 @@ class OpSet9():
'paddle.reshape', 'paddle.reshape',
inputs={"x": val_x.name}, inputs={"x": val_x.name},
outputs=[node.name], outputs=[node.name],
shape=[b, blocksize, blocksize, c // (blocksize**2), h, w] shape=[b, blocksize, blocksize, c // (blocksize**2), h, w])
)
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
'paddle.transpose', 'paddle.transpose',
inputs={"x": node.name}, inputs={"x": node.name},
outputs=[node.name], outputs=[node.name],
perm=[0, 3, 4, 1, 5, 2] perm=[0, 3, 4, 1, 5, 2])
)
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
'paddle.reshape', 'paddle.reshape',
inputs={"x": node.name}, inputs={"x": node.name},
outputs=[node.name], outputs=[node.name],
shape=[b, c // (blocksize**2), h * blocksize, w * blocksize] shape=[b, c // (blocksize**2), h * blocksize, w * blocksize])
)
else: else:
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
'paddle.reshape', 'paddle.reshape',
inputs={"x": val_x.name}, inputs={"x": val_x.name},
outputs=[node.name], outputs=[node.name],
shape=[b, c // (blocksize ** 2), blocksize, blocksize, h, w] shape=[b, c // (blocksize**2), blocksize, blocksize, h, w])
)
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
'paddle.transpose', 'paddle.transpose',
inputs={"x": node.name}, inputs={"x": node.name},
outputs=[node.name], outputs=[node.name],
perm=[0, 1, 4, 2, 5, 3] perm=[0, 1, 4, 2, 5, 3])
)
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
'paddle.reshape', 'paddle.reshape',
inputs={"x": node.name}, inputs={"x": node.name},
outputs=[node.name], outputs=[node.name],
shape=[b, c // (blocksize ** 2), h * blocksize, w * blocksize] shape=[b, c // (blocksize**2), h * blocksize, w * blocksize])
)
@print_mapping_info
\ No newline at end of file def NonMaxSuppression(self, node):
nn_op_name = name_generator("nms", self.nn_name2id)
output_name = node.name
layer_outputs = [nn_op_name, output_name]
boxes = self.graph.get_input_node(node, idx=0, copy=True)
scores = self.graph.get_input_node(node, idx=1, copy=True)
inputs_len = len(node.layer.input)
layer_attrs = dict()
if inputs_len > 2:
max_output_boxes_per_class = self.graph.get_input_node(
node, idx=2, copy=True)
layer_attrs["nms_top_k"] = _const_weight_or_none(
max_output_boxes_per_class).tolist()[0]
else:
layer_attrs["nms_top_k"] = 0
if inputs_len > 3:
iou_threshold = self.graph.get_input_node(node, idx=3, copy=True)
layer_attrs["nms_threshold"] = _const_weight_or_none(
iou_threshold).tolist()[0]
else:
layer_attrs["nms_threshold"] = 0.0
if inputs_len > 4:
score_threshold = self.graph.get_input_node(node, idx=4, copy=True)
layer_attrs["score_threshold"] = _const_weight_or_none(
score_threshold).tolist()[0]
else:
layer_attrs["score_threshold"] = 0.0
self.paddle_graph.add_layer(
"custom_layer:NMS",
inputs={"bboxes": boxes.name,
"scores": scores.name},
outputs=layer_outputs,
**layer_attrs)
x2paddle/op_mapper/static/onnx2paddle/opset9/opset.py
浏览文件 @ 4a2df7e1
...@@ -89,48 +89,60 @@ class OpSet9(): ...@@ -89,48 +89,60 @@ class OpSet9():
'Sub': 'paddle.subtract', 'Sub': 'paddle.subtract',
'Mul': 'paddle.multiply', 'Mul': 'paddle.multiply',
'Pow': 'paddle.pow', 'Pow': 'paddle.pow',
'Less': 'paddle.less_than',
} }
directly_map_ops = { directly_map_ops = {
'Ceil': ['paddle.ceil'], 'Ceil': ['paddle.ceil'],
# reduce function # reduce function
'ReduceMean': ['paddle.mean', 'ReduceMean': [
dict(axes='axis', keepdims='keepdim'), 'paddle.mean', dict(
dict(axes=None, keepdims=1)], axes='axis', keepdims='keepdim'), dict(
'ReduceSum': ['paddle.sum', axes=None, keepdims=1)
dict(axes='axis', keepdims='keepdim'), ],
dict(axes=None, keepdims=1)], 'ReduceSum': [
'ReduceMin': ['paddle.min', 'paddle.sum', dict(
dict(axes='axis', keepdims='keepdim'), axes='axis', keepdims='keepdim'), dict(
dict(axes=None, keepdim=1)], axes=None, keepdims=1)
'ReduceMax': ['paddle.max', ],
dict(axes='axis', keepdims='keepdim'), 'ReduceMin': [
dict(axes=None, keepdim=1)], 'paddle.min', dict(
'ReduceProd': ['paddle.prod', axes='axis', keepdims='keepdim'), dict(
dict(axes='axis', keepdims='keepdim'), axes=None, keepdim=1)
dict(axes=None, keepdim=1)], ],
'ReduceMax': [
'paddle.max', dict(
axes='axis', keepdims='keepdim'), dict(
axes=None, keepdim=1)
],
'ReduceProd': [
'paddle.prod', dict(
axes='axis', keepdims='keepdim'), dict(
axes=None, keepdim=1)
],
# active function # active function
'Relu': ['paddle.nn.functional.relu'], 'Relu': ['paddle.nn.functional.relu'],
'LeakyRelu': ['paddle.nn.functional.leaky_relu', 'LeakyRelu': [
dict(alpha='negative_slope'), 'paddle.nn.functional.leaky_relu', dict(alpha='negative_slope'),
dict(negative_slope=.01)], dict(negative_slope=.01)
'Elu': ['paddle.nn.functional.elu', ],
dict(alpha='alpha'), 'Elu':
dict(alpha=1.)], ['paddle.nn.functional.elu', dict(alpha='alpha'), dict(alpha=1.)],
'ThresholdedRelu': ['paddle.nn.functional.thresholded_relu', 'ThresholdedRelu': [
dict(alpha='threshold'), 'paddle.nn.functional.thresholded_relu', dict(alpha='threshold'),
dict(alpha=1.)], dict(alpha=1.)
],
'Tanh': ['paddle.nn.functional.tanh'], 'Tanh': ['paddle.nn.functional.tanh'],
'Sigmoid': ['paddle.nn.functional.sigmoid'], 'Sigmoid': ['paddle.nn.functional.sigmoid'],
'Softsign': ['paddle.nn.functional.softsign'], 'Softsign': ['paddle.nn.functional.softsign'],
'Softplus': ['paddle.nn.functional.softplus', 'Softplus': [
dict(threshold='threshold'), 'paddle.nn.functional.softplus', dict(threshold='threshold'),
dict(threshold=float(sys.maxsize))], dict(threshold=float(sys.maxsize))
],
'Exp': ['paddle.exp'], 'Exp': ['paddle.exp'],
'Log': ['paddle.log'], 'Log': ['paddle.log'],
'Softmax': ['paddle.nn.functional.softmax', 'Softmax':
dict(axis='axis'), ['paddle.nn.functional.softmax', dict(axis='axis'), dict(axis=1)],
dict(axis=1)],
'Sqrt': ['paddle.sqrt'], 'Sqrt': ['paddle.sqrt'],
'Floor': ['paddle.floor'], 'Floor': ['paddle.floor'],
'Abs': ['paddle.abs'], 'Abs': ['paddle.abs'],
...@@ -170,19 +182,16 @@ class OpSet9(): ...@@ -170,19 +182,16 @@ class OpSet9():
inputs={"x": input.name}, inputs={"x": input.name},
outputs=[node.name], outputs=[node.name],
**layer_attrs) **layer_attrs)
@print_mapping_info @print_mapping_info
def elementwise_map(self, node): def elementwise_map(self, node):
op_type = self.elementwise_ops[node.layer_type] op_type = self.elementwise_ops[node.layer_type]
val_x = self.graph.get_input_node(node, idx=0, copy=True) val_x = self.graph.get_input_node(node, idx=0, copy=True)
val_y = self.graph.get_input_node(node, idx=1, copy=True) val_y = self.graph.get_input_node(node, idx=1, copy=True)
inputs_dict = {'x': val_x.name, inputs_dict = {'x': val_x.name, 'y': val_y.name}
'y': val_y.name}
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
op_type, op_type, inputs=inputs_dict, outputs=[node.name])
inputs=inputs_dict,
outputs=[node.name])
@print_mapping_info @print_mapping_info
def place_holder(self, node): def place_holder(self, node):
shape = node.out_shapes[0] shape = node.out_shapes[0]
...@@ -209,8 +218,8 @@ class OpSet9(): ...@@ -209,8 +218,8 @@ class OpSet9():
shape = node.out_shapes[0] shape = node.out_shapes[0]
if hasattr(node.weight, "shape") and len(node.weight.shape) == 0: if hasattr(node.weight, "shape") and len(node.weight.shape) == 0:
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
"paddle.full", "paddle.full",
inputs={}, inputs={},
outputs=[node.name], outputs=[node.name],
dtype=string(dtype), dtype=string(dtype),
shape=[1], shape=[1],
...@@ -247,17 +256,19 @@ class OpSet9(): ...@@ -247,17 +256,19 @@ class OpSet9():
if len(node.layer.input) == 2: if len(node.layer.input) == 2:
# opset 10 # opset 10
val_scales = self.graph.get_input_node(node, idx=1, copy=True) val_scales = self.graph.get_input_node(node, idx=1, copy=True)
# TODO(syf): paddle.nn.functional.interpolate will support the length # TODO(syf): paddle.nn.functional.interpolate will support the length
# which is the same as the rank of input. # which is the same as the rank of input.
# inputs['scale_factor'] = val_scales.name # inputs['scale_factor'] = val_scales.name
attrs['scale_factor'] = self.params[val_scales.name].tolist()[2:] attrs['scale_factor'] = self.params[val_scales.name].tolist()[
2:]
elif len(node.layer.input) == 3: elif len(node.layer.input) == 3:
# opset 11 # opset 11
val_scales = self.graph.get_input_node(node, idx=2, copy=True) val_scales = self.graph.get_input_node(node, idx=2, copy=True)
# TODO(syf): paddle.nn.functional.interpolate will support the length # TODO(syf): paddle.nn.functional.interpolate will support the length
# which is the same as the rank of input. # which is the same as the rank of input.
# inputs['scale_factor'] = val_scales.name # inputs['scale_factor'] = val_scales.name
attrs['scale_factor'] = self.params[val_scales.name].tolist()[2:] attrs['scale_factor'] = self.params[val_scales.name].tolist()[
2:]
elif len(node.layer.input) == 4: elif len(node.layer.input) == 4:
# opset 11 # opset 11
val_sizes = self.graph.get_input_node(node, idx=3, copy=True) val_sizes = self.graph.get_input_node(node, idx=3, copy=True)
...@@ -274,8 +285,17 @@ class OpSet9(): ...@@ -274,8 +285,17 @@ class OpSet9():
outputs=[var_hw], outputs=[var_hw],
dtype=string('int32')) dtype=string('int32'))
inputs['size'] = var_hw inputs['size'] = var_hw
attrs = {"align_corners": False, attrs = {
"mode": string(node.get_attr('mode', 'nearest'))} "align_corners": False,
"mode": string(node.get_attr('mode', 'nearest'))
}
mode = node.get_attr('mode', 'nearest')
if mode == "linear":
attrs["mode"] = string("bilinear")
if node.get_attr('coordinate_transformation_mode',
'half_pixel') == 'pytorch_half_pixel':
attrs["align_corners"] = False
attrs["align_mode"] = 0
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
kernel="paddle.nn.functional.interpolate", kernel="paddle.nn.functional.interpolate",
inputs=inputs, inputs=inputs,
...@@ -294,19 +314,26 @@ class OpSet9(): ...@@ -294,19 +314,26 @@ class OpSet9():
inputs['scale_factor'] = val_scales.name inputs['scale_factor'] = val_scales.name
mode = node.get_attr('mode', 'nearest') mode = node.get_attr('mode', 'nearest')
attrs.update({"align_corners": False, attrs.update({
"mode": string(mode), "align_corners": False,
"align_mode": 1}) "mode": string(mode),
"align_mode": 1
})
val_x_shape = val_x.out_shapes[0] val_x_shape = val_x.out_shapes[0]
if mode == "linear" and len(val_x_shape) == 4: if mode == "linear" and len(val_x_shape) == 4:
attrs["mode"] = string("bilinear") attrs["mode"] = string("bilinear")
attrs["align_corners"] = True if node.get_attr('coordinate_transformation_mode',
'half_pixel') == 'pytorch_half_pixel':
attrs["align_corners"] = False
attrs["align_mode"] = 0
else:
attrs["align_corners"] = True
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
kernel="paddle.nn.functional.interpolate", kernel="paddle.nn.functional.interpolate",
inputs=inputs, inputs=inputs,
outputs=[node.name], outputs=[node.name],
**attrs) **attrs)
@print_mapping_info @print_mapping_info
def HardSigmoid(self, node): def HardSigmoid(self, node):
val_x = self.graph.get_input_node(node, idx=0, copy=True) val_x = self.graph.get_input_node(node, idx=0, copy=True)
...@@ -323,8 +350,8 @@ class OpSet9(): ...@@ -323,8 +350,8 @@ class OpSet9():
inputs={"x": node.name + "_val"}, inputs={"x": node.name + "_val"},
outputs=[node.name], outputs=[node.name],
min=0.0, min=0.0,
max=1.0) max=1.0)
@print_mapping_info @print_mapping_info
def Shape(self, node): def Shape(self, node):
val_x = self.graph.get_input_node(node, idx=0, copy=True) val_x = self.graph.get_input_node(node, idx=0, copy=True)
...@@ -333,10 +360,10 @@ class OpSet9(): ...@@ -333,10 +360,10 @@ class OpSet9():
inputs={"input": val_x.name}, inputs={"input": val_x.name},
outputs=[node.name]) outputs=[node.name])
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
'paddle.cast', 'paddle.cast',
inputs={"x": node.name}, inputs={"x": node.name},
outputs=[node.name], outputs=[node.name],
dtype=string('int64')) dtype=string('int64'))
@print_mapping_info @print_mapping_info
def RoiAlign(self, node): def RoiAlign(self, node):
...@@ -408,12 +435,14 @@ class OpSet9(): ...@@ -408,12 +435,14 @@ class OpSet9():
paddings = [] paddings = []
paddle_op = 'paddle.nn.functional.pad' paddle_op = 'paddle.nn.functional.pad'
if len(pads) == 10 and sum(pads) == 0: if len(pads) == 10 and sum(pads) == 0:
pads = pads[0: 6] pads = pads[0:6]
if len(pads) in [2, 4, 6]: if len(pads) in [2, 4, 6]:
if data_shape: if data_shape:
assume_pad |= data_shape and 2 * (len(data_shape) - 2) == len(pads) # NCHW assume_pad |= data_shape and 2 * (len(data_shape) - 2
) == len(pads) # NCHW
if output_shape: if output_shape:
assume_pad |= output_shape and 2 * (len(output_shape) - 2) == len(pads) # NCHW assume_pad |= output_shape and 2 * (len(output_shape) - 2
) == len(pads) # NCHW
if assume_pad: if assume_pad:
if len(pads) == 2: if len(pads) == 2:
data_format = "NCL" data_format = "NCL"
...@@ -421,7 +450,7 @@ class OpSet9(): ...@@ -421,7 +450,7 @@ class OpSet9():
data_format = "NCHW" data_format = "NCHW"
else: else:
data_format = "NCDHW" data_format = "NCDHW"
paddings = np.array(pads).reshape( paddings = np.array(pads).reshape(
(2, -1)).transpose().astype("int32") (2, -1)).transpose().astype("int32")
paddings = np.flip(paddings, axis=0).flatten().tolist() paddings = np.flip(paddings, axis=0).flatten().tolist()
...@@ -429,20 +458,26 @@ class OpSet9(): ...@@ -429,20 +458,26 @@ class OpSet9():
layer_attrs['data_format'] = string(data_format) layer_attrs['data_format'] = string(data_format)
else: else:
if data_shape: if data_shape:
assume_pad |= data_shape and 2 * len(data_shape) == len(pads) # NCHW assume_pad |= data_shape and 2 * len(data_shape) == len(
pads) # NCHW
if output_shape: if output_shape:
assume_pad |= output_shape and 2 * len(output_shape) == len(pads) # NCHW assume_pad |= output_shape and 2 * len(
output_shape) == len(pads) # NCHW
if assume_pad: if assume_pad:
paddings = np.array(pads).reshape( paddings = np.array(pads).reshape(
(2, -1)).transpose().astype("int32").flatten().tolist() (2,
-1)).transpose().astype("int32").flatten().tolist()
layer_attrs['pad'] = paddings layer_attrs['pad'] = paddings
else: else:
raise Exception("The padding value {} is wrong!".format(pads)) raise Exception("The padding value {} is wrong!".format(
pads))
elif len(pads) == 8: elif len(pads) == 8:
if data_shape: if data_shape:
assume_pad |= data_shape and 2 * len(data_shape) == len(pads) # NCHW assume_pad |= data_shape and 2 * len(data_shape) == len(
pads) # NCHW
if output_shape: if output_shape:
assume_pad |= output_shape and 2 * len(output_shape) == len(pads) # NCHW assume_pad |= output_shape and 2 * len(output_shape) == len(
pads) # NCHW
if assume_pad: if assume_pad:
paddings = np.array(pads).reshape( paddings = np.array(pads).reshape(
(2, -1)).transpose().astype("int32") (2, -1)).transpose().astype("int32")
...@@ -454,11 +489,11 @@ class OpSet9(): ...@@ -454,11 +489,11 @@ class OpSet9():
layer_attrs['pad'] = paddings layer_attrs['pad'] = paddings
paddle_op = "custom_layer:pad_all_dim4_one_input" paddle_op = "custom_layer:pad_all_dim4_one_input"
else: else:
raise Exception("The padding value {} is wrong!".format(pads)) raise Exception("The padding value {} is wrong!".format(pads))
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
paddle_op, paddle_op,
inputs={'x': val_x.name}, inputs={'x': val_x.name},
outputs=layer_outputs, outputs=layer_outputs,
**layer_attrs) **layer_attrs)
if not op_independent: if not op_independent:
return node.name + '_paded' return node.name + '_paded'
...@@ -466,9 +501,11 @@ class OpSet9(): ...@@ -466,9 +501,11 @@ class OpSet9():
pads_len = val_pad.out_shapes[0][0] pads_len = val_pad.out_shapes[0][0]
if pads_len in [2, 4, 6]: if pads_len in [2, 4, 6]:
if data_shape: if data_shape:
assume_pad |= data_shape and 2 * (len(data_shape) - 2) == pads_len # NCHW assume_pad |= data_shape and 2 * (len(data_shape) - 2
) == pads_len # NCHW
if output_shape: if output_shape:
assume_pad |= output_shape and 2 * (len(output_shape) - 2) == pads_len # NCHW assume_pad |= output_shape and 2 * (len(output_shape) - 2
) == pads_len # NCHW
if assume_pad: if assume_pad:
if pads_len == 2: if pads_len == 2:
data_format = "NCL" data_format = "NCL"
...@@ -477,42 +514,49 @@ class OpSet9(): ...@@ -477,42 +514,49 @@ class OpSet9():
else: else:
data_format = "NCDHW" data_format = "NCDHW"
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
"custom_layer:pad_with_two_input", "custom_layer:pad_with_two_input",
inputs={'x': val_x.name, 'pad': val_pad.name}, inputs={'x': val_x.name,
'pad': val_pad.name},
outputs=layer_outputs, outputs=layer_outputs,
value=value, value=value,
mode=string(mode), mode=string(mode),
data_format=string(data_format)) data_format=string(data_format))
else: else:
if data_shape: if data_shape:
assume_pad |= data_shape and 2 * len(data_shape) == pads_len # NCHW assume_pad |= data_shape and 2 * len(
data_shape) == pads_len # NCHW
if output_shape: if output_shape:
assume_pad |= output_shape and 2 * len(output_shape) == pads_len # NCHW assume_pad |= output_shape and 2 * len(
output_shape) == pads_len # NCHW
if assume_pad: if assume_pad:
if pads_len == 4: if pads_len == 4:
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
"custom_layer:pad_all_dim2", "custom_layer:pad_all_dim2",
inputs={'x': val_x.name, 'pad': val_pad.name}, inputs={'x': val_x.name,
outputs=layer_outputs, 'pad': val_pad.name},
outputs=layer_outputs,
value=value, value=value,
mode=string(mode)) mode=string(mode))
else: else:
raise Exception("The padding value is wrong!") raise Exception("The padding value is wrong!")
elif pads_len == 8: elif pads_len == 8:
if data_shape: if data_shape:
assume_pad |= data_shape and 2 * len(data_shape) == pads_len # NCHW assume_pad |= data_shape and 2 * len(
data_shape) == pads_len # NCHW
if output_shape: if output_shape:
assume_pad |= output_shape and 2 * len(output_shape) == pads_len # NCHW assume_pad |= output_shape and 2 * len(
output_shape) == pads_len # NCHW
if assume_pad: if assume_pad:
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
"custom_layer:pad_all_dim4", "custom_layer:pad_all_dim4",
inputs={'x': val_x.name, 'pad': val_pad.name}, inputs={'x': val_x.name,
outputs=layer_outputs, 'pad': val_pad.name},
outputs=layer_outputs,
value=value, value=value,
mode=string(mode)) mode=string(mode))
else: else:
print(pads_len) print(pads_len)
raise Exception("The padding value is wrong!") raise Exception("The padding value is wrong!")
if not op_independent: if not op_independent:
return node.name + '_paded' return node.name + '_paded'
...@@ -530,8 +574,8 @@ class OpSet9(): ...@@ -530,8 +574,8 @@ class OpSet9():
shape=[1]) shape=[1])
else: else:
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
'paddle.unsqueeze', 'paddle.unsqueeze',
inputs={"x": val_x.name}, inputs={"x": val_x.name},
outputs=[node.name], outputs=[node.name],
**layer_attrs) **layer_attrs)
...@@ -542,9 +586,9 @@ class OpSet9(): ...@@ -542,9 +586,9 @@ class OpSet9():
lambd = node.get_attr('lambd') lambd = node.get_attr('lambd')
assert bias == 0.0, 'not support bias!=0' assert bias == 0.0, 'not support bias!=0'
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
'paddle.nn.functional.hardshrink', 'paddle.nn.functional.hardshrink',
inputs={"x": val_x.name}, inputs={"x": val_x.name},
outputs=[node.name], outputs=[node.name],
threshold=lambd) threshold=lambd)
@print_mapping_info @print_mapping_info
...@@ -571,8 +615,8 @@ class OpSet9(): ...@@ -571,8 +615,8 @@ class OpSet9():
value = value.tolist() value = value.tolist()
value = value[0] value = value[0]
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
"paddle.full", "paddle.full",
inputs={}, inputs={},
outputs=[node.name], outputs=[node.name],
dtype=string(dtype), dtype=string(dtype),
shape=[1], shape=[1],
...@@ -603,11 +647,9 @@ class OpSet9(): ...@@ -603,11 +647,9 @@ class OpSet9():
val_scale = self.graph.get_input_node(node, idx=1, copy=True) val_scale = self.graph.get_input_node(node, idx=1, copy=True)
val_b = self.graph.get_input_node(node, idx=2, copy=True) val_b = self.graph.get_input_node(node, idx=2, copy=True)
epsilon = node.get_attr('epsilon', 1e-5) epsilon = node.get_attr('epsilon', 1e-5)
layer_attrs = { layer_attrs = {'eps': epsilon, }
'eps': epsilon,
}
dim = len(val_x.out_shapes[0]) dim = len(val_x.out_shapes[0])
if dim ==2 : if dim == 2:
layer_attrs["data_format"] = string("NC") layer_attrs["data_format"] = string("NC")
elif dim == 3: elif dim == 3:
layer_attrs["data_format"] = string("NCL") layer_attrs["data_format"] = string("NCL")
...@@ -616,13 +658,17 @@ class OpSet9(): ...@@ -616,13 +658,17 @@ class OpSet9():
elif dim == 5: elif dim == 5:
layer_attrs["data_format"] = string("NCDHW") layer_attrs["data_format"] = string("NCDHW")
else: else:
raise Exception("The paddle only support 2D, 3D, 4D or 5D input in InstanceNormalization.") raise Exception(
"The paddle only support 2D, 3D, 4D or 5D input in InstanceNormalization."
)
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
"paddle.nn.functional.instance_norm", "paddle.nn.functional.instance_norm",
inputs={"x": val_x.name, inputs={
"weight": val_scale.name, "x": val_x.name,
"bias": val_b.name}, "weight": val_scale.name,
outputs=[node.name], "bias": val_b.name
},
outputs=[node.name],
**layer_attrs) **layer_attrs)
@print_mapping_info @print_mapping_info
...@@ -637,16 +683,10 @@ class OpSet9(): ...@@ -637,16 +683,10 @@ class OpSet9():
'fill_value': 1 'fill_value': 1
} }
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
'paddle.full', 'paddle.full', inputs={}, outputs=[name_ones], **attr_ones)
inputs={}, inputs_dict = {'x': name_ones, 'y': val_x.name}
outputs=[name_ones],
**attr_ones)
inputs_dict = {'x': name_ones,
'y': val_x.name}
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
'paddle.multiply', 'paddle.multiply', inputs=inputs_dict, outputs=[node.name])
inputs=inputs_dict,
outputs=[node.name])
@print_mapping_info @print_mapping_info
def Gather(self, node): def Gather(self, node):
...@@ -700,9 +740,9 @@ class OpSet9(): ...@@ -700,9 +740,9 @@ class OpSet9():
for i in range(len(perm)): for i in range(len(perm)):
new_perm[perm[i]] = i new_perm[perm[i]] = i
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
'paddle.transpose', 'paddle.transpose',
inputs={"x": node.name}, inputs={"x": node.name},
outputs=[node.name], outputs=[node.name],
perm=new_perm) perm=new_perm)
if len(indices_shape) < 1: if len(indices_shape) < 1:
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
...@@ -804,9 +844,11 @@ class OpSet9(): ...@@ -804,9 +844,11 @@ class OpSet9():
if len(indices.out_shapes[0]) == 1: if len(indices.out_shapes[0]) == 1:
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
'paddle.scatter', 'paddle.scatter',
inputs={'x': val_x.name, inputs={
'index': indices.name, 'x': val_x.name,
'updates': updates.name}, 'index': indices.name,
'updates': updates.name
},
outputs=[node.name]) outputs=[node.name])
else: else:
input_inner_indices = node.name + '_input_inner_indices' input_inner_indices = node.name + '_input_inner_indices'
...@@ -881,9 +923,11 @@ class OpSet9(): ...@@ -881,9 +923,11 @@ class OpSet9():
val_limit = self.graph.get_input_node(node, idx=1, copy=True) val_limit = self.graph.get_input_node(node, idx=1, copy=True)
val_delta = self.graph.get_input_node(node, idx=2, copy=True) val_delta = self.graph.get_input_node(node, idx=2, copy=True)
dtype = val_start.dtype dtype = val_start.dtype
inputs = {'start': val_start.name, inputs = {
'end': val_limit.name, 'start': val_start.name,
'step': val_delta.name} 'end': val_limit.name,
'step': val_delta.name
}
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
'paddle.arange', 'paddle.arange',
inputs=inputs, inputs=inputs,
...@@ -913,7 +957,7 @@ class OpSet9(): ...@@ -913,7 +957,7 @@ class OpSet9():
if len(node.inputs) > 4: if len(node.inputs) > 4:
steps = self.graph.get_input_node(node, idx=4, copy=True) steps = self.graph.get_input_node(node, idx=4, copy=True)
steps = _const_weight_or_none(steps).tolist() steps = _const_weight_or_none(steps).tolist()
layer_attrs = { layer_attrs = {
"axes": axes, "axes": axes,
"starts": starts.name, "starts": starts.name,
...@@ -927,7 +971,8 @@ class OpSet9(): ...@@ -927,7 +971,8 @@ class OpSet9():
# ends_value[idx] = 2**31 - 1 # ends_value[idx] = 2**31 - 1
#print(val_x.out_shapes) #print(val_x.out_shapes)
for idx in range(len(ends_value)): for idx in range(len(ends_value)):
if starts_value[idx] >= val_x.out_shapes[0][axes[idx]] and val_x.out_shapes[0][axes[idx]] > 0: if starts_value[idx] >= val_x.out_shapes[0][axes[
idx]] and val_x.out_shapes[0][axes[idx]] > 0:
starts_value[idx] = val_x.out_shapes[0][axes[idx]] - 1 starts_value[idx] = val_x.out_shapes[0][axes[idx]] - 1
ends_value[idx] = val_x.out_shapes[0][axes[idx]] ends_value[idx] = val_x.out_shapes[0][axes[idx]]
elif ends_value[idx] > 2**31 - 1: elif ends_value[idx] > 2**31 - 1:
...@@ -965,19 +1010,18 @@ class OpSet9(): ...@@ -965,19 +1010,18 @@ class OpSet9():
ends[idx] = 2**31 - 1 ends[idx] = 2**31 - 1
layer_attrs = {"axes": axes, "starts": starts, "ends": ends} layer_attrs = {"axes": axes, "starts": starts, "ends": ends}
if steps is not None: if steps is not None:
layer_attrs['strides'] = steps layer_attrs['strides'] = steps
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
'paddle.strided_slice', 'paddle.strided_slice',
inputs={"x": val_x.name}, inputs={"x": val_x.name},
outputs=[node.name], outputs=[node.name],
**layer_attrs) **layer_attrs)
else: else:
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
'paddle.slice', 'paddle.slice',
inputs={"input": val_x.name}, inputs={"input": val_x.name},
outputs=[node.name], outputs=[node.name],
**layer_attrs) **layer_attrs)
@print_mapping_info @print_mapping_info
...@@ -992,13 +1036,10 @@ class OpSet9(): ...@@ -992,13 +1036,10 @@ class OpSet9():
'this is not supported') 'this is not supported')
if len(value) == 1: if len(value) == 1:
value = value[0] value = value[0]
layer_attrs = { layer_attrs = {'dtype': string(dtype), 'fill_value': value}
'dtype': string(dtype),
'fill_value': value
}
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
"paddle.full", "paddle.full",
inputs={'shape': val_shape.name}, inputs={'shape': val_shape.name},
outputs=[node.name], outputs=[node.name],
**layer_attrs) **layer_attrs)
...@@ -1015,9 +1056,9 @@ class OpSet9(): ...@@ -1015,9 +1056,9 @@ class OpSet9():
'min': min_value, 'min': min_value,
} }
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
'paddle.clip', 'paddle.clip',
inputs={"x": val_x.name}, inputs={"x": val_x.name},
outputs=[node.name], outputs=[node.name],
**layer_attrs) **layer_attrs)
else: else:
min_ipt = self.graph.get_input_node(node, idx=1, copy=True) min_ipt = self.graph.get_input_node(node, idx=1, copy=True)
...@@ -1031,9 +1072,9 @@ class OpSet9(): ...@@ -1031,9 +1072,9 @@ class OpSet9():
if max_value is not None and min_value is not None: if max_value is not None and min_value is not None:
layer_attrs = {'max': max_value, 'min': min_value} layer_attrs = {'max': max_value, 'min': min_value}
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
'paddle.clip', 'paddle.clip',
inputs={"x": val_x.name}, inputs={"x": val_x.name},
outputs=[node.name], outputs=[node.name],
**layer_attrs) **layer_attrs)
else: else:
raise raise
...@@ -1058,9 +1099,9 @@ class OpSet9(): ...@@ -1058,9 +1099,9 @@ class OpSet9():
else: else:
outputs_list.append(node.name) outputs_list.append(node.name)
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
'paddle.split', 'paddle.split',
inputs={"x": val_x.name}, inputs={"x": val_x.name},
outputs=outputs_list, outputs=outputs_list,
**layer_attrs) **layer_attrs)
@print_mapping_info @print_mapping_info
...@@ -1117,17 +1158,18 @@ class OpSet9(): ...@@ -1117,17 +1158,18 @@ class OpSet9():
if output_dtype: if output_dtype:
assert dtype == output_dtype, 'dtype of to unmatches output' assert dtype == output_dtype, 'dtype of to unmatches output'
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
'paddle.cast', 'paddle.cast',
inputs={'x': val_input.name}, inputs={'x': val_input.name},
outputs=[node.name], outputs=[node.name],
dtype=string(dtype)) dtype=string(dtype))
@print_mapping_info @print_mapping_info
def Not(self, node): def Not(self, node):
val_input = self.graph.get_input_node(node, idx=0, copy=True) val_input = self.graph.get_input_node(node, idx=0, copy=True)
self.paddle_graph.add_layer('paddle.logical_not', self.paddle_graph.add_layer(
inputs={'x': val_input.name}, 'paddle.logical_not',
outputs=[node.name]) inputs={'x': val_input.name},
outputs=[node.name])
@print_mapping_info @print_mapping_info
def AveragePool(self, node): def AveragePool(self, node):
...@@ -1162,9 +1204,9 @@ class OpSet9(): ...@@ -1162,9 +1204,9 @@ class OpSet9():
"name": string(node.name) "name": string(node.name)
} }
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
paddle_op, paddle_op,
inputs={'x': val_x if isinstance(val_x, str) else val_x.name}, inputs={'x': val_x if isinstance(val_x, str) else val_x.name},
outputs=[node.name], outputs=[node.name],
**layer_attrs) **layer_attrs)
@print_mapping_info @print_mapping_info
...@@ -1179,9 +1221,9 @@ class OpSet9(): ...@@ -1179,9 +1221,9 @@ class OpSet9():
assert 'Unspported situation happened, please create issue on https://github.com/PaddlePaddle/X2Paddle/issues.' assert 'Unspported situation happened, please create issue on https://github.com/PaddlePaddle/X2Paddle/issues.'
axis = node.get_attr('axis') axis = node.get_attr('axis')
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
'paddle.concat', 'paddle.concat',
inputs={"x": inputs_list}, inputs={"x": inputs_list},
outputs=[node.name], outputs=[node.name],
axis=axis) axis=axis)
@print_mapping_info @print_mapping_info
...@@ -1199,8 +1241,8 @@ class OpSet9(): ...@@ -1199,8 +1241,8 @@ class OpSet9():
for s in output_shape[axis:]: for s in output_shape[axis:]:
shape_list[1] *= s shape_list[1] *= s
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
'paddle.reshape', 'paddle.reshape',
inputs={"x": val_x.name}, inputs={"x": val_x.name},
outputs=[node.name], outputs=[node.name],
shape=shape_list) shape=shape_list)
...@@ -1215,8 +1257,7 @@ class OpSet9(): ...@@ -1215,8 +1257,7 @@ class OpSet9():
trans_a = bool(node.get_attr('transA', 0)) # optional trans_a = bool(node.get_attr('transA', 0)) # optional
trans_b = bool(node.get_attr('transB', 0)) # optional trans_b = bool(node.get_attr('transB', 0)) # optional
val_mm = node.name + '_mm' val_mm = node.name + '_mm'
matmul_inputs = {"x": val_a.name, matmul_inputs = {"x": val_a.name, "y": val_b.name}
"y": val_b.name}
attr_matmul = { attr_matmul = {
"transpose_x": trans_a, "transpose_x": trans_a,
"transpose_y": trans_b, "transpose_y": trans_b,
...@@ -1227,19 +1268,13 @@ class OpSet9(): ...@@ -1227,19 +1268,13 @@ class OpSet9():
outputs=[val_mm], outputs=[val_mm],
**attr_matmul) **attr_matmul)
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
"paddle.scale", "paddle.scale", inputs={"x": val_mm}, outputs=[val_mm], scale=alpha)
inputs={"x": val_mm},
outputs=[val_mm],
scale=alpha)
if beta != 0: if beta != 0:
if beta == 1.: if beta == 1.:
add_inputs = {"x": val_mm, add_inputs = {"x": val_mm, "y": val_c.name}
"y": val_c.name}
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
"paddle.add", "paddle.add", inputs=add_inputs, outputs=[node.name])
inputs=add_inputs,
outputs=[node.name])
else: else:
var_beta = node.name + '_beta' var_beta = node.name + '_beta'
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
...@@ -1249,9 +1284,7 @@ class OpSet9(): ...@@ -1249,9 +1284,7 @@ class OpSet9():
scale=beta) scale=beta)
add_inputs = {"x": val_mm, "y": var_beta} add_inputs = {"x": val_mm, "y": var_beta}
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
"paddle.add", "paddle.add", inputs=add_inputs, outputs=[node.name])
inputs=add_inputs,
outputs=[node.name])
@print_mapping_info @print_mapping_info
def Sum(self, node): def Sum(self, node):
...@@ -1262,9 +1295,8 @@ class OpSet9(): ...@@ -1262,9 +1295,8 @@ class OpSet9():
"y": self.graph.get_input_node( "y": self.graph.get_input_node(
node, idx=1, copy=True).name, node, idx=1, copy=True).name,
} }
self.paddle_graph.add_layer("paddle.add", self.paddle_graph.add_layer(
inputs=inputs_dict, "paddle.add", inputs=inputs_dict, outputs=[node.name])
outputs=[node.name])
for idx, ipt in enumerate(val_inps[2:]): for idx, ipt in enumerate(val_inps[2:]):
y = self.graph.get_input_node(node, idx=idx, copy=True) y = self.graph.get_input_node(node, idx=idx, copy=True)
...@@ -1273,9 +1305,7 @@ class OpSet9(): ...@@ -1273,9 +1305,7 @@ class OpSet9():
"y": y.name, "y": y.name,
} }
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
"paddle.add", "paddle.add", inputs=inputs_dict, outputs=[node.name])
inputs=inputs_dict,
outputs=[node.name])
@print_mapping_info @print_mapping_info
def MatMul(self, node): def MatMul(self, node):
...@@ -1283,8 +1313,7 @@ class OpSet9(): ...@@ -1283,8 +1313,7 @@ class OpSet9():
val_y = self.graph.get_input_node(node, idx=1, copy=True) val_y = self.graph.get_input_node(node, idx=1, copy=True)
x_shape = val_x.out_shapes[0] x_shape = val_x.out_shapes[0]
y_shape = val_y.out_shapes[0] y_shape = val_y.out_shapes[0]
inputs_dict = {"x": val_x.name, inputs_dict = {"x": val_x.name, "y": val_y.name}
"y": val_y.name}
if y_shape[0] == 1 and x_shape[-1] != 1 and x_shape[0] != 1: if y_shape[0] == 1 and x_shape[-1] != 1 and x_shape[0] != 1:
y_squeeze = val_y.name + '_squeeze' y_squeeze = val_y.name + '_squeeze'
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
...@@ -1294,15 +1323,11 @@ class OpSet9(): ...@@ -1294,15 +1323,11 @@ class OpSet9():
axis=[0]) axis=[0])
inputs_dict['y'] = y_squeeze inputs_dict['y'] = y_squeeze
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
"paddle.matmul", "paddle.matmul", inputs=inputs_dict, outputs=[node.name])
inputs=inputs_dict,
outputs=[node.name])
else: else:
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
"paddle.matmul", "paddle.matmul", inputs=inputs_dict, outputs=[node.name])
inputs=inputs_dict,
outputs=[node.name])
@print_mapping_info @print_mapping_info
def BatchNormalization(self, node): def BatchNormalization(self, node):
val_x = self.graph.get_input_node(node, idx=0, copy=True) val_x = self.graph.get_input_node(node, idx=0, copy=True)
...@@ -1321,15 +1346,17 @@ class OpSet9(): ...@@ -1321,15 +1346,17 @@ class OpSet9():
"epsilon": epsilon, "epsilon": epsilon,
} }
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
"paddle.nn.functional.batch_norm", "paddle.nn.functional.batch_norm",
inputs={"x": val_x.name, inputs={
"weight": val_scale.name, "x": val_x.name,
"bias": val_b.name, "weight": val_scale.name,
"running_mean": val_mean.name, "bias": val_b.name,
"running_var": val_var.name}, "running_mean": val_mean.name,
outputs=[node.name], "running_var": val_var.name
},
outputs=[node.name],
**layer_attrs) **layer_attrs)
@print_mapping_info @print_mapping_info
def Transpose(self, node): def Transpose(self, node):
val_x = self.graph.get_input_node(node, idx=0, copy=True) val_x = self.graph.get_input_node(node, idx=0, copy=True)
...@@ -1338,9 +1365,9 @@ class OpSet9(): ...@@ -1338,9 +1365,9 @@ class OpSet9():
perm_default.reverse() perm_default.reverse()
perm = node.get_attr('perm', perm_default) perm = node.get_attr('perm', perm_default)
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
"paddle.transpose", "paddle.transpose",
inputs={"x": val_x.name}, inputs={"x": val_x.name},
outputs=[node.name], outputs=[node.name],
perm=perm) perm=perm)
@print_mapping_info @print_mapping_info
...@@ -1355,23 +1382,23 @@ class OpSet9(): ...@@ -1355,23 +1382,23 @@ class OpSet9():
if mode == "element": if mode == "element":
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
"paddle.static.nn.prelu", "paddle.static.nn.prelu",
inputs={"x": val_x.name, inputs={"x": val_x.name,
"param_attr": val_slope.name}, "param_attr": val_slope.name},
outputs=[node.name], outputs=[node.name],
mode="element") mode="element")
else: else:
if mode == 'channel': if mode == 'channel':
if len(shape_slope) > 1: if len(shape_slope) > 1:
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
"paddle.reshape", "paddle.reshape",
inputs={"x": val_slope.name}, inputs={"x": val_slope.name},
outputs=[val_slope.name], outputs=[val_slope.name],
shape=[shape_slope[0]]) shape=[shape_slope[0]])
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
"paddle.nn.functional.prelu", "paddle.nn.functional.prelu",
inputs={"x": val_x.name, inputs={"x": val_x.name,
"weight": val_slope.name}, "weight": val_slope.name},
outputs=[node.name]) outputs=[node.name])
@print_mapping_info @print_mapping_info
...@@ -1386,9 +1413,9 @@ class OpSet9(): ...@@ -1386,9 +1413,9 @@ class OpSet9():
dtype=string(val_x.dtype)) dtype=string(val_x.dtype))
else: else:
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
"paddle.squeeze", "paddle.squeeze",
inputs={"x": val_x.name}, inputs={"x": val_x.name},
outputs=[node.name], outputs=[node.name],
axis=axes) axis=axes)
@print_mapping_info @print_mapping_info
...@@ -1409,8 +1436,7 @@ class OpSet9(): ...@@ -1409,8 +1436,7 @@ class OpSet9():
"paddle.greater_than", "paddle.greater_than",
inputs={'x': val_x.name, inputs={'x': val_x.name,
'y': val_y.name}, 'y': val_y.name},
outputs=[node.name], outputs=[node.name])
param_attr=None)
@print_mapping_info @print_mapping_info
def Where(self, node): def Where(self, node):
...@@ -1460,8 +1486,8 @@ class OpSet9(): ...@@ -1460,8 +1486,8 @@ class OpSet9():
val_x_dim = len(val_x.out_shapes[0]) val_x_dim = len(val_x.out_shapes[0])
if val_x_dim == 1: if val_x_dim == 1:
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
"paddle.nonzero", "paddle.nonzero",
inputs={"x": val_x.name}, inputs={"x": val_x.name},
outputs=[val_x.name]) outputs=[val_x.name])
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
"paddle.transpose", "paddle.transpose",
...@@ -1470,28 +1496,24 @@ class OpSet9(): ...@@ -1470,28 +1496,24 @@ class OpSet9():
perm=[1, 0]) perm=[1, 0])
if val_x_dim > 1: if val_x_dim > 1:
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
"paddle.nonzero", "paddle.nonzero",
inputs={"x": val_x.name}, inputs={"x": val_x.name},
outputs=[val_x.name]) outputs=[val_x.name])
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
"paddle.split", "paddle.split",
inputs={"x": val_x.name}, inputs={"x": val_x.name},
outputs=[val_x.name], outputs=[val_x.name],
num_or_sections=1, num_or_sections=1,
axis=val_x_dim) axis=val_x_dim)
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
"paddle.concat", "paddle.concat", inputs={"x": val_x.name}, outputs=[node.name])
inputs={"x": val_x.name},
outputs=[node.name])
@print_mapping_info @print_mapping_info
def Identity(self, node): def Identity(self, node):
val_x = self.graph.get_input_node(node, idx=0, copy=True) val_x = self.graph.get_input_node(node, idx=0, copy=True)
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
"paddle.assign", "paddle.assign", inputs={"x": val_x.name}, outputs=[node.name])
inputs={"x": val_x.name},
outputs=[node.name])
@print_mapping_info @print_mapping_info
def Tile(self, node): def Tile(self, node):
val_x = self.graph.get_input_node(node, idx=0, copy=True) val_x = self.graph.get_input_node(node, idx=0, copy=True)
...@@ -1504,22 +1526,25 @@ class OpSet9(): ...@@ -1504,22 +1526,25 @@ class OpSet9():
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
"paddle.cast", "paddle.cast",
inputs={"x": repeats}, inputs={"x": repeats},
outputs=["{}.tmp".format(repeats)], outputs=["{}_tmp".format(repeats)],
dtype=string("int32")) dtype=string("int32"))
repeats = "{}.tmp".format(repeats) repeats = "{}_tmp".format(repeats)
elif isinstance(repeats, int): elif isinstance(repeats, int):
repeats = [repeats] repeats = [repeats]
elif type(repeats) is np.ndarray:
repeats = repeats.tolist()
attr = { attr = {
'expand_times': repeats, 'expand_times': repeats,
"name": string(node.name), "name": string(node.name),
} }
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
"paddle.tile", "paddle.tile",
inputs={"x": val_x.name}, inputs={"x": val_x.name},
outputs=[node.name], outputs=[node.name],
repeat_times=repeats) repeat_times=repeats)
@print_mapping_info @print_mapping_info
def MaxPool(self, node): def MaxPool(self, node):
...@@ -1546,7 +1571,7 @@ class OpSet9(): ...@@ -1546,7 +1571,7 @@ class OpSet9():
pad_w = _get_same_padding(input_shape[3], kernel_shape[1], pad_w = _get_same_padding(input_shape[3], kernel_shape[1],
strides[1]) strides[1])
paddings = pad_h + pad_w paddings = pad_h + pad_w
layer_attrs = { layer_attrs = {
"kernel_size": kernel_shape, "kernel_size": kernel_shape,
"stride": strides, "stride": strides,
...@@ -1554,9 +1579,9 @@ class OpSet9(): ...@@ -1554,9 +1579,9 @@ class OpSet9():
"ceil_mode": ceil_mode, "ceil_mode": ceil_mode,
} }
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
paddle_op, paddle_op,
inputs={'x': val_x if isinstance(val_x, str) else val_x.name}, inputs={'x': val_x if isinstance(val_x, str) else val_x.name},
outputs=[node.name], outputs=[node.name],
**layer_attrs) **layer_attrs)
@print_mapping_info @print_mapping_info
...@@ -1573,11 +1598,11 @@ class OpSet9(): ...@@ -1573,11 +1598,11 @@ class OpSet9():
assert 1 <= poolnd <= 3, 'only adaptive_max_pool1d, adaptive_max_pool2d and adaptive_max_pool3d are supported' assert 1 <= poolnd <= 3, 'only adaptive_max_pool1d, adaptive_max_pool2d and adaptive_max_pool3d are supported'
output_shape = node.out_shapes[0] output_shape = node.out_shapes[0]
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
paddle_op, paddle_op,
inputs={'x': val_x.name}, inputs={'x': val_x.name},
outputs=[node.name], outputs=[node.name],
output_size=output_shape[2:]) output_size=output_shape[2:])
@print_mapping_info @print_mapping_info
def GlobalAveragePool(self, node): def GlobalAveragePool(self, node):
val_x = self.graph.get_input_node(node, idx=0, copy=True) val_x = self.graph.get_input_node(node, idx=0, copy=True)
...@@ -1592,9 +1617,9 @@ class OpSet9(): ...@@ -1592,9 +1617,9 @@ class OpSet9():
assert 1 <= poolnd <= 3, 'only Pool1D, Pool2D and Pool3D are supported' assert 1 <= poolnd <= 3, 'only Pool1D, Pool2D and Pool3D are supported'
output_shape = node.out_shapes[0] output_shape = node.out_shapes[0]
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
paddle_op, paddle_op,
inputs={'x': val_x.name}, inputs={'x': val_x.name},
outputs=[node.name], outputs=[node.name],
output_size=output_shape[2:]) output_size=output_shape[2:])
@print_mapping_info @print_mapping_info
...@@ -1640,20 +1665,18 @@ class OpSet9(): ...@@ -1640,20 +1665,18 @@ class OpSet9():
} }
if has_bias: if has_bias:
layer_inputs["bias"] = val_b.name layer_inputs["bias"] = val_b.name
if reduce(lambda x,y:x*y, input_shape) in [1, -1] and 1 not in input_shape: if reduce(lambda x, y: x * y,
input_shape) in [1, -1] and 1 not in input_shape:
input_shape[1] = num_in_channels * num_groups input_shape[1] = num_in_channels * num_groups
input_shape[0] = 0 input_shape[0] = 0
input_shape[2] = 0 input_shape[2] = 0
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
"paddle.reshape", "paddle.reshape",
inputs={"x": layer_inputs["x"]}, inputs={"x": layer_inputs["x"]},
outputs=[layer_inputs["x"]], outputs=[layer_inputs["x"]],
shape=input_shape) shape=input_shape)
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
paddle_op, paddle_op, inputs=layer_inputs, outputs=[node.name], **layer_attrs)
inputs=layer_inputs,
outputs=[node.name],
**layer_attrs)
@print_mapping_info @print_mapping_info
def ConvTranspose(self, node): def ConvTranspose(self, node):
...@@ -1688,14 +1711,14 @@ class OpSet9(): ...@@ -1688,14 +1711,14 @@ class OpSet9():
output_size[1] = (val_x.out_shapes[0][3] - 1 output_size[1] = (val_x.out_shapes[0][3] - 1
) * strides[1] - 2 * paddings[1] + dilations[1] * ( ) * strides[1] - 2 * paddings[1] + dilations[1] * (
kernel_shape[1] - 1) + 1 + out_padding[1] kernel_shape[1] - 1) + 1 + out_padding[1]
layer_inputs = {'x': val_x.name, layer_inputs = {'x': val_x.name, "weight": val_w.name}
"weight": val_w.name}
layer_attrs = { layer_attrs = {
"stride": strides, "stride": strides,
"dilation": dilations, "dilation": dilations,
"padding": paddings, "padding": paddings,
"groups": num_groups, "groups": num_groups,
"output_size": node.out_shapes[0][2:]} "output_size": node.out_shapes[0][2:]
}
if val_b is not None: if val_b is not None:
layer_inputs["bias"] = val_b.name layer_inputs["bias"] = val_b.name
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
...@@ -1703,54 +1726,47 @@ class OpSet9(): ...@@ -1703,54 +1726,47 @@ class OpSet9():
inputs=layer_inputs, inputs=layer_inputs,
outputs=[node.name], outputs=[node.name],
**layer_attrs) **layer_attrs)
@print_mapping_info @print_mapping_info
def ArgMax(self, node): def ArgMax(self, node):
val_x = self.graph.get_input_node(node, idx=0, copy=True) val_x = self.graph.get_input_node(node, idx=0, copy=True)
axis = node.get_attr('axis') axis = node.get_attr('axis')
keepdims = False if node.get_attr('keepdims') == 0 else True keepdims = False if node.get_attr('keepdims') == 0 else True
layer_attrs = {'axis': axis, layer_attrs = {'axis': axis, 'keepdim': keepdims}
'keepdim': keepdims}
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
'paddle.argmax', 'paddle.argmax',
inputs={"x": val_x.name}, inputs={"x": val_x.name},
outputs=[node.name], outputs=[node.name],
**layer_attrs) **layer_attrs)
@print_mapping_info @print_mapping_info
def Size(self, node): def Size(self, node):
val_x = self.graph.get_input_node(node, idx=0, copy=True) val_x = self.graph.get_input_node(node, idx=0, copy=True)
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
"paddle.shape", "paddle.shape", inputs={"input": val_x.name}, outputs=[node.name])
inputs={"input": val_x.name},
outputs=[node.name])
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
'paddle.cast', 'paddle.cast',
inputs={"x": node.name}, inputs={"x": node.name},
outputs=[node.name], outputs=[node.name],
dtype=string('int64')) dtype=string('int64'))
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
"paddle.prod", "paddle.prod", inputs={"x": node.name}, outputs=[node.name])
inputs={"x": node.name},
outputs=[node.name])
@print_mapping_info @print_mapping_info
def Sign(self, node): def Sign(self, node):
val_x = self.graph.get_input_node(node, idx=0, copy=True) val_x = self.graph.get_input_node(node, idx=0, copy=True)
if node.dtype not in ["float16", "float32", "float64"]: if node.dtype not in ["float16", "float32", "float64"]:
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
"paddle.cast", "paddle.cast",
inputs={"x": val_x.name}, inputs={"x": val_x.name},
outputs=[val_x.name], outputs=[val_x.name],
dtype=string("float32")) dtype=string("float32"))
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
"paddle.sign", "paddle.sign", inputs={"x": val_x.name}, outputs=[node.name])
inputs={"x": val_x.name},
outputs=[node.name])
if node.dtype not in ["float16", "float32", "float64"]: if node.dtype not in ["float16", "float32", "float64"]:
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
"paddle.cast", "paddle.cast",
inputs={"x": node.name}, inputs={"x": node.name},
outputs=[node.name], outputs=[node.name],
dtype=string(node.dtype)) dtype=string(node.dtype))
...@@ -1761,10 +1777,12 @@ class OpSet9(): ...@@ -1761,10 +1777,12 @@ class OpSet9():
values = self.graph.get_input_node(node, idx=2, copy=True) values = self.graph.get_input_node(node, idx=2, copy=True)
axis = node.get_attr('axis', -1) axis = node.get_attr('axis', -1)
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
"custom_layer:one_hot", "custom_layer:one_hot",
inputs={"indices": indices.name, inputs={
"depth": depth.name, "indices": indices.name,
"values": values.name}, "depth": depth.name,
"values": values.name
},
outputs=[node.name], outputs=[node.name],
axis=axis) axis=axis)
...@@ -1772,9 +1790,7 @@ class OpSet9(): ...@@ -1772,9 +1790,7 @@ class OpSet9():
def Reciprocal(self, node): def Reciprocal(self, node):
val_x = self.graph.get_input_node(node, idx=0, copy=True) val_x = self.graph.get_input_node(node, idx=0, copy=True)
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
"paddle.reciprocal", "paddle.reciprocal", inputs={"x": val_x.name}, outputs=[node.name])
inputs={"x": val_x.name},
outputs=[node.name])
@print_mapping_info @print_mapping_info
def TopK(self, node): def TopK(self, node):
...@@ -1782,15 +1798,20 @@ class OpSet9(): ...@@ -1782,15 +1798,20 @@ class OpSet9():
val_k = self.graph.get_input_node(node, idx=1, copy=True) val_k = self.graph.get_input_node(node, idx=1, copy=True)
layer_attrs = dict() layer_attrs = dict()
layer_attrs["axis"] = node.get_attr('axis', -1) layer_attrs["axis"] = node.get_attr('axis', -1)
layer_attrs["largest"] = True if node.get_attr('largest', 1) == 1 else False layer_attrs["largest"] = True if node.get_attr('largest',
layer_attrs["sorted"] = True if node.get_attr('sorted', 1) == 1 else False 1) == 1 else False
layer_attrs["sorted"] = True if node.get_attr('sorted',
1) == 1 else False
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
"paddle.topk", "paddle.topk",
inputs={"x": val_x.name, inputs={"x": val_x.name,
"k": val_k.name}, "k": val_k.name},
outputs=["{}_p{}".format(node.layer_name, 0), "{}_p{}".format(node.layer_name, 1)], outputs=[
"{}_p{}".format(node.layer_name, 0),
"{}_p{}".format(node.layer_name, 1)
],
**layer_attrs) **layer_attrs)
@print_mapping_info @print_mapping_info
def LRN(self, node): def LRN(self, node):
val_x = self.graph.get_input_node(node, idx=0, copy=True) val_x = self.graph.get_input_node(node, idx=0, copy=True)
...@@ -1799,16 +1820,11 @@ class OpSet9(): ...@@ -1799,16 +1820,11 @@ class OpSet9():
beta = node.get_attr('beta', 0.75) beta = node.get_attr('beta', 0.75)
bias = node.get_attr('bias', 1.0) bias = node.get_attr('bias', 1.0)
size = node.get_attr('size') size = node.get_attr('size')
layer_attrs = { layer_attrs = {'size': size, 'alpha': alpha, 'beta': beta, 'k': bias}
'size': size,
'alpha': alpha,
'beta': beta,
'k': bias
}
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
"custom_layer:local_response_norm", "custom_layer:local_response_norm",
inputs={"x": val_x.name}, inputs={"x": val_x.name},
outputs=[node.name], outputs=[node.name],
**layer_attrs) **layer_attrs)
@print_mapping_info @print_mapping_info
...@@ -1823,37 +1839,30 @@ class OpSet9(): ...@@ -1823,37 +1839,30 @@ class OpSet9():
'paddle.reshape', 'paddle.reshape',
inputs={"x": val_x.name}, inputs={"x": val_x.name},
outputs=[node.name], outputs=[node.name],
shape=[b, blocksize, blocksize, c // (blocksize**2), h, w] shape=[b, blocksize, blocksize, c // (blocksize**2), h, w])
)
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
'paddle.transpose', 'paddle.transpose',
inputs={"x": node.name}, inputs={"x": node.name},
outputs=[node.name], outputs=[node.name],
perm=[0, 3, 4, 1, 5, 2] perm=[0, 3, 4, 1, 5, 2])
)
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
'paddle.reshape', 'paddle.reshape',
inputs={"x": node.name}, inputs={"x": node.name},
outputs=[node.name], outputs=[node.name],
shape=[b, c // (blocksize**2), h * blocksize, w * blocksize] shape=[b, c // (blocksize**2), h * blocksize, w * blocksize])
)
else: else:
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
'paddle.reshape', 'paddle.reshape',
inputs={"x": val_x.name}, inputs={"x": val_x.name},
outputs=[node.name], outputs=[node.name],
shape=[b, c // (blocksize ** 2), blocksize, blocksize, h, w] shape=[b, c // (blocksize**2), blocksize, blocksize, h, w])
)
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
'paddle.transpose', 'paddle.transpose',
inputs={"x": node.name}, inputs={"x": node.name},
outputs=[node.name], outputs=[node.name],
perm=[0, 1, 4, 2, 5, 3] perm=[0, 1, 4, 2, 5, 3])
)
self.paddle_graph.add_layer( self.paddle_graph.add_layer(
'paddle.reshape', 'paddle.reshape',
inputs={"x": node.name}, inputs={"x": node.name},
outputs=[node.name], outputs=[node.name],
shape=[b, c // (blocksize ** 2), h * blocksize, w * blocksize] shape=[b, c // (blocksize**2), h * blocksize, w * blocksize])
)
\ No newline at end of file
Markdown is supported
0% 或 .
You are about to add 0 people to the discussion. Proceed with caution.
先完成此消息的编辑！
想要评论请 注册 或