Topaz train error in CSparc 3.2

Troubleshooting
1

Hi,

When I run Topaz train in v3.2, after running apparently normally for some time it terminates with the attached error. This is on CentOS 7, with RTX-3090 cards, CUDA 11.2, Topaz 0.2.4. Thoughts?

Cheers
Oli

image
image2098Ă—1120 505 KB

2

Hi @olibclarke,

There seems to be a conflict between the last Topaz and Cryosparc v3.2. You can try to install another Topaz. Have a look to this thread…
cryoSPARC v.3.1.0 and Topaz

Best,

Juan

3

Thanks Juan - but this doesn’t seem to be the same error, and I am using the version of Topaz that is apparently ok per that thread… I am suspecting that maybe it has something to do with these new cards, which require CUDA 11.1, but not entirely sure

Cheers
Oli

4

Hi Oli,

I just tried it on our Cryosparc 3.2.0 with Topaz 0.2.3 and CUDA 10.1 + RTX 5000 and it worked fine. My guess is Topaz and Cuda 11.1 are not compatible yet. I will make an issue on the Topaz github. Thanks!

Best,
-Alex

1 Like
5

Hi @olibclarke, @alexjamesnoble,

I just tested this on our machine with a 3090 on cryoSPARC v3.2, and I was able to get the job to complete successfully:

platform_release: "5.4.0-65-generic"
platform_version: "#73~18.04.1-Ubuntu SMP Tue Jan 19 09:02:24 UTC 2021"
platform_architecture: "x86_64"
name: "GeForce RTX 3090"
CUDA_version: "11.1.0" # this is the version of CUDA that pyCUDA was built with

nvidia-smi:

Thu Apr  8 15:15:40 2021
+-----------------------------------------------------------------------------+
| NVIDIA-SMI 460.32.03    Driver Version: 460.32.03    CUDA Version: 11.2     |
|-------------------------------+----------------------+----------------------+
| GPU  Name        Persistence-M| Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |
|                               |                      |               MIG M. |
|===============================+======================+======================|
|   0  GeForce RTX 3090    Off  | 00000000:0A:00.0 Off |                  N/A |
|  0%   49C    P8    18W / 350W |      2MiB / 24268MiB |      0%   E. Process |
|                               |                      |                  N/A |
+-------------------------------+----------------------+----------------------+
|   1  GeForce RTX 3090    Off  | 00000000:42:00.0 Off |                  N/A |
|  0%   47C    P8    23W / 350W |      2MiB / 24265MiB |      0%   E. Process |
|                               |                      |                  N/A |
+-------------------------------+----------------------+----------------------+

+-----------------------------------------------------------------------------+
| Processes:                                                                  |
|  GPU   GI   CI        PID   Type   Process name                  GPU Memory |
|        ID   ID                                                   Usage      |
|=============================================================================|
|  No running processes found                                                 |
+-----------------------------------------------------------------------------+

One difference I made was to install pyTorch 1.8.1 with CUDA Toolkit 11.1:
conda activate topaz
conda install pytorch cudatoolkit=11.1 -c pytorch -c conda-forge

4 Likes
6

Thanks @stephan, we will give this a go!

7

Thanks for checking Stephan! We closed the Topaz issue. Can this be added to the Cryosparc Topaz installation recommendations?

-Alex

1 Like
8

Yes thanks @stephan - after doing that it works on our system too

1 Like
9

Definitely, done!

https://guide.cryosparc.com/processing-data/all-job-types-in-cryosparc/deep-picking/topaz#python-environment

2 Likes
10

Hi Stephan,
does Topaz now work with cryoSPARC 3.2.0 without deactivating the cryoSPARC anaconda environment, as described here?
Cheers,
Dirk.