EliasRas opened a new issue #20531:
URL: https://github.com/apache/incubator-mxnet/issues/20531
## Description
Program gets stuck when trying to call "mx.ndarray.ones" on a NVIDIA Turing
GPU if mxnet version is >= 1.6.0. Program does not crash and GPU Memory usage
keeps increasing.
Everything works if I use Pascal GPU (Quadro P400) or CPU. Everything works
if I use mxnet 1.5.0.
### Error Message
No error message is given.
## To Reproduce
```
import os
with os.add_dll_directory(os.environ["CUDA_PATH"]):
import mxnet as mx
testArray = mx.ndarray.ones((1,1),ctx=mx.gpu(1))
```
### Steps to reproduce
1. Install CUDA, Python
2. Install mxnet + dependencies using pip
3. Run the script using IDLE
## What have you tried to solve it?
1. Reinstalled CUDA, Python and mxnet
2. Tried various versions of CUDA and mxnet
3. Tried another version of cuDNN
4. Upgraded graphics drivers
5. Tried using the GPUs both in TCC mode and in WDDM mode
## Environment
- Windows Server 2019 Essentials
- Python 3.9.5
- pip 21.1.1
I've tried
- Quadro RTX 4000 and Tesla T4
- NVIDIA Graphics driver 441.66 and 471.11
- CUDA 9.2.148 and 10.2.86
- cuDNN v.7.6.5.32 and v8.2.2.26
- mxnet-cu92 1.6.0
- mxnet-cu102 1.6.0, 1.7.0 and 2.0.0b20201108
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
