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 c05a67a4 编写于 作者: J Jason 提交者： GitHub
浏览文件
操作

Merge pull request #435 from SunAhong1993/paddle-2.0 

add pytorch2paddle
上级 07c16e50 b2b5d719
展开全部 隐藏空白更改
内联 并排
Showing with 2777 addition and 2341 deletion
README.md
浏览文件 @ c05a67a4
......@@ -16,6 +16,7 @@ paddlepaddle >= 1.8.0
tensorflow ： tensorflow == 1.14.0
caffe ： 无
onnx ： onnx >= 1.6.0
pytorch：torch >=1.5.0 (script方式暂不支持1.7.0)
## 安装
### 安装方式一（推荐）
......@@ -45,11 +46,10 @@ x2paddle --framework=caffe --prototxt=deploy.prototxt --weight=deploy.caffemodel
x2paddle --framework=onnx --model=onnx_model.onnx --save_dir=pd_model
```
### Paddle2ONNX
```
# 注意：paddle_infer_model_dir下需包含__model__和__params__两个文件
x2paddle --framework=paddle2onnx --model=paddle_infer_model_dir --save_dir=onnx_model
```
### PyTorch
> PyTorch不支持命令行使用方式，详见[PyTorch2Paddle](pytorch2paddle.md)
### 参数选项
| 参数 | |
|----------|--------------|
......
pytorch2paddle.md 0 → 100644
浏览文件 @ c05a67a4
# PyTorch2Paddle
PyTorch2Paddle支持trace和script两种方式的转换，均是PyTorch动态图到Paddle动态图的转换，转换后的Paddle动态图运用动转静可转换为静态图模型。trace方式生成的代码可读性较强，较为接近原版PyTorch代码的组织结构；script方式不需要知道输入数据的类型和大小即可转换，使用上较为方便，但目前PyTorch支持的script代码方式有所限制，所以支持转换的代码也有所限制。用户可根据自身需求，选择转换方式。
## 环境依赖
python == 2.7 | python >= 3.5
paddlepaddle >= 1.8.0
pytorch：torch >=1.5.0 (script方式暂不支持1.7.0)
**使用trace方式需安装以下依赖**
pandas
treelib
## 使用方式
```
from x2paddle.convert import pytorch2paddle
pytorch2paddle(module=torch_module,
save_dir="./pd_model",
jit_type="trace",
input_examples=[torch_input])
# module (torch.nn.Module): PyTorch的Module。
# save_dir (str): 转换后模型的保存路径。
# jit_type (str): 转换方式。默认为"trace"。
# input_examples (list[torch.tensor]): torch.nn.Module的输入示例，list的长度必须与输入的长度一致。默认为None。
```
**注意：** 当jit_type为"trace"时，input_examples不可为None，转换后自动进行动转静；
当jit_type为"script"时"，input_examples不为None时，才可以进行动转静。
## 使用示例
```
import torch
import numpy as np
from torchvision.models import AlexNet
from torchvision.models.utils import load_state_dict_from_url
# 构建输入
input_data = np.random.rand(1, 3, 224, 224).astype("float32")
# 获取PyTorch Module
torch_module = AlexNet()
torch_state_dict = load_state_dict_from_url('https://download.pytorch.org/models/alexnet-owt-4df8aa71.pth')
torch_module.load_state_dict(torch_state_dict)
# 设置为eval模式
torch_module.eval()
# 进行转换
from x2paddle.convert import pytorch2paddle
pytorch2paddle(torch_model,
save_dir="pd_model_trace",
jit_type="trace",
input_examples=[torch.tensor(input_data)])
```
\ No newline at end of file
x2paddle/convert.py
浏览文件 @ c05a67a4
......@@ -209,7 +209,7 @@ def onnx2paddle(model_path, save_dir, paddle_type, params_merge=False):
mapper.save_inference_model(save_dir, params_merge)
def pytorch2paddle(model_path, save_dir, jit_type, input_files):
def pytorch2paddle(module, save_dir, jit_type="trace", input_examples=None):
# check pytorch installation and version
try:
import torch
......@@ -225,21 +225,22 @@ def pytorch2paddle(model_path, save_dir, jit_type, input_files):
)
return
print("Now translating model from pytorch to paddle.")
from x2paddle.decoder.pytorch_decoder import ScriptDecoder, TraceDecoder
from x2paddle.op_mapper.pytorch2paddle import pytorch_op_mapper
from x2paddle.op_mapper.dygraph.pytorch2paddle.pytorch_op_mapper import PyTorchOpMapper
if jit_type == "trace":
model = TraceDecoder(model_path, input_files)
model = TraceDecoder(module, input_examples)
else:
model = ScriptDecoder(model_path)
mapper = pytorch_op_mapper.PyTorchOpMapper(model)
mapper.graph.build()
model = ScriptDecoder(module, input_examples)
mapper = PyTorchOpMapper(model)
mapper.paddle_graph.build()
print("Model optimizing ...")
from x2paddle.optimizer.pytorch_optimizer.optimizer import GraphOptimizer
graph_opt = GraphOptimizer()
graph_opt.optimize(mapper.graph)
from x2paddle.optimizer.optimizer import GraphOptimizer
graph_opt = GraphOptimizer(source_frame="pytorch", paddle_type="dygraph", jit_type=jit_type)
graph_opt.optimize(mapper.paddle_graph)
print("Model optimized.")
mapper.graph.gen_model(save_dir, jit_type, input_files)
mapper.paddle_graph.gen_model(save_dir, jit_type=jit_type)
def paddle2onnx(model_path, save_dir, opset_version=10):
......@@ -323,10 +324,6 @@ def main():
if args.params_merge:
params_merge = True
onnx2paddle(args.model, args.save_dir, args.paddle_type, params_merge)
elif args.framework == "pytorch":
assert args.model is not None, "--model should be defined while translating pytorch model"
pytorch2paddle(args.model, args.save_dir, args.jit_type, args.input_files)
elif args.framework == "paddle2onnx":
assert args.model is not None, "--model should be defined while translating paddle model to onnx"
paddle2onnx(args.model, args.save_dir, opset_version=args.onnx_opset)
......
x2paddle/core/program.py
浏览文件 @ c05a67a4
......@@ -281,13 +281,14 @@ class PaddleGraph(object):
else:
self.gen_dygraph_code(save_dir)
self.dump_dygraph_parameter(save_dir)
input_shapes = list()
input_types = list()
for input_name in self.inputs:
input_shapes.append(self.inputs_info[input_name][0])
input_types.append(self.inputs_info[input_name][1])
# 如果input_files非空，则导出推理模型；其值类似[[None, 3, 224, 224]]
self.dygraph2static(save_dir, input_shapes, input_types)
# 动转静
if len(self.inputs_info) > 0:
input_shapes = list()
input_types = list()
for input_name in self.inputs:
input_shapes.append(self.inputs_info[input_name][0])
input_types.append(self.inputs_info[input_name][1])
self.dygraph2static(save_dir, input_shapes, input_types)
def gen_static_code(self, code_dir):
def write_code(f, code_list, indent=0):
......@@ -424,9 +425,7 @@ class PaddleGraph(object):
if self.edges_out.get(layer_id, 0) == 0:
for i, output_name in enumerate(layer.outputs):
if ("paddle.nn" in layer.kernel and "functional" not in layer.kernel) or \
(layer.kernel == "paddle.to_tensor" and layer.attrs["data"].startswith("params["))or \
"paddle.fluid.dygraph" in layer.kernel:
if ("paddle.nn" in layer.kernel and "functional" not in layer.kernel):
if i == 0:
continue
if output_name not in self.outputs:
......@@ -512,6 +511,8 @@ class PaddleGraph(object):
return_code = "return {}".format(", ".join(self.outputs))
self.forward_func.extend(gen_codes([return_code], indent=2))
for code_line in self.forward_func:
if "assert [1, 1] == 1 or [1, 1] == [1, 1], 'The [1, 1] must be [1, [1, 1]]!'" in code_line:
continue
f.write(code_line)
for code_line in self.run_func:
f.write(code_line)
......@@ -525,7 +526,6 @@ class PaddleGraph(object):
for layer_id, layer in self.layers.items():
if ("paddle.nn" in layer.kernel and "functional" not in layer.kernel
) or layer.kernel == "paddle.to_tensor" or \
"paddle.fluid.dygraph" in layer.kernel or \
layer.kernel.startswith("custom_layer"):
line = "{}".format(
layer.outputs[0]
......@@ -566,7 +566,7 @@ class PaddleGraph(object):
self.forward_func.extend(gen_codes([line], indent=indent))
elif "prim" in layer.kernel:
func_name = layer.kernel.replace(".", "_")
from x2paddle.op_mapper.dygraph import prim2code
from x2paddle.op_mapper.dygraph.pytorch2paddle import prim2code
if hasattr(prim2code, func_name):
func = getattr(prim2code, func_name)
func(
......@@ -594,7 +594,7 @@ class PaddleGraph(object):
line = line.strip(", ")
line += ")"
if layer.kernel == "self.create_parameter":
self.init_func.extend(gen_codes(["self." + line], indent=indent))
self.init_func.extend(gen_codes(["self." + line], indent=2))
self.forward_func.extend(gen_codes(["{} = self.{}".format(layer.outputs[0],
layer.outputs[0])], indent=indent))
else:
......@@ -614,7 +614,6 @@ class PaddleGraph(object):
from paddle.fluid.dygraph.jit import declarative
sepc_list = list()
for i, name in enumerate(self.inputs):
input_shapes[i][0] = -1
sepc_list.append(
paddle.static.InputSpec(
shape=input_shapes[i], name=name, dtype=input_types[i]))
......@@ -631,4 +630,11 @@ class PaddleGraph(object):
model.set_dict(restore)
model.eval()
static_model = paddle.jit.to_static(model, input_spec=sepc_list)
paddle.jit.save(static_model, osp.join(save_dir, "inference_model/model"))
\ No newline at end of file
try:
paddle.jit.save(static_model, osp.join(save_dir, "inference_model/model"))
except ValueError as e:
if str(e) == "'target_vars' should be a list of Variable.":
print("[DyGraph2StaticGraph Error] Can not convert the dygraph to static! The output of PyTorch mustbe Variable or a list of Variable.")
else:
print(e)
exit(0)
x2paddle/decoder/pytorch_decoder.py
浏览文件 @ c05a67a4
......@@ -41,9 +41,10 @@ class ScriptDecoder(Decoder):
script_path (str): ScriptModule保存路径。
model_path (str): PyTorchModule保存路径。
"""
def __init__(self, script_path=None):
self.script = torch.jit.load(script_path)
def __init__(self, module, input_examples=None):
self.script = torch.jit.script(module)
self.graph = self._optimize_graph(self.script.inlined_graph)
self.input_examples = input_examples
class TraceDecoder(Decoder):
""" PyTorchModule后使用trace方式转换为ScriptModule。
......@@ -53,14 +54,16 @@ class TraceDecoder(Decoder):
input_files (list): 输入网络的numpy，每个numpy保存成.npy文件,
文件路径存储在input_files中。
"""
def __init__(self, model_path, input_files=list()):
# TODO(syf): 传入pytorch的Module(即import)，否则出错
model = torch.load(model_path)
model.eval()
input_list = list()
for npy_file in input_files:
input_list.append(torch.tensor(np.load(npy_file)))
self.script = torch.jit.trace(model, input_list, strict=False)
def __init__(self, module, input_examples):
try:
self.script = torch.jit.trace(module, input_examples)
except RuntimeError as e:
if "strict" in str(e):
self.script = torch.jit.trace(module, input_examples, strict=False)
else:
print(e)
exit(0)
self.graph = self._optimize_graph(self.script.inlined_graph)
# print(self.graph)
# print(getattr(getattr(self.script.decoder.block, "5").layer, "2"))
self.input_examples = input_examples
x2paddle/op_mapper/dygraph/onnx2paddle/opset9/opset.py
浏览文件 @ c05a67a4
......@@ -1180,7 +1180,7 @@ class OpSet9():
scale=beta)
add_inputs = {"x": val_mm, "y": var_beta}
self.paddle_graph.add_layer(
"paddle.addd",
"paddle.add",
inputs=add_inputs,
outputs=[node.layer_name])
......
x2paddle/op_mapper/dygraph/pytorch2paddle/prim.py
浏览文件 @ c05a67a4
......@@ -25,6 +25,7 @@ def prim_Constant(mapper, graph, node):
参数含义:
%2 (常量类型由赋值类型定义，该示例中为int型): 常量赋值结果输出。
"""
scope_name = mapper.normalize_scope_name(node)
output_name = mapper._get_outputs_name(node)[0]
output = list(node.outputs())[0]
value = output.toIValue()
......@@ -32,7 +33,10 @@ def prim_Constant(mapper, graph, node):
if isinstance(value, str):
value = string(value)
if str(output_type) == "Tensor":
tensor_value = value
value = "{}".format(value)
if "tensor" in value:
mapper.pytorch_params[output_name] = tensor_value.cpu().detach().numpy()
if "inf" in str(value):
t = str(type(value)).split("'")[1]
......@@ -45,7 +49,7 @@ def prim_Constant(mapper, graph, node):
value = int(math.pow(2, 31) - 1)
mapper.attrs[output_name] = value
graph.add_layer(
"prim.constant", inputs={}, outputs=[output_name], value=value)
"prim.constant", inputs={}, outputs=[output_name], scope_name=scope_name, value=value)
return [], [output_name]
......@@ -60,6 +64,7 @@ def prim_data(mapper, graph, node):
【注意】Paddle中无此用法，所以此处翻译成赋值。
"""
scope_name = mapper.normalize_scope_name(node)
output_name = mapper._get_outputs_name(node)[0]
layer_outputs = [output_name]
layer_inputs = {}
......@@ -68,15 +73,53 @@ def prim_data(mapper, graph, node):
# 获取当前节点输出的list
current_outputs = [output_name]
# 处理输入0，即%4336
mapper._check_input(graph, inputs_node[0], inputs_name[0], current_outputs)
mapper._check_input(graph, inputs_node[0], inputs_name[0], current_outputs, scope_name)
layer_inputs["input"] = inputs_name[0]
# 获取当前节点输入的list
current_inputs = list(layer_inputs.values())
graph.add_layer("prim.equal", inputs=layer_inputs, outputs=layer_outputs)
graph.add_layer("prim.equal", inputs=layer_inputs, outputs=layer_outputs, scope_name=scope_name)
return current_inputs, current_outputs
def prim_DictConstruct(mapper, graph, node):
""" 构建dict。
TorchScript示例:
%32 : Dict(str, Tensor) = prim::DictConstruct(%30, %23, %31, %29)
参数含义:
%32 (dict): 组成的字典。
%30 (str): key。
%23 (-): value。
%31 (str): key。
%29 (-): value。
"""
scope_name = mapper.normalize_scope_name(node)
output_name = mapper._get_outputs_name(node)[0]
layer_outputs = [output_name]
layer_inputs = {}
layer_attrs = {}
inputs_name, inputs_node = mapper._get_inputs_name(node)
# 获取当前节点输出的list
current_outputs = [output_name]
# 处理每个输入
for i, input_name in enumerate(inputs_name):
if i%2 == 0:
layer_attrs["key{}".format(int(i/2))] = mapper.attrs[input_name]
else:
layer_inputs["value{}".format(int(i/2))] = input_name
# 获取当前节点输入的list
current_inputs = list(layer_inputs.values())
graph.add_layer("prim.dict_construct",
inputs=layer_inputs,
outputs=layer_outputs,
scope_name=scope_name,
**layer_attrs)
return current_inputs, current_outputs
def prim_GetAttr(mapper, graph, node):
""" 获取attribute信息。
......@@ -86,6 +129,7 @@ def prim_GetAttr(mapper, graph, node):
%7 (Tensor): 输入Tensor。
%27 (Tensor): 输入Tensor。
"""
scope_name = mapper.normalize_scope_name(node)
current_node = node
field_name_list = [node.s('name')]
while True:
......@@ -102,7 +146,7 @@ def prim_GetAttr(mapper, graph, node):
if hasattr(part_script, field_name):
param = getattr(part_script, field_name)
if isinstance(param, torch.Tensor):
param = param.detach().numpy()
param = param.cpu().detach().numpy()
if len(param.shape) == 0:
param = np.reshape(param, 1)
if str(param.dtype) == "uint8":
......@@ -129,14 +173,15 @@ def prim_If(mapper, graph, node):
%107 (bool): if判断条件。
%input.5 (Tensor): if控制流的输出，与%output.4对应。
"""
scope_name = mapper.normalize_scope_name(node)
outputs_name = mapper._get_outputs_name(node)
node_outputs = outputs_name.copy()
current_outputs = outputs_name.copy()
input_node = list(node.inputs())[0].node()
script_input_unique_id = list(node.inputs())[0].unique()
input_node_name = mapper.outputs_info[script_input_unique_id]
mapper._check_input(graph, input_node, input_node_name, current_outputs)
graph.add_layer("prim.if", {'input': input_node_name}, node_outputs)
mapper._check_input(graph, input_node, input_node_name, current_outputs, scope_name)
graph.add_layer("prim.if", inputs={'input': input_node_name}, outputs=node_outputs, scope_name=scope_name)
current_layer = list(graph.layers.values())[-1]
block0 = list(node.blocks())[0]
block0_graph, graph_inputs0 = mapper.traverse(block0, current_layer)
......@@ -163,6 +208,7 @@ def prim_ListConstruct(mapper, graph, node):
%84 (int/其他): list第一个元素信息。
%85 (int/其他): list第二个元素信息。
"""
scope_name = mapper.normalize_scope_name(node)
output_name = mapper._get_outputs_name(node)[0]
layer_outputs = [output_name]
layer_inputs = {}
......@@ -175,7 +221,7 @@ def prim_ListConstruct(mapper, graph, node):
# 获取当前节点输入的list
current_inputs = list(layer_inputs.values())
graph.add_layer("prim.list", inputs=layer_inputs, outputs=layer_outputs)
graph.add_layer("prim.list", inputs=layer_inputs, outputs=layer_outputs, scope_name=scope_name)
return current_inputs, current_outputs
......@@ -189,6 +235,7 @@ def prim_ListUnpack(mapper, graph, node):
%x2.4 (Tensor): 输出，list的第二个元素。
%4354 (list): 列表。
"""
scope_name = mapper.normalize_scope_name(node)
outputs_name = mapper._get_outputs_name(node)
layer_outputs = outputs_name.copy()
layer_inputs = {}
......@@ -196,13 +243,13 @@ def prim_ListUnpack(mapper, graph, node):
# 获取当前节点输出的list
current_outputs = layer_outputs.copy()
# 处理输入0，即%4354
mapper._check_input(graph, inputs_node[0], inputs_name[0], current_outputs)
mapper._check_input(graph, inputs_node[0], inputs_name[0], current_outputs, scope_name)
layer_inputs["input"] = inputs_name[0]
# 获取当前节点输入的list
current_inputs = list(layer_inputs.values())
graph.add_layer(
"prim.list_unpack", inputs=layer_inputs, outputs=layer_outputs)
"prim.list_unpack", inputs=layer_inputs, outputs=layer_outputs, scope_name=scope_name)
mapper.split_len[list(layer_inputs.values())[0]] = len(layer_outputs)
return current_inputs, current_outputs
......@@ -223,6 +270,7 @@ def prim_Loop(mapper, graph, node):
%x.3 (Tensor): 循环中修改的Tensor。
%x (Tensor): loop循环的输出，与%x.5对应。
"""
scope_name = mapper.normalize_scope_name(node)
node_outputs = mapper._get_outputs_name(node)
loop_inputs = {}
block = list(node.blocks())[0]
......@@ -242,7 +290,7 @@ def prim_Loop(mapper, graph, node):
loop_input_node_name = mapper.outputs_info[
script_loop_input_unique_id]
mapper._check_input(graph, loop_input_node, loop_input_node_name,
node_outputs)
node_outputs, scope_name)
loop_inputs['input'] = loop_input_node_name
loop_outputs.append(block_input_node_name)
node_outputs.append(block_input_node_name)
......@@ -252,14 +300,15 @@ def prim_Loop(mapper, graph, node):
loop_input_node_name = mapper.outputs_info[
script_loop_input_unique_id]
mapper._check_input(graph, loop_input_node, loop_input_node_name,
node_outputs)
node_outputs, scope_name)
graph.add_layer(
"prim.equal",
inputs={'input': loop_input_node_name},
outputs=[block_input_node_name])
outputs=[block_input_node_name],
scope_name=scope_name)
node_outputs.append(block_input_node_name)
graph.add_layer("prim.loop", inputs=loop_inputs, outputs=loop_outputs)
graph.add_layer("prim.loop", inputs=loop_inputs, outputs=loop_outputs, scope_name=scope_name)
current_layer = list(graph.layers.values())[-1]
block_graph, graph_inputs = mapper.traverse(block, current_layer)
for i, input_name in enumerate(graph_inputs):
......@@ -279,6 +328,7 @@ def prim_min(mapper, graph, node):
%86 (list): 输入。
%87 (int): 输出。
"""
scope_name = mapper.normalize_scope_name(node)
output_name = mapper._get_outputs_name(node)[0]
layer_outputs = [output_name]
layer_inputs = {}
......@@ -286,12 +336,12 @@ def prim_min(mapper, graph, node):
# 获取当前节点输出的list
current_outputs = [output_name]
# 处理输入0，即%86
mapper._check_input(graph, inputs_node[0], inputs_name[0], current_outputs)
mapper._check_input(graph, inputs_node[0], inputs_name[0], current_outputs, scope_name)
layer_inputs["input"] = inputs_name[0]
# 获取当前节点输入的list
current_inputs = list(layer_inputs.values())
graph.add_layer("prim.min", inputs=layer_inputs, outputs=layer_outputs)
graph.add_layer("prim.min", inputs=layer_inputs, outputs=layer_outputs, scope_name=scope_name)
return current_inputs, current_outputs
......@@ -304,6 +354,7 @@ def prim_NumToTensor(mapper, graph, node):
%other.2 (Tensor): 输出。
%1736 (-): 输入。
"""
scope_name = mapper.normalize_scope_name(node)
output_name = mapper._get_outputs_name(node)[0]
layer_outputs = [output_name]
layer_inputs = {}
......@@ -312,25 +363,26 @@ def prim_NumToTensor(mapper, graph, node):
# 获取当前节点输出的list
current_outputs = [output_name]
# 处理输入0，即%86
mapper._check_input(graph, inputs_node[0], inputs_name[0], current_outputs)
if inputs_node[0].kind() == "aten::size":
mapper._check_input(graph, inputs_node[0], inputs_name[0], current_outputs, scope_name)
inputs_inputs_name, inputs_inputs_node = mapper._get_inputs_name(inputs_node[0])
if inputs_node[0].kind() == "aten::size" and len(inputs_inputs_name) > 1:
layer_inputs["input"] = inputs_name[0]
# 获取当前节点输入的list
current_inputs = list(layer_inputs.values())
graph.add_layer(
"prim_equal", inputs=layer_inputs, outputs=layer_outputs)
"prim_equal", inputs=layer_inputs, outputs=layer_outputs, scope_name=scope_name)
else:
layer_inputs["value"] = inputs_name[0]
layer_inputs["fill_value"] = inputs_name[0]
# 获取当前节点输入的list
current_inputs = list(layer_inputs.values())
input_type = list(node.inputs())[0].type()
layer_attrs["dtype"] = input_type
layer_attrs["persistable"] = True
layer_attrs["shape"] = [1]
graph.add_layer(
"fluid.layers.create_global_var",
"paddle.full",
inputs=layer_inputs,
outputs=layer_outputs,
scope_name=scope_name,
**layer_attrs)
return current_inputs, current_outputs
......@@ -343,6 +395,7 @@ def prim_RaiseException(mapper, graph, node):
参数含义:
%76 (str): 异常信息。
"""
scope_name = mapper.normalize_scope_name(node)
output_name = mapper._get_outputs_name(node)[0]
layer_outputs = [output_name]
layer_inputs = {}
......@@ -350,13 +403,13 @@ def prim_RaiseException(mapper, graph, node):
# 获取当前节点输出的list
current_outputs = [output_name]
# 处理输入0，即%76
mapper._check_input(graph, inputs_node[0], inputs_name[0], current_outputs)
mapper._check_input(graph, inputs_node[0], inputs_name[0], current_outputs, scope_name)
layer_inputs["input"] = inputs_name[0]
# 获取当前节点输入的list
current_inputs = list(layer_inputs.values())
graph.add_layer(
"prim.exception", inputs=layer_inputs, outputs=layer_outputs)
"prim.exception", inputs=layer_inputs, outputs=layer_outputs, scope_name=scope_name)
return current_inputs, current_outputs
......@@ -369,6 +422,7 @@ def prim_requires_grad(mapper, graph, node):
%356 (bool): 输出，当前Tensor是否计算梯度。
%tensor.31 (Tensor): 输入的Tensor。
"""
scope_name = mapper.normalize_scope_name(node)
output_name = mapper._get_outputs_name(node)[0]
layer_outputs = [output_name]
layer_inputs = {}
......@@ -376,13 +430,13 @@ def prim_requires_grad(mapper, graph, node):
# 获取当前节点输出的list
current_outputs = [output_name]
# 处理输入0，即%86
mapper._check_input(graph, inputs_node[0], inputs_name[0], current_outputs)
mapper._check_input(graph, inputs_node[0], inputs_name[0], current_outputs, scope_name)
layer_inputs["input"] = inputs_name[0]
# 获取当前节点输入的list
current_inputs = list(layer_inputs.values())
graph.add_layer(
"prim.requires_grad", inputs=layer_inputs, outputs=layer_outputs)
"prim.requires_grad", inputs=layer_inputs, outputs=layer_outputs, scope_name=scope_name)
return current_inputs, current_outputs
......@@ -395,6 +449,7 @@ def prim_SetAttr(mapper, graph, node):
%260 (-): 属性名前缀。
%277 (-): 需要设置的值。
"""
scope_name = mapper.normalize_scope_name(node)
output_name = mapper._get_outputs_name(node)[0]
field_name_list = []
tmp_node = node
......@@ -416,7 +471,8 @@ def prim_SetAttr(mapper, graph, node):
graph.add_layer(
"prim.set_attr",
inputs={"input": inputs_name[1]},
outputs=["self." + ".".join(field_name_list).replace(".", "_")])
outputs=["self." + ".".join(field_name_list).replace(".", "_")],
scope_name=scope_name)
return [], [output_name]
......@@ -429,6 +485,7 @@ def prim_shape(mapper, graph, node):
%4701 (list): 输出，shape信息。
%result.1 (Tensor): 需要获取shape的值。
"""
scope_name = mapper.normalize_scope_name(node)
output_name = mapper._get_outputs_name(node)[0]
layer_outputs = [output_name]
layer_inputs = {}
......@@ -436,13 +493,13 @@ def prim_shape(mapper, graph, node):
# 获取当前节点输出的list
current_outputs = [output_name]
# 处理输入0，即%input.8
mapper._check_input(graph, inputs_node[0], inputs_name[0], current_outputs)
mapper._check_input(graph, inputs_node[0], inputs_name[0], current_outputs, scope_name)
layer_inputs["input"] = inputs_name[0]
# 获取当前节点输入的list
current_inputs = list(layer_inputs.values())
graph.add_layer(
"fluid.layers.shape", inputs=layer_inputs, outputs=layer_outputs)
"paddle.shape", inputs=layer_inputs, outputs=layer_outputs, scope_name=scope_name)
return current_inputs, current_outputs
......@@ -456,6 +513,7 @@ def prim_TupleConstruct(mapper, graph, node):
%x.46 (Tensor/其他): tuple第一个元素信息。
%aux (Tensor/其他): tuple第二个元素信息。
"""
scope_name = mapper.normalize_scope_name(node)
output_name = mapper._get_outputs_name(node)[0]
layer_outputs = [output_name]
layer_inputs = {}
......@@ -468,7 +526,7 @@ def prim_TupleConstruct(mapper, graph, node):
# 获取当前节点输入的list
current_inputs = list(layer_inputs.values())
graph.add_layer("prim.tuple", inputs=layer_inputs, outputs=layer_outputs)
graph.add_layer("prim.tuple", inputs=layer_inputs, outputs=layer_outputs, scope_name=scope_name)
return current_inputs, current_outputs
......@@ -482,6 +540,7 @@ def prim_TupleUnpack(mapper, graph, node):
%aux.3 (Tensor/其他): 输出，tuple第二个元素信息。
%4492 (tuple): 需要获取元素的tuple。
"""
scope_name = mapper.normalize_scope_name(node)
outputs_name = mapper._get_outputs_name(node)
layer_outputs = outputs_name
layer_inputs = {}
......@@ -493,7 +552,7 @@ def prim_TupleUnpack(mapper, graph, node):
current_inputs = list(layer_inputs.values())
graph.add_layer(
"prim.tuple_unpack", inputs=layer_inputs, outputs=layer_outputs)
"prim.tuple_unpack", inputs=layer_inputs, outputs=layer_outputs, scope_name=scope_name)
return current_inputs, current_outputs
......@@ -508,6 +567,7 @@ def prim_unchecked_cast(mapper, graph, node):
【注意】Paddle中无此用法，所以此处翻译成赋值。
"""
scope_name = mapper.normalize_scope_name(node)
output_name = mapper._get_outputs_name(node)[0]
layer_outputs = [output_name]
layer_inputs = {}
......@@ -516,12 +576,12 @@ def prim_unchecked_cast(mapper, graph, node):
# 获取当前节点输出的list
current_outputs = [output_name]
# 处理输入0，即%size.63
mapper._check_input(graph, inputs_node[0], inputs_name[0], current_outputs)
mapper._check_input(graph, inputs_node[0], inputs_name[0], current_outputs, scope_name)
layer_inputs["input"] = inputs_name[0]
# 获取当前节点输入的list
current_inputs = list(layer_inputs.values())
graph.add_layer("prim.equal", inputs=layer_inputs, outputs=layer_outputs)
graph.add_layer("prim.equal", inputs=layer_inputs, outputs=layer_outputs, scope_name=scope_name)
return current_inputs, current_outputs
......@@ -533,9 +593,10 @@ def prim_Uninitialized(mapper, graph, node):
参数含义:
%345 (bool): 输出，为赋值的bool。
"""
scope_name = mapper.normalize_scope_name(node)
output_name = mapper._get_outputs_name(node)[0]
output = list(node.outputs())[0]
mapper.attrs[output_name] = None
graph.add_layer(
"prim.constant", inputs={}, outputs=[output_name], value=None)
"prim.constant", inputs={}, outputs=[output_name], scope_name=scope_name, value=None)
return [], [output_name]
x2paddle/op_mapper/dygraph/pytorch2paddle/pytorch_op_mapper.py
浏览文件 @ c05a67a4
......@@ -17,24 +17,29 @@ import numpy as np
from x2paddle.core.op_mapper import OpMapper
from x2paddle.core.util import *
from x2paddle.core.program import PaddleGraph
from x2paddle.op_mapper.pytorch2paddle import prim
from x2paddle.op_mapper.pytorch2paddle import aten
from x2paddle.op_mapper.dygraph.pytorch2paddle import prim
from x2paddle.op_mapper.dygraph.pytorch2paddle import aten
class PyTorchOpMapper(OpMapper):
def __init__(self, decoder):
super(PyTorchOpMapper, self).__init__()
self.script = decoder.script
self.input_examples = decoder.input_examples
self.paddle_params = dict()
self.outputs_info = {} # key为output unique id，value为当前节点的输出名字
self.pytorch_params = {} # key为节点名，value为参数
self.attrs = {} # key为节点名，value为属性值
self.output_index = 0
self.dygraph_name_id = {} # 动态图__init__输出名字中的id，key为kernel类型，value为id
self.nn_name2id = {} # 动态图__init__输出名字中的id，key为kernel类型，value为id
self.split_len = {} # split的长度
self.scope_name_list = list()
self.scope_name2id = dict()
self.inputs_info = dict()
# 转换
self.check_op(decoder.graph)
self.graph, _ = self.traverse(decoder.graph)
self.paddle_graph, _ = self.traverse(decoder.graph)
self.paddle_graph.set_inputs_info(self.inputs_info)
def check_op(self, script_graph):
def _update_op_list(graph):
......@@ -72,17 +77,21 @@ class PyTorchOpMapper(OpMapper):
current_node_outputs.extend(outputs)
# 初始化
graph = PaddleGraph(parent_layer, graph_type="dygraph")
graph = PaddleGraph(parent_layer=parent_layer, graph_type="dygraph")
if "TopLevelTracedModule" in str(type(self.script)):
graph.set_script(self.script)
current_node_outputs = []
graph_inputs = []
# 转换输入节点
if isinstance(script_graph, torch._C.Graph):
input_ct = 0
for i, ivalue in enumerate(script_graph.inputs()):
node = ivalue.node()
if str(ivalue.type()) != "Tensor":
if str(ivalue.type()) not in ["Tensor", "Dict[str, Tensor]"]:
graph.set_name(str(ivalue.type()).split(".")[-1])
continue
inputs, outputs = self.data(graph, node, ivalue.unique())
inputs, outputs = self.data(graph, node, ivalue.unique(), input_ct)
input_ct += 1
# 转换中间节点
for node in script_graph.nodes():
kind = node.kind()
......@@ -120,7 +129,7 @@ class PyTorchOpMapper(OpMapper):
graph.outputs = inputs_name
# 更新split参数
for layer in graph.layers.values():
if layer.kernel == "fluid.layers.split" and "num_or_sections" in layer.attrs:
if layer.kernel == "paddle.split" and "num_or_sections" in layer.attrs:
layer.attrs["num_or_sections"] = self.split_len[layer.outputs[
0]]
return graph, graph_inputs
......@@ -151,6 +160,7 @@ class PyTorchOpMapper(OpMapper):
node,
output_name,
node_outputs,
scope_name,
add_dim=False):
if node.kind() == "prim::GetAttr":
param = self.pytorch_params[output_name]
......@@ -159,10 +169,13 @@ class PyTorchOpMapper(OpMapper):
param = param[np.newaxis, :]
self.paddle_params[output_name] = param
graph.add_layer(
"fluid.dygraph.base.to_variable",
"self.create_parameter",
inputs={},
outputs=[output_name],
value="params[{}]".format(string(output_name)))
scope_name=scope_name,
dtype=string(str(param.dtype)),
shape = param.shape,
default_initializer="paddle.nn.initializer.Constant(value=0.0)")
else:
if isinstance(param, dict) and "Tensor" in param and \
"parent_layer_id" in param:
......@@ -183,11 +196,13 @@ class PyTorchOpMapper(OpMapper):
param = param[np.newaxis, :]
self.paddle_params[output_name] = param
graph.add_layer(
"fluid.dygraph.base.to_variable",
"self.create_parameter",
inputs={},
outputs=[output_name],
value="params[{}]".format(
string(output_name)))
scope_name=scope_name,
dtype=string(str(param.dtype)),
shape = param.shape,
default_initializer="paddle.nn.initializer.Constant(value=0.0)")
node_outputs.append(output_name)
return
# 若if-else外，则可直接引用if-else中的赋值结果
......@@ -195,16 +210,30 @@ class PyTorchOpMapper(OpMapper):
"prim.constant",
inputs={},
outputs=[output_name],
scope_name=scope_name,
value=param["Tensor"])
else:
graph.add_layer(
"prim.constant",
inputs={},
outputs=[output_name],
scope_name=scope_name,
value=string(param)
if isinstance(param, str) else param)
node_outputs.append(output_name)
elif node.kind() == "prim::Constant" and output_name in self.pytorch_params:
param = self.pytorch_params[output_name]
self.paddle_params[output_name] = param
graph.add_layer(
"self.create_parameter",
inputs={},
outputs=[output_name],
scope_name=scope_name,
dtype=string(str(param.dtype)),
shape = param.shape,
default_initializer="paddle.nn.initializer.Constant(value=0.0)")
def _get_inputs_name(self, node):
inputs_name = []
inputs_node = []
......@@ -215,8 +244,10 @@ class PyTorchOpMapper(OpMapper):
inputs_node.append(script_input_node)
inputs_name.append(input_name)
return inputs_name, inputs_node
def data(self, graph, node, uid):
def data(self, graph, node, uid, input_ct):
scope_name = self.normalize_scope_name(node)
for output_ivalue in node.outputs():
script_unique_id = output_ivalue.unique()
if script_unique_id in self.outputs_info or script_unique_id != uid:
......@@ -226,13 +257,18 @@ class PyTorchOpMapper(OpMapper):
self.output_index += 1
output_name = self.outputs_info[uid]
graph.add_layer(
"fluid.dygraph.base.to_variable",
"paddle.to_tensor",
inputs={},
outputs=[node_name],
value=output_name)
scope_name=scope_name,
data=output_name)
if self.input_examples is not None:
input_np = self.input_examples[input_ct].detach().numpy()
self.inputs_info[output_name] = [list(input_np.shape), str(input_np.dtype)]
return [], [output_name]
def equal(self, graph, node, uid=None, parent_layer=None, index=None):
scope_name = self.normalize_scope_name(node)
if parent_layer is not None and index is not None:
# block的输出
input_node_name = self.outputs_info[uid]
......@@ -241,9 +277,61 @@ class PyTorchOpMapper(OpMapper):
control_output_id = index - 1
output_node_name = parent_layer.outputs[control_output_id]
current_outputs = [output_node_name]
self._check_input(graph, node, input_node_name, current_outputs)
self._check_input(graph, node, input_node_name, current_outputs, scope_name)
graph.add_layer(
"prim.equal",
inputs={'input': input_node_name},
outputs=[output_node_name])
outputs=[output_node_name],
scope_name=scope_name)
return [input_node_name], current_outputs
def normalize_scope_name(self, node):
""" 对scope的名字进行标准化。
"""
scope_name = node.scopeName()
if scope_name == "":
return scope_name
scope_name_part = scope_name.split("/")
for index in range(len(scope_name_part) - 1):
if scope_name_part[index] in scope_name_part[index + 1]:
continue
last_name_segments = scope_name_part[index].split(".")
name_segments = scope_name_part[index + 1].split(".")
for j, name in enumerate(last_name_segments):
name_segments[j] = name
scope_name_part[index + 1] = ".".join(name_segments)
last_name = scope_name_part[-1]
name_segments = last_name.split(".")
for i, ns in enumerate(name_segments):
if i not in self.scope_name2id:
self.scope_name2id[i] = dict()
if ns not in self.scope_name2id[i]:
self.scope_name2id[i][ns] = 0
real_scope_name = "/".join(name_segments[1:])
real_father_scope_name = "/".join(name_segments[1:-1])
for i, ns in enumerate(name_segments):
if i == 0:
continue
if self.scope_name2id[i][ns] != 0:
name_segments[i] = name_segments[i] + \
"__{}".format(self.scope_name2id[i][ns])
prefix_scope_name = "/".join(name_segments[1 :i + 1])
is_found = False
for j in range(len(self.scope_name_list)):
last_scope_name = self.scope_name_list[-1-j]
if last_scope_name.startswith(prefix_scope_name + "/") \
or last_scope_name == prefix_scope_name:
if j != 0: # and i != len(name_segments) - 1:
is_found = True
origin_name_segment_i = name_segments[i].split("__")[0]
self.scope_name2id[i][origin_name_segment_i] += 1
name_segments[i] = origin_name_segment_i + \
"__" + str(self.scope_name2id[i][origin_name_segment_i])
break
if is_found:
break
real_scope_name = "/".join(name_segments[1:])
self.scope_name_list.append(real_scope_name)
return real_scope_name
\ No newline at end of file
x2paddle/optimizer/code_optimizer/__init__.py 0 → 100644
浏览文件 @ c05a67a4
# 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.
from x2paddle.optimizer.code_optimizer.hierachical_tree import HierarchicalTree
\ No newline at end of file
x2paddle/optimizer/code_optimizer/hierachical_tree.py 0 → 100644
浏览文件 @ c05a67a4
此差异已折叠。
x2paddle/optimizer/code_optimizer/parameter_tree.py 0 → 100644
浏览文件 @ c05a67a4
# 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.
class PamareterNode(object):
def __init__(self, old_name=None, new_name=None):
self.old_name = old_name
self.new_name = new_name
self.childs = list()
def add_child(self, child):
self.childs.append(child)
def has_child(self):
if len(self.childs) == 0:
return False
else:
return True
class PamareterTree(object):
def __init__(self):
self.nodes = list()
self.old2new = dict()
def add_node(self, node):
self.nodes.append(node)
def traverse(self):
tmp = list()
def recurs(node, prefix_name):
for child in node.childs:
child_prefix_name = prefix_name + "." + child.new_name
if child.has_child():
recurs(child, child_prefix_name)
else:
self.old2new[child.old_name] = child_prefix_name[1:]
recurs(self.nodes[-1], "")
def get_node(self, old_name):
for node in self.nodes:
if node.old_name == old_name:
return node
\ No newline at end of file
x2paddle/optimizer/code_optimizer/subgraphs_union.py 0 → 100644
浏览文件 @ c05a67a4
此差异已折叠。
x2paddle/optimizer/fusion/dygraph/__init__.py
浏览文件 @ c05a67a4
......@@ -32,3 +32,5 @@ from .reshape_fuser import DygraphReshapeFuser
from .reshape_fuse_pass import DygraphReshapeFusePass
from .tf_batchnorm_fuser import DygraphTFBatchNormFuser
from .tf_batchnorm_fuse_pass import DygraphTFBatchNormFusePass
from .trace_fc_fuser import TraceFcFuser
from .trace_fc_fuse_pass import TraceFcFusePass
x2paddle/optimizer/fusion/dygraph/trace_fc_fuse_pass.py 0 → 100644
浏览文件 @ c05a67a4
# 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.
from x2paddle.optimizer.pass_ import Pass
from x2paddle.optimizer.fusion.dygraph import TraceFcFuser
from x2paddle.optimizer.pass_manager import pass_register
@pass_register
class TraceFcFusePass(Pass):
name = "trace_fc_fuse_pass"
def __init__(self):
Pass.__init__(self)
def apply(self, graph):
fuser = TraceFcFuser()
fuser.operate(graph, match_kind="topo")
# 用于注册
trace_fc_fuse_pass = TraceFcFusePass()
\ No newline at end of file
x2paddle/optimizer/fusion/dygraph/trace_fc_fuser.py 0 → 100644
浏览文件 @ c05a67a4
此差异已折叠。
x2paddle/optimizer/optimizer.py
浏览文件 @ c05a67a4
......@@ -18,14 +18,20 @@ from x2paddle.optimizer.fusion.static import *
from x2paddle.optimizer.elimination.dygraph import *
class GraphOptimizer(object):
def __init__(self, source_frame, paddle_type="dygraph"):
def __init__(self, source_frame, paddle_type="dygraph", jit_type="trace"):
if source_frame == "pytorch":
self.passes = [
"dygraph_constant_fuse_pass", "dygraph_batchnorm2d_fuse_pass",
"dygraph_interpolate_bilinear_fuse_pass", "dygraph_fc_fuse_pass",
"dygraph_adaptive_pool2d_fuse_pass", "dygraph_reshape_fuse_pass",
"dygraph_dropout_fuse_pass"
]
if jit_type == "trace":
self.passes = ["trace_fc_fuse_pass"]
else:
self.passes = [
"dygraph_constant_fuse_pass",
"dygraph_batchnorm2d_fuse_pass",
"dygraph_interpolate_bilinear_fuse_pass",
"dygraph_fc_fuse_pass",
"dygraph_adaptive_pool2d_fuse_pass",
"dygraph_reshape_fuse_pass",
"dygraph_dropout_fuse_pass"
]
elif source_frame == "caffe":
if paddle_type == "dygraph":
self.passes = ["dygraph_bn_scale_fuse_pass"]
......@@ -38,8 +44,7 @@ class GraphOptimizer(object):
"transpose_eliminate_pass"
]
else:
# TODO
pass
self.passes = []
def optimize(self, graph):
for pass_name in self.passes:
......
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