Digits, model fail

Following details from :
https://docs.nvidia.com/deeplearning/digits/digits-user-guide/index.html

I’ve imported the MNIST dataset without problems. Setting up the model, however, raises the following error:

conv1 -> conv1
Check failed: status == CUDNN_STATUS_SUCCESS (6 vs. 0) CUDNN_STATUS_ARCH_MISMATCH, device 0

Within the docker image, the GPU seems available:
Sun Jul 8 12:14:24 2018
±----------------------------------------------------------------------------+
| NVIDIA-SMI 390.30 Driver Version: 390.30 |
|-------------------------------±---------------------±---------------------+
| GPU Name Persistence-M| Bus-Id Disp.A | Volatile Uncorr. ECC |
| Fan Temp Perf Pwr:Usage/Cap| Memory-Usage | GPU-Util Compute M. |
|===============================+======================+======================|
| 0 Quadro 4000 Off | 00000000:03:00.0 On | N/A |
| 40% 71C P12 N/A / N/A | 233MiB / 1980MiB | 0% Default |
±------------------------------±---------------------±---------------------+

±----------------------------------------------------------------------------+
| Processes: GPU Memory |
| GPU PID Type Process name Usage |
|=============================================================================|
±----------------------------------------------------------------------------+

Any thoughts ?
Regards
Pete

Quadro 4000 is Fermi architecture. During container image build process, we remove support on that architecture. That’s why you saw ARCH_MISMATCH status.