paddlepaddle / paconvert Goto Github PK

View Code? Open in Web Editor NEW

67.0 67.0 40.0 1.75 MB

Code Convert to PaddlePaddle Toolkit

License: Apache License 2.0

Python 98.17% Makefile 0.03% Shell 0.54% C++ 0.09% Cuda 1.17%

paconvert's People

Contributors

Stargazers

Watchers

paconvert's Issues

got the wrong result when using paconvet

Pytorch

import torch.backends.cudnn as cudnn

cudnn.benchmark = True

Paddle

import paddle

False = True

转换出现bug

torch

import torch
import torch.nn as nn
import torch.nn.functional as F
from typing import List, Tuple, Optional
import math

# Calculate asymmetric TensorFlow-like 'SAME' padding for a convolution
def get_same_padding(x: int, kernel_size: int, stride: int, dilation: int):
    if isinstance(x, torch.Tensor):
        return torch.clamp(((x / stride).ceil() - 1) * stride + (kernel_size - 1) * dilation + 1 - x, min=0)
    else:
        return max((math.ceil(x / stride) - 1) * stride + (kernel_size - 1) * dilation + 1 - x, 0)


# Dynamically pad input x with 'SAME' padding for conv with specified args
def pad_same(
        x,
        kernel_size: List[int],
        stride: List[int],
        dilation: List[int] = (1, 1),
        value: float = 0,
):
    ih, iw = x.size()[-2:]
    pad_h = get_same_padding(ih, kernel_size[0], stride[0], dilation[0])
    pad_w = get_same_padding(iw, kernel_size[1], stride[1], dilation[1])
    x = F.pad(x, (pad_w // 2, pad_w - pad_w // 2, pad_h // 2, pad_h - pad_h // 2), value=value)
    return x


def avg_pool2d_same(x, kernel_size: List[int], stride: List[int], padding: List[int] = (0, 0),
                    ceil_mode: bool = False, count_include_pad: bool = True):
    # FIXME how to deal with count_include_pad vs not for external padding?
    x = pad_same(x, kernel_size, stride)
    return F.avg_pool2d(x, kernel_size, stride, (0, 0), ceil_mode, count_include_pad)

paddlepaddle

import sys
sys.path.append('/home/greatx/repos/PaConvert/paddle_project/utils')
import paddle_aux
import paddle
from typing import List, Tuple, Optional
import math


def get_same_padding(x: int, kernel_size: int, stride: int, dilation: int):
    if isinstance(x, paddle.Tensor):
        return paddle.clip(x=((x / stride).ceil() - 1) * stride + (
            kernel_size - 1) * dilation + 1 - x, min=0)
    else:
        return max((math.ceil(x / stride) - 1) * stride + (kernel_size - 1) *
            dilation + 1 - x, 0)


def pad_same(x, kernel_size: List[int], stride: List[int], dilation: List[
    int]=(1, 1), value: float=0):
    ih, iw = x.shape[-2:]
    pad_h = get_same_padding(ih, kernel_size[0], stride[0], dilation[0])
    pad_w = get_same_padding(iw, kernel_size[1], stride[1], dilation[1])
    x = paddle_aux._FUNCTIONAL_PAD(pad=(pad_w // 2, pad_w - pad_w // 2, 
        pad_h // 2, pad_h - pad_h // 2), value=value, x=x)
    return x


def avg_pool2d_same(x, kernel_size: List[int], stride: List[int], padding:
    List[int]=(0, 0), ceil_mode: bool=False, count_include_pad: bool=True):
    x = pad_same(x, kernel_size, stride)
    return paddle.nn.functional.avg_pool2d(kernel_size=kernel_size, stride=
        stride, padding=(0, 0), ceil_mode=ceil_mode, x=x, exclusive=
        notcount_include_pad)

count_include_pad -> not count_include_pad -> notcount_include_pad

【社区治理】co63oc 发起 Committer 身份申请

问题描述 Please describe your issue

基本信息

申请人 GitHub ID	PaConvert Repo 整体 merge PR 数	PaConvert Repo 整体 review PR 数	PaConvert Repo 整体报告 Issue 数
@co63oc	50	0	0

社区贡献概况

以下是此前在本 repo 里做过的贡献：
实现若干算子转换规则

高质量 merge PR 展示

PR 号	PR 标题	PR 简介	Reviewer
#197	转换规则 No. 196/197/232/233/319	-	zhwesky2010
#198	转换规则 No. 323/333	-	zhwesky2010
#199	转换规则 No. 349/350	-	zhwesky2010

担保人意见

@Ligoml @luotao1

`torch.distributed.is_available()` and `torch.distributed.is_initialized()` cannot be converted

torch.distributed.is_available()
torch.distributed.is_initialized()
torch.distributed.get_world_size()

import torch.distributed as dist

def is_dist_avail_and_initialized():
    if not dist.is_available():
        return False
    if not dist.is_initialized():
        return False
    return True

Pytorch-Paddle代码转换工具开源任务

✨ 问题描述：

大家好，为了实现将 PyTorch 代码自动化的转写成 Paddle 代码，从而提升模型迁移的效率，我们建设了 代码自动转换工具: PaddlePaddle Code Convert Toolkits，目前已支持了1000+个Pytorch API的自动转换，我们此次对外开放408个API的转换规则开发，欢迎大家提 PR 来一起支持自动转换 🎉🎉🎉。

通过本次活动，你可以更详细地了解 PyTorch 框架与 Paddle 框架用法及设计差异，提升自己对深度学习框架的熟悉程度。

🍻 你需要做的是：

我们已将任务记录在《在线任务明细表》，为方便大家自由选择所熟悉的API，此次不对API进行分组，大家可自由选择一个或多个API来实现自动转换，认领时直接在本issue下回复认领的任务ID（至少1个，建议一次认领多个，且越多约好）。欢迎大家认领任务和提 PR~

Fork PaddlePaddle/docs 、PaddlePaddle/PaConvert 两个Github Reop。
书写API映射关系：PR提交到 PaddlePaddle/docs 下，需要为每个 API 新增对应的 md 文件并放入docs/guides/model_convert/convert_from_pytorch/api_difference 对应的目录下，文件名为PyTorch API名。如果已存在该API的映射关系，则无需新增 md 文件，只需要检查并校正之前的文档是否正确，如果与后面的AST规则有差异，则需要修改文档。
API映射关系请参考《API映射关系-格式规范》，PR标题格式：映射文档 No. xxx/yyy/zzz，PR描述附上本issue。请严格根据格式规范来书写文档，避免因格式问题增加不必要的review成本，不满足格式规范的PR将不予合入。
API映射关系相当于人工转换的思路，在其完成后，即可开发AST自动转换规则：参考《AST转换规则开发步骤》中步骤3~5，PR标题格式：转换规则 No. xxx/yyy/zzz，PR描述附上本issue与上述文档PR。请严格根据文档要求来开发代码，避免增加不必要的review成本，不满足要求的PR将不予合入。
每1个任务No.xxx 均包含 1个映射关系文档PR + 1个转换规则PR，两者均合入该任务才算完成，请尽可能一次性提交多个任务，以提高review效率。
review方式：评论里 @zhwesky2010 review，请及时修改review意见，review通过后将合入代码。

✨ 注：

该任务时间：PR 截止合入时间是2023/10/30。
认领规则：直接在 issue 下回复认领的任务 ID（建议一次认领多个）。
提交PR前请参照官网安装pre-commit，检查代码格式。否则CI可能无法通过。
PR请先通过CI检查后再发起review，避免增加不必要的review成本。
PR标题格式：映射文档 No. xxx/yyy/zzz，转换规则 No. xxx/yyy/zzz，PR描述均需附上本issue，后者PR描述的 PR Docs 里需要写上前者PR的链接。
如果该API在paddle中存在 功能缺失、功能Bug、功能diff 等问题，导致无法转换，请直接在本issue下回复。我们会定期确认并记录API功能问题，同时对于此问题点，可暂不开发Matcher但仍需开发 屏蔽版本的测试case（单测屏蔽方式可查询开发文档或参考已有单测的代码）。问题描述参考如下格式：
- torch.diff 问题：paddle仅支持n=1，torch的n可支持任意值
- torch.nonzero 问题：指定as_tuple时行为不一致，paddle会多一维且不合理
- torch.dstack 问题：功能缺失
任务明细表中已对映射关系的分类进行了初步粗略标注，仅供参考，最终以开发者分析为主。
历史上的 good first issue 列表，也欢迎来提 PR 解决~ 欢迎联系花花加入社区，和我们一起快乐开源！

got the wrong result when using paconvet

pytorch

import torch.nn as nn
from functools import partial


class test(nn.Module):
    def __init__(self, in_channels, out_channels, norm_func=nn.LayerNorm):
        super(test, self).__init__()
        self.norm = norm_func(in_channels)
        self.linear = nn.Linear(in_channels, out_channels)
    
    def forward(self, x):
        x = self.norm(x)
        x = self.linear(x)
        return x

if __name__ == "__main__":
    model = test(10, 10, partial(nn.LayerNorm, eps=0.2))

paddle

import paddle
from functools import partial


class test(paddle.nn.Layer):

    def __init__(self, in_channels, out_channels, norm_func=paddle.nn.LayerNorm
        ):
        super(test, self).__init__()
        self.norm = norm_func(in_channels)
        self.linear = paddle.nn.Linear(in_features=in_channels,
            out_features=out_channels)

    def forward(self, x):
        x = self.norm(x)
        x = self.linear(x)
        return x


if __name__ == '__main__':
    model = test(10, 10, partial(paddle.nn.LayerNorm, eps=0.2))

Traceback (most recent call last):
  File "/home/greatx/repos/PaConvert/paddle_project/test.py", line 21, in <module>
    model = test(10, 10, partial(paddle.nn.LayerNorm, eps=0.2))
  File "/home/greatx/repos/PaConvert/paddle_project/test.py", line 10, in __init__
    self.norm = norm_func(in_channels)
TypeError: LayerNorm.__init__() got an unexpected keyword argument 'eps'

eps should be converted to epsilon.

Torch code 转 Paddle code 失败

使用命令pytest tests/test_cummin.py测试测试案例时失败，发现在test_project/paddle_temp.py中的代码没有更改为Paddle code，请问该如何解决该问题。

报错信息如下：

================================================= test session starts =================================================
platform win32 -- Python 3.9.18, pytest-7.4.2, pluggy-1.3.0 -- D:\anaconda2\envs\hackthon\python.exe
cachedir: .pytest_cache
rootdir: C:\Users\lfy\Desktop\PaConvert\tests
configfile: pytest.ini
plugins: anyio-4.0.0, cov-4.1.0
collected 1 item

tests\test_cummin.py::test_case_1 FAILED

====================================================== FAILURES =======================================================
_____________________________________________________ test_case_1 _____________________________________________________

    def test_case_1():
        pytorch_code = textwrap.dedent(
            """
            import torch
            x = torch.tensor([[1.0, 1.0, 1.0],
                            [2.0, 2.0, 2.0],
                            [3.0, 3.0, 3.0]])
            result = torch.cummin(x, 0)
            """
        )
>       obj.run(pytorch_code, ["result"])

tests\test_cummin.py:32:
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _
tests\apibase.py:89: in run
    self.compare(
tests\apibase.py:165: in compare
    self.compare(
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _

self = <apibase.APIBase object at 0x0000024B5A25EF70>, name = 'torch.cummin'
pytorch_result = tensor([[1., 1., 1.],
        [1., 1., 1.],
        [1., 1., 1.]])
paddle_result = tensor([[1., 1., 1.],
        [1., 1., 1.],
        [1., 1., 1.]]), check_value = True
check_dtype = True, check_stop_gradient = True, rtol = 1e-06, atol = 0.0

    def compare(
        self,
        name,
        pytorch_result,
        paddle_result,
        check_value=True,
        check_dtype=True,
        check_stop_gradient=True,
        rtol=1.0e-6,
        atol=0.0,
    ):
        """
        compare tensors' data, shape, requires_grad, dtype
        args:
            name: pytorch api name
            pytorch_result: pytorch Tensor
            paddle_result: paddle Tensor
            check_value: If false, the value will not be checked
            check_dtype: If false, the dtype will not be checked
            check_stop_gradient: If false, the stop gradient will not be checked
        """
        if isinstance(pytorch_result, dict):
            assert isinstance(paddle_result, dict), "paddle result should be dict"
            assert len(pytorch_result) == len(
                paddle_result
            ), "paddle result have different length with pytorch"
            pytorch_result_k = [k for k in pytorch_result.keys()]
            pytorch_result_v = [v for v in pytorch_result.values()]
            paddle_result_k = [k for k in paddle_result.keys()]
            paddle_result_v = [v for v in paddle_result.values()]
            self.compare(
                self.pytorch_api,
                pytorch_result_k,
                paddle_result_k,
                check_value,
                check_dtype,
                check_stop_gradient,
                rtol,
                atol,
            )
            self.compare(
                self.pytorch_api,
                pytorch_result_v,
                paddle_result_v,
                check_value,
                check_dtype,
                check_stop_gradient,
                rtol,
                atol,
            )
            return

        if isinstance(pytorch_result, (tuple, list)):
            assert isinstance(
                paddle_result, (tuple, list)
            ), "paddle result should be list/tuple"
            assert len(pytorch_result) == len(
                paddle_result
            ), "paddle result have different length with pytorch"
            for i in range(len(pytorch_result)):
                self.compare(
                    self.pytorch_api,
                    pytorch_result[i],
                    paddle_result[i],
                    check_value,
                    check_dtype,
                    check_stop_gradient,
                    rtol,
                    atol,
                )
            return

        if isinstance(pytorch_result, (bool, np.number, int, str, type(None))):
            assert type(paddle_result) == type(
                pytorch_result
            ), "paddle result's type [{}] should be the same with pytorch's type [{}]".format(
                type(paddle_result), type(pytorch_result)
            )
            if check_value:
                assert (
                    pytorch_result == paddle_result
                ), "API ({}): pytorch result is {}, but paddle result is {}".format(
                    name, pytorch_result, paddle_result
                )
            return

        if pytorch_result.requires_grad:
            pytorch_numpy, paddle_numpy = (
                pytorch_result.detach().numpy(),
                paddle_result.numpy(False),
            )
        elif pytorch_result.is_conj():
            pytorch_numpy, paddle_numpy = (
                pytorch_result.resolve_conj().numpy(),
                paddle_result.numpy(False),
            )
        else:
            (
                pytorch_numpy,
                paddle_numpy,
>           ) = pytorch_result.cpu().numpy(), paddle_result.numpy(False)
E           TypeError: numpy() takes 0 positional arguments but 1 was given

tests\apibase.py:205: TypeError
-------------------------------------------------- Captured log call --------------------------------------------------
INFO     Converter_0:utils.py:91 ===========================================
INFO     Converter_0:utils.py:91 PyTorch to Paddle Convert Start ------>:
INFO     Converter_0:utils.py:91 ===========================================
INFO     Converter_0:utils.py:91 Start convert file: C:\Users\lfy\Desktop\PaConvert\test_project\pytorch_temp.py --> C:\Users\lfy\Desktop\PaConvert\test_project\paddle_temp.py
INFO     Converter_0:utils.py:91 Finish convert C:\Users\lfy\Desktop\PaConvert\test_project\pytorch_temp.py --> C:\Users\lfy\Desktop\PaConvert\test_project\paddle_temp.py

INFO     Converter_0:utils.py:91
===========================================
INFO     Converter_0:utils.py:91 Convert Summary:
INFO     Converter_0:utils.py:91 ===========================================
INFO     Converter_0:utils.py:91 There are 0 Pytorch APIs in this Project:
INFO     Converter_0:utils.py:91  0  Pytorch APIs have been converted to Paddle successfully!
INFO     Converter_0:utils.py:91  0  Pytorch APIs are not supported to convert to Paddle currently!
INFO     Converter_0:utils.py:91  Convert Rate is: 0.000%
INFO     Converter_0:utils.py:91
Thank you to use Paddle Code Convert Tool. You can make any suggestions to us.
=============================================== short test summary info ===============================================
FAILED tests\test_cummin.py::test_case_1 - TypeError: numpy() takes 0 positional arguments but 1 was given
================================================== 1 failed in 3.25s ==================================================

torch.nn.functional.pad无法转换成功

结果如下：

    def _relative_position_to_absolute_position(self, x):
        """
        x: [b, h, l, 2*l-1]
        ret: [b, h, l, l]
        """
        batch, heads, length, _ = x.shape
>>>        x = torch.nn.functional.pad(x, commons.convert_pad_shape([[0, 0], [
            0, 0], [0, 0], [0, 1]]))
        """Class Method: *.view, can not convert, please check whether it is torch.Tensor.*/Optimizer.*/nn.Module.*/torch.distributions.Distribution.*/torch.autograd.function.FunctionCtx.*/torch.profiler.profile.*/torch.autograd.profiler.profile.*, and convert manually"""
>>>        x_flat = x.view([batch, heads, length * 2 * length])
>>>        x_flat = torch.nn.functional.pad(x_flat, commons.convert_pad_shape(
            [[0, 0], [0, 0], [0, length - 1]]))
        """Class Method: *.view, can not convert, please check whether it is torch.Tensor.*/Optimizer.*/nn.Module.*/torch.distributions.Distribution.*/torch.autograd.function.FunctionCtx.*/torch.profiler.profile.*/torch.autograd.profiler.profile.*, and convert manually"""
>>>        x_final = x_flat.view([batch, heads, length + 1, 2 * length - 1])[:,
            :, :length, length - 1:]
        return x_final

    def _absolute_position_to_relative_position(self, x):
        """
        x: [b, h, l, l]
        ret: [b, h, l, 2*l-1]
        """
        batch, heads, length, _ = x.shape
>>>        x = torch.nn.functional.pad(x, commons.convert_pad_shape([[0, 0], [
            0, 0], [0, 0], [0, length - 1]]))
        """Class Method: *.view, can not convert, please check whether it is torch.Tensor.*/Optimizer.*/nn.Module.*/torch.distributions.Distribution.*/torch.autograd.function.FunctionCtx.*/torch.profiler.profile.*/torch.autograd.profiler.profile.*, and convert manually"""
>>>        x_flat = x.view([batch, heads, length ** 2 + length * (length - 1)])
>>>        x_flat = torch.nn.functional.pad(x_flat, commons.convert_pad_shape(
            [[0, 0], [0, 0], [length, 0]]))
        """Class Method: *.view, can not convert, please check whether it is torch.Tensor.*/Optimizer.*/nn.Module.*/torch.distributions.Distribution.*/torch.autograd.function.FunctionCtx.*/torch.profiler.profile.*/torch.autograd.profiler.profile.*, and convert manually"""
>>>        x_final = x_flat.view([batch, heads, length, 2 * length])[:, :, :, 1:]
        return x_final

got wrong output

 git clone https://github.com/facebookresearch/detectron2.git
cd PaConvert
python paconvert/main.py --in_dir ../detectron2/ --out_dir ../detectron2.paddle

original: detectron2/detectron2/utils/comm.py

# Copyright (c) Facebook, Inc. and its affiliates.
"""
This file contains primitives for multi-gpu communication.
This is useful when doing distributed training.
"""

import functools
import numpy as np
import torch
import torch.distributed as dist

_LOCAL_PROCESS_GROUP = None
_MISSING_LOCAL_PG_ERROR = (
    "Local process group is not yet created! Please use detectron2's `launch()` "
    "to start processes and initialize pytorch process group. If you need to start "
    "processes in other ways, please call comm.create_local_process_group("
    "num_workers_per_machine) after calling torch.distributed.init_process_group()."
)

Converted: detectron2.paddle/detectron2/utils/comm.py

import paddle
"""
This file contains primitives for multi-gpu communication.
This is useful when doing distributed training.
"""
import functools
import numpy as np
_LOCAL_PROCESS_GROUP = None
_MISSING_LOCAL_PG_ERROR = (
>>>>>>    "Local process group is not yet created! Please use detectron2's `launch()` to start processes and initialize pytorch process group. If you need to start processes in other ways, please call comm.create_local_process_group(num_workers_per_machine) after calling torch.distributed.init_process_group()."
    )

torch_scatter库的paddle转换

torch_scatter 是一个torch的扩增库，目前在科学计算、神经辐射场等领域均有用到，例如第四期黑客松的赛题No.173 Point-NeRF就涉及torch_scatter中的scatter_min算子，无法通过paddle复现，像简单的scatter_add倒是可以组合复现，但是也很大的增加了复现的成本。请问您这边有计划推出torch_scatter的paddle转换吗

Conversion fails when `\u` is in the code

code

import numpy as np

def extract_vertices(lines):
	'''extract vertices info from txt lines
	Input:
		lines   : list of string info
	Output:
		vertices: vertices of text regions <numpy.ndarray, (n,8)>
		labels  : 1->valid, 0->ignore, <numpy.ndarray, (n,)>
	'''
	labels = []
	vertices = []
	for line in lines:
		vertices.append(list(map(int,line.rstrip('\n').lstrip('\ufeff').split(',')[:8])))
		label = 0 if '###' in line else 1
		labels.append(label)
	return np.array(vertices), np.array(labels)

output

python paconvert/main.py --in_dir ~/repos/bug_test/ --out_dir ~/repos/bug_test_
===========================================
PyTorch to Paddle Convert Start ------>:
===========================================
Start convert file: /home/greatx/repos/bug_test/test.py --> /home/greatx/repos/bug_test_/test.py
Traceback (most recent call last):
  File "/home/greatx/miniconda3/envs/jit-env/lib/python3.10/sre_parse.py", line 1051, in parse_template
    this = chr(ESCAPES[this][1])
KeyError: '\\u'

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
  File "/home/greatx/repos/PaConvert/paconvert/main.py", line 145, in <module>
    main()
  File "/home/greatx/repos/PaConvert/paconvert/main.py", line 131, in main
    converter.run(args.in_dir, args.out_dir, args.exclude_dirs)
  File "/home/greatx/repos/PaConvert/paconvert/../paconvert/converter.py", line 88, in run
    self.transfer_dir(in_dir, out_dir, exclude_dir_list)
  File "/home/greatx/repos/PaConvert/paconvert/../paconvert/converter.py", line 186, in transfer_dir
    self.transfer_dir(old_path, new_path, exclude_dir_list)
  File "/home/greatx/repos/PaConvert/paconvert/../paconvert/converter.py", line 164, in transfer_dir
    self.transfer_file(old_path, new_path)
  File "/home/greatx/repos/PaConvert/paconvert/../paconvert/converter.py", line 202, in transfer_file
    self.transfer_node(root, old_path)
  File "/home/greatx/repos/PaConvert/paconvert/../paconvert/converter.py", line 242, in transfer_node
    trans.transform()
  File "/home/greatx/repos/PaConvert/paconvert/../paconvert/base.py", line 81, in transform
    self.visit(self.root)
  File "/home/greatx/repos/PaConvert/paconvert/../paconvert/base.py", line 86, in visit
    node = super(BaseTransformer, self).visit(node)
  File "/home/greatx/miniconda3/envs/jit-env/lib/python3.10/ast.py", line 418, in visit
    return visitor(node)
  File "/home/greatx/repos/PaConvert/paconvert/../paconvert/base.py", line 295, in visit_Module
    super(BaseTransformer, self).generic_visit(node)
  File "/home/greatx/miniconda3/envs/jit-env/lib/python3.10/ast.py", line 494, in generic_visit
    value = self.visit(value)
  File "/home/greatx/repos/PaConvert/paconvert/../paconvert/base.py", line 86, in visit
    node = super(BaseTransformer, self).visit(node)
  File "/home/greatx/miniconda3/envs/jit-env/lib/python3.10/ast.py", line 418, in visit
    return visitor(node)
  File "/home/greatx/repos/PaConvert/paconvert/../paconvert/base.py", line 247, in visit_FunctionDef
    super(BaseTransformer, self).generic_visit(node)
  File "/home/greatx/miniconda3/envs/jit-env/lib/python3.10/ast.py", line 494, in generic_visit
    value = self.visit(value)
  File "/home/greatx/repos/PaConvert/paconvert/../paconvert/base.py", line 86, in visit
    node = super(BaseTransformer, self).visit(node)
  File "/home/greatx/miniconda3/envs/jit-env/lib/python3.10/ast.py", line 418, in visit
    return visitor(node)
  File "/home/greatx/repos/PaConvert/paconvert/../paconvert/base.py", line 277, in visit_For
    super(BaseTransformer, self).generic_visit(node)
  File "/home/greatx/miniconda3/envs/jit-env/lib/python3.10/ast.py", line 494, in generic_visit
    value = self.visit(value)
  File "/home/greatx/repos/PaConvert/paconvert/../paconvert/base.py", line 86, in visit
    node = super(BaseTransformer, self).visit(node)
  File "/home/greatx/miniconda3/envs/jit-env/lib/python3.10/ast.py", line 418, in visit
    return visitor(node)
  File "/home/greatx/repos/PaConvert/paconvert/../paconvert/transformer/basic_transformer.py", line 666, in visit_Expr
    new_node = self.visit(old_value)
  File "/home/greatx/repos/PaConvert/paconvert/../paconvert/base.py", line 86, in visit
    node = super(BaseTransformer, self).visit(node)
  File "/home/greatx/miniconda3/envs/jit-env/lib/python3.10/ast.py", line 418, in visit
    return visitor(node)
  File "/home/greatx/repos/PaConvert/paconvert/../paconvert/transformer/basic_transformer.py", line 363, in visit_Call
    super(BasicTransformer, self).generic_visit(node)
  File "/home/greatx/miniconda3/envs/jit-env/lib/python3.10/ast.py", line 494, in generic_visit
    value = self.visit(value)
  File "/home/greatx/repos/PaConvert/paconvert/../paconvert/base.py", line 86, in visit
    node = super(BaseTransformer, self).visit(node)
  File "/home/greatx/miniconda3/envs/jit-env/lib/python3.10/ast.py", line 418, in visit
    return visitor(node)
  File "/home/greatx/repos/PaConvert/paconvert/../paconvert/transformer/basic_transformer.py", line 363, in visit_Call
    super(BasicTransformer, self).generic_visit(node)
  File "/home/greatx/miniconda3/envs/jit-env/lib/python3.10/ast.py", line 494, in generic_visit
    value = self.visit(value)
  File "/home/greatx/repos/PaConvert/paconvert/../paconvert/base.py", line 86, in visit
    node = super(BaseTransformer, self).visit(node)
  File "/home/greatx/miniconda3/envs/jit-env/lib/python3.10/ast.py", line 418, in visit
    return visitor(node)
  File "/home/greatx/repos/PaConvert/paconvert/../paconvert/transformer/basic_transformer.py", line 363, in visit_Call
    super(BasicTransformer, self).generic_visit(node)
  File "/home/greatx/miniconda3/envs/jit-env/lib/python3.10/ast.py", line 494, in generic_visit
    value = self.visit(value)
  File "/home/greatx/repos/PaConvert/paconvert/../paconvert/base.py", line 86, in visit
    node = super(BaseTransformer, self).visit(node)
  File "/home/greatx/miniconda3/envs/jit-env/lib/python3.10/ast.py", line 418, in visit
    return visitor(node)
  File "/home/greatx/miniconda3/envs/jit-env/lib/python3.10/ast.py", line 503, in generic_visit
    new_node = self.visit(old_value)
  File "/home/greatx/repos/PaConvert/paconvert/../paconvert/base.py", line 86, in visit
    node = super(BaseTransformer, self).visit(node)
  File "/home/greatx/miniconda3/envs/jit-env/lib/python3.10/ast.py", line 418, in visit
    return visitor(node)
  File "/home/greatx/repos/PaConvert/paconvert/../paconvert/transformer/basic_transformer.py", line 539, in visit_Call
    return self.trans_class_method(node, torch_class_api)
  File "/home/greatx/repos/PaConvert/paconvert/../paconvert/transformer/basic_transformer.py", line 556, in trans_class_method
    node_list = matcher.get_paddle_class_nodes(
  File "/home/greatx/repos/PaConvert/paconvert/../paconvert/base.py", line 508, in get_paddle_class_nodes
    self.parse_func(func)
  File "/home/greatx/repos/PaConvert/paconvert/../paconvert/base.py", line 390, in parse_func
    new_paddle_api = re.sub(
  File "/home/greatx/miniconda3/envs/jit-env/lib/python3.10/re.py", line 209, in sub
    return _compile(pattern, flags).sub(repl, string, count)
  File "/home/greatx/miniconda3/envs/jit-env/lib/python3.10/re.py", line 326, in _subx
    template = _compile_repl(template, pattern)
  File "/home/greatx/miniconda3/envs/jit-env/lib/python3.10/re.py", line 317, in _compile_repl
    return sre_parse.parse_template(repl, pattern)
  File "/home/greatx/miniconda3/envs/jit-env/lib/python3.10/sre_parse.py", line 1054, in parse_template
    raise s.error('bad escape %s' % this, len(this))
re.error: bad escape \u at position 26