11-03-2020 03:17 PM
I am just met an error output when i try to use "./docker_run.sh xilinx/vitis-ai-gpu:latest" to boot docker environment for Vitis AI in GPU
but i get an error like this
after i try it again by running "./docker_run.sh xilinx/vitis-ai-gpu:latest" under "su root", but the error is the same like above
if there has somebody who share the same experience with me? I can make sure my network is ok
waiting for response online, thanks
11-03-2020 06:16 PM
You need to build the GPU docker image before run the docker.
cd ./docker ./docker_build_gpu.sh
11-04-2020 01:11 AM
thanks for your response
actually, I have run these commands yet
cd ./docker ./docker_build_gpu.sh
but I can not make sure if the installation is success or not.
i have run both "./docker_build_gpu.sh" and "./docker_build_cpu.sh", I have run both of them under a "su root", after I running "./docker_build_cpu.sh", the very end is just like this
because it return a non-zero code:4, so I believe there have some errors. but because there has no red highlights, so I can not find where the errors come from.
and the output I got by running a "./docker_build_gpu.sh" is similar to the output above, for make sure of it, I will run "./docker_build_gpu.sh" again at night and send screenshot to you.
could you please provide some additional suggestion for me?
thanks a lot:)
11-04-2020 07:42 AM
thanks for your instructions, i am just input what you suggested and get output like this
per this output, if i can assume that my "./docker_build_cpu.sh" is success and "./docker_build_gpu.sh" is failure?
11-04-2020 08:05 AM
I don't think either succeeded. The two <none> entries are probably the failed builds. Part of the reason I suspect that is due to the CREATED dates. The xilinx/vitis-ai-cpu:latest image was created 7 months ago. If you built it using the docker_build_cpu.sh script then it would show a newer CREATED date.
Additionally, the xilinx/vitis-ai-cpu:latest image on your machine has an IMAGE ID of 798f6eaea389, which is the same as the prebuilt 1.1.56 vitis-ai image that corresponds to Vitis-AI v1.1. The image with the a7eb601784e9 tag corresponds to the 1.2.82 prebuilt image.
11-04-2020 03:04 PM
for fixing problems mentioned above, I think i need run "./docker_build_cpu.sh" and "./docker_build_gpu.sh" again until the two <none> entries disappeared, is that right?
and in addition, my current Vitis AI package's edition is 1.2.1
11-04-2020 06:13 PM
You may also try to remove the seeming incomplete images and rebuild the gpu docker image from scratch.
docker image rm <image ID>
11-04-2020 06:36 PM
There must be some errors when you run docker_build_gpu.sh. I believe you need to check the error information print in the terminal.
You need to provide more information to let me know what's wrong, like you install steps, did you add related repository and key?
You need read https://docs.docker.com/engine/install/
hope it can help you!
Don't forget to reply, kudo, and accept as solution.
11-04-2020 07:03 PM
Since you got error no 4 (Interrupt system call), docker couldn't build image.
In this case, I guess your network was unstable. So, wget couldn't get proper image from Internet.
Would you make sure your network ?
As you know, docker puts a heavy load onto network during build flow with Dockerfile.
So you might retry this flow until success.
11-04-2020 07:04 PM
thanks for your response, i think my docker is ok, because i can run "docker run hello-world" without sudo
and i just try to run "./docker_build_gpu.sh" under a "su root" again, in the very end i get error output like this
so how can i deal with it now, appreciating for any help, thanks a lot
11-04-2020 07:11 PM
>and i just try to run "./docker_build_gpu.sh" under a "su root" again, in the very end i get error output like this
It seems proxy issue.
Would you make sure your proxy setting or your network ?
11-04-2020 07:13 PM
I also think sometimes there have some unstable effects on my network, so i just try to run "./docker_build_gpu.sh" under a "su root" again, in the very end i get error output like this
there still have some problems, appreciating for any support, thanks a lot
11-04-2020 07:14 PM
11-04-2020 07:22 PM
currently i am in china, sometimes i will open my VPN to download some resources, but i just run "./docker_build_gpu.sh" under a "su root" in a VPN disconnected environment.
And in addition, if there is a VPN connected environment, i can open google or youtube quickly, but the running of "./docker_build_gpu.sh" can not access to the network, the screenshot is like this
11-04-2020 07:47 PM
i have another question for you.
because i find that i can run "./docker_run.sh xilinx/vitis-ai-cpu:latest" successfully without sudo like this
if that means my running of "./docker_build_cpu.sh" is successful? if it is successful, why there have two <none> here(just under REPOSITORY and above docker)
just waiting for your helps, thanks
11-04-2020 08:07 PM
followed the question for watari above, i have a confusion for this screenshot
there are information like this
Docker Image Version: latest
Build Date: Wed Nov 4 20:36:57 MST 2020
For TensorFlow Workflows do:
conda activate vitis-ai-tensorflow
For Caffe Workflows do:
conda activate vitis-ai-caffe
For Neptune Workflows do:
conda activate vitis-ai-neptune
More detail on conda packages included in container: /opt/vitis_ai/conda/conda_packages.txt
More detail on other 3rd party package source included in container: https://www.xilinx.com/products/design-tools/guest-resources.html
As the bold part indicated, there should have a vitis_ai folder under the folder of opt,
but actually, there has no vitis_ai in opt
just wait for your response, thanks a lot
11-04-2020 08:43 PM
>because i find that i can run "./docker_run.sh xilinx/vitis-ai-cpu:latest" successfully without sudo like this
If your account have same group id or same privilege, you can launch docker command without root privilege.
Would you make sure it in /etc/passwd and/or /etc/group, if you want to know details ?
>if that means my running of "./docker_build_cpu.sh" is successful? if it is successful, why there have two <none> here(just under REPOSITORY and above docker)
Because of lack of tag by ex. your flow or mistake or wrong tag, you can see <none> on "docker image ls".
Would you try to login with /bin/bash by docker command on <none>'s docker container image, if you want to know detail ?
11-04-2020 08:53 PM
Thanks for your reply,
For your another problem, you want know where is the /opt/vitis_ai?
Actually, It is the directory in docker image, you can run docker.sh <>, and print command "cd var/vitis_ai" to find it,
but totally, I think it is not necessary
Please mark the Answer as "Accept as solution" if the information provided is helpful.
Give Kudos to a post which you think is helpful and reply oriented.
11-04-2020 10:11 PM
as this part
i remembered that i use the shells shown below to make sure running "docker run hello-world" without sudo
that is also come from the instructions of tutorial shown below
and in the files of /etc/passwd and /etc/group i can find names just like "root" or "shixian".
11-04-2020 10:26 PM
and as this part
please provide some more details about "lack of tag by ex", i think i can not fully understand it.
just wait for your response, thanks a lot
11-05-2020 02:53 PM
>please provide some more details about "lack of tag by ex", i think i can not fully understand it.
As I already mentioned before, if your flow is wrong or mistake or you attach wrong tag in procedure, you can see <none> on "docker image ls".
# Procedure (a part of example)
1. delete tag information in docker image files before building procudure, if there is already same tag.
2. build docker container with Dockerfile.
3. If building procedure is failed, you can see <none> on "docker image ls".
Hope your helps.