isonet troubleshooting advice
120 views
Skip to first unread message
Oleksiy Kovtun
Feb 28, 2024, 5:37:34 AM2/28/24
to IsoNet
Dear colleagues,
IsoNet installation on a workstation (ubuntu 20.04 kernel 5.14, RTX3090 under 495 driver, 11.5 cuda) completes test_tf.py but crushes upon refine, see below. I have attached files with system config and installation notes. I will be very grateful for any advice on how to troubleshoot.
Thank you and best,
Oleksiy
>>isonet.py refine subtomo.star --gpuID 0 --result_dir t25_results
02-28 10:54:14, INFO
######Isonet starts refining######
02-28 10:54:15, INFO Start Iteration1!
02-28 10:54:19, INFO Noise Level:0.0
^T02-28 10:54:57, INFO Done preparing subtomograms!
02-28 10:54:57, INFO Start training!
02-28 10:54:58, INFO Loaded model from disk
02-28 10:54:58, INFO begin fitting
Epoch 1/10
2024-02-28 10:55:04.172548: E tensorflow/compiler/xla/stream_executor/cuda/cuda_dnn.cc:433] Could not create cudnn handle: CUDNN_STATUS_NOT_INITIALIZED
2024-02-28 10:55:04.172633: E tensorflow/compiler/xla/stream_executor/cuda/cuda_dnn.cc:442] Possibly insufficient driver version: 495.29.5
02-28 10:54:14, INFO
######Isonet starts refining######
02-28 10:54:15, INFO Start Iteration1!
02-28 10:54:19, INFO Noise Level:0.0
^T02-28 10:54:57, INFO Done preparing subtomograms!
02-28 10:54:57, INFO Start training!
02-28 10:54:58, INFO Loaded model from disk
02-28 10:54:58, INFO begin fitting
Epoch 1/10
2024-02-28 10:55:04.172548: E tensorflow/compiler/xla/stream_executor/cuda/cuda_dnn.cc:433] Could not create cudnn handle: CUDNN_STATUS_NOT_INITIALIZED
2024-02-28 10:55:04.172633: E tensorflow/compiler/xla/stream_executor/cuda/cuda_dnn.cc:442] Possibly insufficient driver version: 495.29.5
Oleksiy Kovtun
Feb 28, 2024, 10:21:47 AM2/28/24
to IsoNet
Dear all.
We found the fix.
Thank you.
Oleksiy
Reply all
Reply to author
Forward
0 new messages