Closed pescap closed 1 month ago
Hi @pescap ,
Thanks for raising a request for the enhancement.
Yes as you have mentioned libgl1-mesa-glx
library will be needed as additional installation, which was mentioned in the installation section of the documentation Link. However, in our tests we did not find the need to install other libraries, which were mentioned above. If you could provide us an example where libgl1-mesa-glx
is not sufficient, then we can go ahead and add the additional libraries to the documentation page.
Currently the code is tested upto tensorflow version 2.13 (as observed from the requirements.txt file). . Depending on your OS, cuDNN, and CUDA version, any version of TensorFlow from 2.9 to 2.13 can be used.
As suggested, I have created a new docker image at the url FastPINNs-Docker Link,You can pull the image using the command:
docker pull thivinanandh/fastvpinns:latest
.
The docker image will have the welcome message, which displays the cuda driver, gpu and tensorflow version
The Docker image is built on Ubuntu 20.04, with CUDA version 11.2 and cuDNN version 8.1, and it includes TensorFlow version 2.11 (since tensorflow 2.11 is compiled with cuDNN 8.1). The additional documentation on how to use the docker image is available at the FastVPINNs-Docker Documentation Page
Please let me know if you need any additional information regarding the same
Dear all, great!
When I run the container
docker run --gpus all -it --rm -v ~/fastvpinns:/fastvpinns thivinanandh/fastvpinns:latest
I get:
Traceback (most recent call last):
File "/docker_initialise.py", line 85, in <module>
main()
File "/docker_initialise.py", line 50, in main
version = get_version_from_toml()
File "/docker_initialise.py", line 6, in get_version_from_toml
with open("pyproject.toml", "r") as file:
FileNotFoundError: [Errno 2] No such file or directory: 'pyproject.toml'
Hi @pescap ,
Bug: [Errno 2] No such file or directory: 'pyproject.toml'
Root Cause : It seems that in the Dockerfile
i have used a git clone command to pull the codes from the github repository to get the updated codes and files. However , during my build, docker seems to have used a cached version of the git clone
command and the recent changes made to the codes were not reflected in the local build.
Fix: The issue is fixed now, by running the build using --no-cache flag
.
Could you kindly try once again to see if the issue persists or not.
Hi @pescap ,
Bug:
[Errno 2] No such file or directory: 'pyproject.toml'
Root Cause : It seems that in the
Dockerfile
i have used a git clone command to pull the codes from the github repository to get the updated codes and files. However , during my build, docker seems to have used a cached version of thegit clone
command and the recent changes made to the codes were not reflected in the local build.Fix: The issue is fixed now, by running the build using
--no-cache flag
.Could you kindly try once again to see if the issue persists or not.
Hi, has the Docker image been updated here? (with docker push)
@pescap , yes it has been updated with docker push. The Current image on dockerhub is the updated image
Thanks
Hi, I am still having the same error.
Hi @pescap ,
Sorry, I was attending a conference past week thats why i was not able to reply back in time. Please find below the fix for the issue reported with docker.
Issue
File not found while trying to launch the docker terminal
[Errno 2] No such file or directory: 'pyproject.toml'
Root Cause
It seems there was no issue with the docker build, rather the issue was with the usage of -v
flag which we used to mount an external volume. When we use -v ~/fastvpinns:/fastvpinns
, we were mounting an empty directory from a host named ~/fastvpinns
to the actual directory within the docker container /fastvpinns
, which contains the code base. So when this command is executed, the empty directory overwrites the core directory /fastvpinns
within our docker environment causing the docker build to look for a file, which has been deleted now due to the mounting.
Fix
The fix is to mount an external directory from host into a new directory such as /fastvpinns/output
within the docker environment, so that it does not overwrite the existing codes. Updated command for this is as follows,
docker run --gpus all -it --rm -v ~/fastvpinns_docker_output:/fastvpinns/output thivinanandh/fastvpinns:latest
Additional Update
The explanation for the same is now added to the docker documentation page of FastVPINNs and the running instructions are updated accordingly. Docker Documentation - FastVPINNs
please pull the new version of the docker image for testing
@pescap , let me know if this fixes the issue, so that i can close this issue
Please let me know if you need any additional information regarding the same.
Fixed, thank you!
Dear all, I successfully installed the library from the DeepXde Docker image, which includes TensorFlow 2.9.2.
For your information, I had to further install the following libraries:
Is it a known behavior?
It could be beneficial to build a dedicated image for fastvpinns, for reproducibility purposes. I could help you with this.
Finally, do you recommend any version of TensorFlow? Best,