search

kevaldoshi17 / Prediction-based-Video-Anomaly-Detection-

MIT License
17 stars 10 forks source link

Setup specifications #2

Open mijalapenos opened 3 years ago

mijalapenos commented 3 years ago

Hello,

I am trying to run the training from scratch with the following command:

python train.py --dataset ped2 --train_folder ../dataset/ped2/training/frames --test_folder ../dataset/ped2/testing/frames --gpu 0 --iters 80000

I downloaded the ped2 dataset, hyper_params.ini file and even the flownet2 sources and pre-trained models to be loaded from here. However the code always fails. Based on my research it is due to incompatibility of Tensorflow 1.4.1 with my CUDA version. It seems like I would need CUDA 8 which is no longer available to install from what I've found. Do you plan to update the code to a more recent Tensorflow version? Or could you provide me with the full specifications of your setup?

Thank you very much!

kevaldoshi17 commented 3 years ago

Hi,

Sorry for the difficulties running the code. The full specifications required for running the code are already available in the readme file. You can always create a virtual environment and install those without removing your existing installation.

mijalapenos commented 3 years ago

Thank you for your reply. Can I ask what CUDA version are you running? I installed the packages according to the specifications in a Conda environment with the exception of CUDA, because there is no version mentioned.

kevaldoshi17 commented 3 years ago

Sorry for the late reply, in my implementation, the version of CUDA is 10.0.130 and for CUDNN is 7.6.5. But this is also dependant on the GPU that you are using, mine is RTX 2070.