isosnovik / nn4topopt Goto Github PK

I am very sorry to bother you again. I would like to know how to test the network and how to get the data value or graph of IoU.

As shown in the above figure, how do you get the input image, input gradient, prediction and ground truth in the code？

I am currently a student working on a project in the field of Topology Optimization.As a first step in understanding the process, I am trying to run and reproduce some of the results you have got from the paper. However, I run into a lot of technical errors while running your GitHub codes on Jupyter Lab that trace back to errors in compatibility between TensorFlow and Keras versions.

Here is my technical stack information:

1. Currently using a Linux-based machine with docker image of TensorFlow and Keras as it is pre-installed. I have tried installing specific versions of known compatible combinations of both the software (eg. TensorFlow v.2.0 and Keras v.2.3 etc. ) but it does not work.
2. Currently using the NIVIDIA DGX GPU cluster machine for training the dataset.

I have used the available TensorFlow image from the NVIDIA NGC catalogue to build my container and have tried testing your code using Jupyter Lab. However, I still run into a bunch of compatibility issues.

Could you please let me know the exact TensorFlow, Keras and Python versions that you used at the time of training in your project? That would then help me run the code and understand your results better.

          Hello,

It is a Windows-specific error related to the read/write rights. I don't work with Windows, so I don't know the exact solution.
Maybe you need to add permissions to prepare_data.py

Originally posted by @ISosnovik in #2 (comment)

hello,Professior!
I have a problem in the first step--to generate the .h5 dataset. To solve it , I have checked File System Permissions and ran as Administrator. It also can't be solved. Can you give me other ways to solve it?The error picture as the follow picture: