Atom pytracking 运行

1

subprocess.CalledProcessError: Command '['which', 'c++']' returned non-zero exit status 1

2

linux下   zip001 zip002 zip003 解压


 cat TRAIN_1.zip* > Tr1.zip

 unzip Tr1.zip -d ../Tracking/

3

Traceback (most recent call last):
  File "run_training.py", line 55, in <module>
    main()
  File "run_training.py", line 50, in main
    run_training(args.train_module, args.train_name, args.cudnn_benchmark)
  File "run_training.py", line 39, in run_training
    expr_func(settings)
  File "../ltr/train_settings/bbreg/atom_default.py", line 105, in run
    trainer.train(40, load_latest=True, fail_safe=False)
  File "../ltr/trainers/base_trainer.py", line 70, in train
    self.train_epoch()
  File "../ltr/trainers/ltr_trainer.py", line 76, in train_epoch
    self.cycle_dataset(loader)
  File "../ltr/trainers/ltr_trainer.py", line 57, in cycle_dataset
    loss, stats = self.actor(data)
  File "../ltr/actors/bbreg.py", line 17, in __call__
    iou_pred = self.net(data['train_images'], data['test_images'], data['train_anno'], data['test_proposals'])
  File "/home/robot/.conda/envs/tracking/lib/python3.7/site-packages/torch/nn/modules/module.py", line 550, in __call__
    result = self.forward(*input, **kwargs)
  File "../ltr/models/bbreg/atom.py", line 38, in forward
    train_feat = self.extract_backbone_features(train_imgs.view(-1, *train_imgs.shape[-3:]))
  File "../ltr/models/bbreg/atom.py", line 53, in extract_backbone_features
    return self.feature_extractor(im, layers)
  File "/home/robot/.conda/envs/tracking/lib/python3.7/site-packages/torch/nn/modules/module.py", line 550, in __call__
    result = self.forward(*input, **kwargs)
  File "../ltr/models/backbone/resnet.py", line 143, in forward
    x = self.bn1(x)
  File "/home/robot/.conda/envs/tracking/lib/python3.7/site-packages/torch/nn/modules/module.py", line 550, in __call__
    result = self.forward(*input, **kwargs)
  File "/home/robot/.conda/envs/tracking/lib/python3.7/site-packages/torch/nn/modules/batchnorm.py", line 106, in forward
    exponential_average_factor, self.eps)
  File "/home/robot/.conda/envs/tracking/lib/python3.7/site-packages/torch/nn/functional.py", line 1923, in batch_norm
    training, momentum, eps, torch.backends.cudnn.enabled
RuntimeError: cuDNN error: CUDNN_STATUS_EXECUTION_FAILED

4

OSError: [Errno 97] Address family not supported by protocol
Can someone tell me what this actually means? When I search via Google, all I can find are posts irrelevant to my problem or unanswered questions. From what I can gather, it is related to IPv6. I tried enabling IPv6 by adding the following to my systcl.conf file and then restarting the networking service.

我忘了怎么解决的了。。。。。。

net.ipv6.conf.default.disable_ipv6 = 0
net.ipv6.conf.all.disable_ipv6 = 0

5

6

Traceback (most recent call last):
  File "run_training.py", line 7, in <module>
    import torch.backends.cudnn
  File "/export/home/user2/.conda/envs/tracking/lib/python3.7/site-packages/torch/__init__.py", line 136, in <module>
    from torch._C import *
ImportError: /export/home/user2/.conda/envs/tracking/bin/../lib/libstdc++.so.6: version `CXXABI_1.3.11' not found (required by /export/home/user2/.conda/envs/tracking/lib/python3.7/site-packages/torch/lib/libtorch_python.so)
~                                                                                                                                                                         
~       
----------------------------------------------------------------------------                                                        
https://blog.csdn.net/Iv_zzy/article/details/109310965

7

subprocess.CalledProcessError: Command '['which', 'c++']' returned non-zero exit status 1.

gcc问题
https://www.zhihu.com/question/56272908
更换了gcc4.9 ，7 两个版本仍然有问题

(atom) [user2@aihpc 7.3.0]$ conda install gcc_linux-64

conda uninstall gcc

conda install cffi
conda install -c conda-forge gcc_impl_linux-64
ln -s /export/home/user2/.conda/envs/atom/libexec/gcc/x86_64-conda_cos6-linux-gnu/7.3.0/gcc /export/home/user2/.conda/envs/atom/bin/gcc

conda install gcc_linux-64


------------------2---------------------------------------------

conda install -c serge-sans-paille gcc_49
conda 下装好gcc-4.9的.so，gcc-4.9 -v 查看
创建软链接，把 gcc-4.9 链接到 gcc



(tra) [user2@aihpc bin]$ ln -s /export/home/user2/.conda/envs/tra/bin/gcc-4.9 /export/home/user2/.conda/envs/tra/bin/gcc
(tra) [user2@aihpc bin]$ ln -s /export/home/user2/.conda/envs/tra/bin/g++-4.9 /export/home/user2/.conda/envs/tra/bin/g++
(tra) [user2@aihpc bin]$ gcc -v
Using built-in specs.
COLLECT_GCC=gcc
COLLECT_LTO_WRAPPER=/state/partition1/home/user2/.conda/envs/tra/bin/../libexec/gcc/x86_64-unknown-linux-gnu/4.9.1/lto-wrapper
Target: x86_64-unknown-linux-gnu
Configured with: ./configure --prefix=/home/conda/anaconda/envs/_build --with-gmp=/home/conda/anaconda/envs/_build --with-mpfr=/home/conda/anaconda/envs/_build --with-mpc=/home/conda/anaconda/envs/_build --with-isl=/home/conda/anaconda/envs/_build --with-cloog=/home/conda/anaconda/envs/_build --disable-multilib --enable-languages=c,c++ --program-suffix=-4.9
Thread model: posix
gcc version 4.9.1 (GCC) 
(tra) [user2@aihpc bin]$ g++ -v
Using built-in specs.
COLLECT_GCC=g++
COLLECT_LTO_WRAPPER=/state/partition1/home/user2/.conda/envs/tra/bin/../libexec/gcc/x86_64-unknown-linux-gnu/4.9.1/lto-wrapper
Target: x86_64-unknown-linux-gnu
Configured with: ./configure --prefix=/home/conda/anaconda/envs/_build --with-gmp=/home/conda/anaconda/envs/_build --with-mpfr=/home/conda/anaconda/envs/_build --with-mpc=/home/conda/anaconda/envs/_build --with-isl=/home/conda/anaconda/envs/_build --with-cloog=/home/conda/anaconda/envs/_build --disable-multilib --enable-languages=c,c++ --program-suffix=-4.9
Thread model: posix
gcc version 4.9.1 (GCC) 

8.ImportError: /lib64/libstdc++.so.6: version `CXXABI_1.3.11’ not found

Traceback (most recent call last):
  File "run_training.py", line 7, in <module>
    import torch.backends.cudnn
  File "/export/home/user2/.conda/envs/tracking/lib/python3.7/site-packages/torch/__init__.py", line 136, in <module>
    from torch._C import *
ImportError: /lib64/libstdc++.so.6: version `CXXABI_1.3.11' not found (required by /export/home/user2/.conda/envs/tracking/lib/python3.7/site-packages/torch/lib/libtorch_python.so)
~                                                                                                                                                                                    
~                                

https://github.com/AllenDowney/ThinkStats2/issues/92
https://stackoverflow.com/questions/49875588/importerror-lib64-libstdc-so-6-version-cxxabi-1-3-9-not-found/49909230
https://blog.csdn.net/qingdu007/article/details/81515984
https://blog.csdn.net/JianJuly/article/details/99678608                                                                                                                                                    
~                                                   


执行以下命令，查找编译 gcc 时生成的最新动态库： 

find / -name "libstdc++.so.*"                                                                                                                                

9
AttributeError: ‘JPEG’ object has no attribute ‘decompressor’

https://blog.csdn.net/weixin_39386392/article/details/112556693?spm=1001.2014.3001.5501

1、libjpeg-turbo源码下载地址：https://sourceforge.net/projects/libjpeg-turbo/files/，选择下载libjpeg-turbo-2.0.6.tar.gz
2、下载源码后，在服务器原目录解压压缩文件，并在源码目录新建一个build-arm的目录。
    进入build-arm目录，执行以下命令，生成Makefile。
    cmake -G"Unix Makefiles"  -DCMAKE_INSTALL_PREFIX=/home/user/libjpeg-turbo/build-arm/libs-build /home/user/libjpeg-turbo
3、执行make -j 8 操作，完成编译。
4、执行make install 操作，完成安装。

pip install jpeg4py==0.1.4
/home/user/anaconda3/envs/pytorch/lib/python3.6/site-packages/jpeg4py/_cffi.py

将此文件中第198行中的

"libturbojpeg.so.0",  #for Ubuntu

修改为

"/home/user/libjpeg-turbo/build-arm/libs-build/lib/libturbojpeg.so.0",  #for Ubuntu

RuntimeError: Error building extension ‘_prroi_pooling’