PaddleVideo segmentation fault

Describe the bug

[10/23 16:37:28] DALI is not installed, you can improve performance if use DALI
Inference models that Paddle provides are listed as follows:

{'ppTSM', 'TSM', 'TSN'}

Inference models that Paddle provides are listed as follows:

{'ppTSM', 'TSM', 'TSN'}

process params are as follows:
Namespace(batch_size=1, enable_mkldnn=False, gpu_mem=8000, ir_optim=True, label_name_path='/home/kula/.local/lib/python3.8/site-packages/ppvideo/tools/../data/k400/Kinetics-400_label_list.txt', model_file='/home/kula/.paddlevideo_inference/inference_model/ppTSM/inference.pdmodel', model_name='ppTSM', normalize=True, num_seg=8, params_file='/home/kula/.paddlevideo_inference/inference_model/ppTSM/inference.pdiparams', seg_len=1, short_size=256, target_size=224, top_k=1, use_fp16=False, use_gpu=True, use_tensorrt=False, video_file='/mnt/d/tiktok/20211023/jp.kbiz.shop/7016678652104543494.mp4')
I1023 16:37:28.573222 30338 analysis_config.cc:424] use_dlnne_:0
I1023 16:37:28.573268 30338 analysis_config.cc:424] use_dlnne_:0
I1023 16:37:28.573279 30338 analysis_config.cc:424] use_dlnne_:0
I1023 16:37:28.573285 30338 analysis_config.cc:424] use_dlnne_:0
**********/mnt/d/tiktok/20211023/jp.kbiz.shop/7016678652104543494.mp4**********


--------------------------------------
C++ Traceback (most recent call last):
--------------------------------------
0   paddle::framework::SignalHandle(char const*, int)
1   paddle::platform::GetCurrentTraceBackString[abi:cxx11]()

----------------------
Error Message Summary:
----------------------
FatalError: `Segmentation fault` is detected by the operating system.
  [TimeInfo: *** Aborted at 1634978252 (unix time) try "date -d @1634978252" if you are using GNU date ***]
  [SignalInfo: *** SIGSEGV (@0x0) received by PID 30338 (TID 0x7f33208ba740) from PID 0 ***]

[1]    30338 segmentation fault  ppvideo --model_name='ppTSM'

To Reproduce

ppvideo --model_name='ppTSM' --video_file='/mnt/d/tiktok/20211023/jp.kbiz.shop/7016678652104543494.mp4'

Expected behavior A clear and concise description of what you expected to happen.

Screenshots If applicable, add screenshots to help explain your problem.

Desktop (please complete the following information):

OS: [e.g. iOS]
Browser [e.g. chrome, safari]
Version [e.g. 22]

Smartphone (please complete the following information):

ubuntu20.04

Additional context

➜  PaddleVideo git:(develop) ✗ python3 -c "import paddle; print(paddle.__version__)"
2.1.0

Oct 23 '21 08:10 kulame

我试着跑了hub里的videotag_tsn_lstm 也是同样的报错

➜  text hub install videotag_tsn_lstm==1.0.0
Download https://bj.bcebos.com/paddlehub/paddlehub_dev/videotag_tsn_lstm_1.0.0.tar.gz
[##################################################] 100.00%
Decompress /home/kula/.paddlehub/tmp/tmp_5ngb47b/videotag_tsn_lstm_1.0.0.tar.gz
[##################################################] 100.00%
[2021-10-25 11:54:31,951] [    INFO] - Successfully installed videotag_tsn_lstm-1.0.0
➜  text hub run videotag_tsn_lstm  --input_path /mnt/d/tiktok/20211025/us.kbiz.shop/6987961095747063046.mp4 --use_gpu True
[2021-10-25 11:55:27,406] [ WARNING] - The _initialize method in HubModule will soon be deprecated, you can use the __init__() to handle the initialization of the object
W1025 11:55:27.920683  5109 device_context.cc:404] Please NOTE: device: 0, GPU Compute Capability: 7.5, Driver API Version: 11.4, Runtime API Version: 10.1
W1025 11:55:27.931650  5109 device_context.cc:422] device: 0, cuDNN Version: 8.2.


--------------------------------------
C++ Traceback (most recent call last):
--------------------------------------
0   paddle::framework::SignalHandle(char const*, int)
1   paddle::platform::GetCurrentTraceBackString[abi:cxx11]()

----------------------
Error Message Summary:
----------------------
FatalError: `Segmentation fault` is detected by the operating system.
  [TimeInfo: *** Aborted at 1635134146 (unix time) try "date -d @1635134146" if you are using GNU date ***]
  [SignalInfo: *** SIGSEGV (@0x0) received by PID 5109 (TID 0x7f05cbff3740) from PID 0 ***]

[1]    5109 segmentation fault  hub run videotag_tsn_lstm --input_path  --use_gpu True

Oct 25 '21 03:10 kulame

把gpu关掉就可以了，是因为这个模型和cuda11.4不兼容吗?

Oct 25 '21 03:10 kulame

你好，CUDA10.2环境下经测试是可以正常运行打印结果的，可以尝试将CUDA和CUDNN从11.4降级到10.2和对应版本再试试？

Oct 25 '21 07:10 HydrogenSulfate

你好，我在460驱动+cuda11.2环境下测试正常可以检查一下你的paddle版本是否安装正确，目前paddlepaddle官方最高支持到cuda11.2，可以尝试安装cuda 11.2版的paddlepaddle-gpu 2.1.3，然后看看是否还存在这个问题

python3.7 -m pip install paddlepaddle-gpu==2.1.3.post112 -f https://www.paddlepaddle.org.cn/whl/linux/mkl/avx/stable.html

在11.4的环境下测试也是正常的，paddlepaddle-gpu的版本是 cuda11.2 develop

python3.7 -m pip install paddlepaddle-gpu==0.0.0.post112 -f https://www.paddlepaddle.org.cn/whl/linux/gpu/develop.html

Please NOTE: device: 0, GPU Compute Capability: 7.0, Driver API Version: 11.4, Runtime API Version: 11.2
device: 0, cuDNN Version: 8.2.
Current video file: data/example.avi
        top-1 class: 5
        top-1 score: 0.990738570690155

Nov 01 '21 04:11 HydrogenSulfate