作者|王益、严浩
翻译|程浩源、董文文

1

GDB Python3

PyTorch官网公布了如何应用GDB对Python触发的C++代码进行调试的指南,详情参考:
https://github.com/pytorch/py...

其外围思路是运行gdb python3。在GDB会话中,能够为给定的C++函数名设置断点,如at::Tensor::neg。GDB以后无奈找到这个函数,prompt中会提醒是否在共享库加载时将断点挂起,答复yes。而后输出run,GDB会启动Python解释器。Python解释器会提醒输出Python源码。输出import torch,而后回车。

当Python解释器执行import语句时,会加载相干的共享库。GDB会监督加载并设置断点。执行Python源码,触发断点,而后关上GDB prompt进行C++调试,例如应用 bt 查看回溯,应用 l 显示Python调用的C++代码。

2

在调试模式下编译OneFlow

Linux零碎

OneFlow 反对 Linux,暂不反对macOS和Windows。本文次要介绍在AWS GPU主机上运行Amazon Linux 2(相似于CentOS)。

(base) [wkyi ~]$ cat /etc/os-releaseNAME="Amazon Linux"VERSION="2"ID="amzn"ID_LIKE="centos rhel fedora"VERSION_ID="2"PRETTY_NAME="Amazon Linux 2"ANSI_COLOR="0;33"CPE_NAME="cpe:2.3:o:amazon:amazon_linux:2"HOME_URL="https://amazonlinux.com/"

Conda或Docker环境

OneFlow官网文档倡议应用Conda或Docker镜像:
https://github.com/Oneflow-In...。本次运行应用Anaconda。应用Conda或Docker是为了修复C++编译器和其余构建工具链的版本。应用新版本的g++须要对源代码进行更新,如
https://github.com/Oneflow-In...

编译调试版本

这里要留神,必须先编译OneFlow的调试版本,因为GDB须要调试符号能力使 btl 的输入有意义。

cd ~/w/oneflow/buildCMAKE_BUILD_TYPE=Debug cmake .. -C ../cmake/caches/international/cpu.cmake

我装置的是CPU版本的OneFlow,创立了cpu.cmake文件。因为我的 AWS 主机不在中国,所以是在international目录下创立文件。

报错

在装置报错时,我在GitHub上提交了相干issue
https://github.com/Oneflow-In...),OneFlow的研发人员疾速给出了回应,向他们致敬!

编译步骤

本大节将展现编译OneFlow的具体步骤:

  1. 下载安装Anaconda。默认装置门路是 ~/anaconda3。装置时将环境变量增加到~/.bashrc。而后,获取环境变量或从新连贯主机使更改失效。
  2. 创立并激活Conda环境,具体步骤参考:
    https://github.com/Oneflow-In...
  3. Git clone源码

mkdir ~/w cd ~/w git clone https://github.com/Oneflow-In...

  1. 编译OneFlow

cd oneflow mkdir build cd build `
CMAKE_BUILD_TYPE=Debug cmake .. -C ../cmake/caches/international/cpu.cmake
make -k -j $(nproc)`

运行和调试

装置好后,在~/w/oneflow/build 目录中会呈现 source.sh文件,这个文件设置了PYTHONPATH环境。运行下列命令使设置失效。

source source.sh

而后,用GDB运行Python编辑器。

gdb python3

在GDB prompt中,我在
oneflow::one::Tensor::is_eager设置了一个断点,在共享库加载时会将断点挂起。

(gdb) b oneflow::one::Tensor::is_eagerFunction "oneflow::one::Tensor::is_eager" not defined.Make breakpoint pending on future shared library load? (y or [n]) yBreakpoint 1 (oneflow::one::Tensor::is_eager) pending.

再而后,输出run运行Python编辑器。在Python prompt中,输出oneflow

(gdb) runStarting program: /home/wkyi/anaconda3/envs/oneflow-dev-gcc7-v2/bin/python3Missing separate debuginfos, use: debuginfo-install glibc-2.26-58.amzn2.x86_64[Thread debugging using libthread_db enabled]Using host libthread_db library "/lib64/libthread_db.so.1".Python 3.7.10 (default, Feb 26 2021, 18:47:35)[GCC 7.3.0] :: Anaconda, Inc. on linuxType "help", "copyright", "credits" or "license" for more information.>>> import oneflow

导入的工夫会比平时更长。如果显示ImportError,要查看是否运行了source source.sh

接下来就能够来创立tensor了。

a = oneflow.tensor(1)Thread 1 "python3" hit Breakpoint 1, oneflow::one::CopyBetweenMirroredTensorAndNumpy<long> (t=..., array=array@entry=0x7fffe5905150, Copy=<optimized out>,    Copy@entry=0x7fffefc977e0 <oneflow::BlobNumpyCopyUtil<long>::From(unsigned long, oneflow::NumPyArrayPtr const&)>, modifier=...,    block_host_until_done=block_host_until_done@entry=false) at /home/wkyi/w/oneflow/oneflow/api/python/utils/tensor_utils.h:9898      CHECK_OR_RETURN(tensor->is_eager()) << "eager tensors supported only.";

输出回车,此行代码执行会触发断点。

上述信息表明在源文件的第 98 行有一个名为tensor->is_eager()的函数 oneflow::CopyBetweenMirroredTensorAndNumpy

要显示更多内容,能够输出l。在第 98 行,调用了tensor->is_eager()

(gdb) l93    inline Maybe<void> CopyBetweenMirroredTensorAndNumpy(94        const std::shared_ptr<Tensor>& t, PyObject* array,95        Maybe<void> (*Copy)(uint64_t, const NumPyArrayPtr&), const std::string& modifier,96        bool block_host_until_done) {97      auto tensor = JUST(t->AsMirroredTensor());98      CHECK_OR_RETURN(tensor->is_eager()) << "eager tensors supported only.";99100      if (block_host_until_done) {101        NumPyArrayPtr array_ptr(array);102        const auto& Callback = [array_ptr, Copy](uint64_t ofblob_ptr) {

你可能会好奇,为什么在Python中创立tensor会触发对Tensor::is_eager的调用?能够输出bt来显示更多信息。

(gdb) bt#0  oneflow::one::CopyBetweenMirroredTensorAndNumpy<long> (t=..., array=array@entry=0x7fffe5905150, Copy=<optimized out>,    Copy@entry=0x7fffefc977e0 <oneflow::BlobNumpyCopyUtil<long>::From(unsigned long, oneflow::NumPyArrayPtr const&)>, modifier=...,    block_host_until_done=block_host_until_done@entry=false) at /home/wkyi/w/oneflow/oneflow/api/python/utils/tensor_utils.h:98#1  0x00007fffefd5aa5c in oneflow::one::CopyMirroredTensorFromUntypedArray<long> (array=0x7fffe5905150, tensor=...)    at /home/wkyi/w/oneflow/oneflow/api/python/utils/tensor_utils.cpp:61#13 0x00007fffefbe433f in oneflow::one::functional::tensor (self=<optimized out>, args=<optimized out>, kwargs=<optimized out>)    at /home/wkyi/w/oneflow/build/oneflow/api/python/functional/tensor_api.yaml.pybind.cpp:96#14 0x00005555556b98b4 in _PyMethodDef_RawFastCallKeywords () at /tmp/build/80754af9/python_1614362349910/work/Objects/call.c:693#15 0x00005555556b99d1 in _PyCFunction_FastCallKeywords (func=0x7ffdc75675a0, args=<optimized out>, nargs=<optimized out>, kwnames=<optimized out>)    at /tmp/build/80754af9/python_1614362349910/work/Objects/call.c:732#29 0x000055555578c22c in _Py_UnixMain () at /tmp/build/80754af9/python_1614362349910/work/Modules/main.c:3495#30 0x00007ffff783113a in __libc_start_main () from /lib64/libc.so.6#31 0x0000555555730e90 in _start () at ../sysdeps/x86_64/elf/start.S:103

调用堆栈的底部是_stack,它是Python编辑器的入口点。从下面的代码中能够看到Python和OneFlow共享库之间的调用边界——Python中的_PyMethodDef_RawFastCallKeywords函数调用了OneFlow的C++函数oneflow::one::functional::tensor,进而触发了对oneflow::one::Tensor::is_eager的调用

上述内容展现了如何应用GDB对C++编写的Python库进行调试。以下将更进一步,展现如何在VSCode中调用GDB,以便更轻松地定位代码。

要想在Python C++库中应用GDB,须要进行如下操作:

  1. 构建C++库的调试版本
  2. 用GDB启动Python解释器
  3. 在C++函数中设置断点,例如b
    oneflow::one::functional::tensor
  4. 在Python REPL中运行代码,触发断点,而后GDB会在断点处暂停

依照以下办法改良VSCode用户界面

  1. 装置Microsoft的“C/C++”扩大,该扩大反对GDB调试器
  2. 增加新的调试启动配置

  1. 增加下列配置
{   "version": "0.2.0",   "configurations": [       {          "type": "cppdbg",          "request": "launch",          "name": "GDB",          "program": "/home/charlieyan/anaconda3/envs/oneflow-dev-gcc7-v2/bin/python          "cwd": ".",          "environment": [              {                  "name": "PYTHONPATH",                  "value": "/home/charlieyan/proj/oneflow/python"              }          ]       }   ]   }
  1. 而后就能够启动调试器“GDB”并设置断点。当调试器在断点处暂停时,VSCode也会跳转到源代码行,侧边栏中的变量会与调用栈一起显示。要想更好地了解代码,也能够单步执行各个函数。

参考起源:

1.https://code.visualstudio.com...

2.https://of-worldwide.quip.com...

本文经受权后编译公布

欢送下载体验 OneFlow v0.7.0 最新版本:
https://github.com/Oneflow-In...