tensorQTL using CPU on a GPU host

[broadinstitute]

tensorQTL using CPU on a GPU host

[francois-a]

Hi,

Have you followed the installation instructions listed at https://github.com/broadinstitute/tensorqtl/tree/master/install? Recent versions of tensorQTL use PyTorch instead of TensorFlow. What output do you get from nvidia-smi?

[Toseph]

This is my nvidia-smi output, I usually rely on environment modules to load cuda to the appropriate env variables.

`[thasan@pe2dg2-01 ~]$ nvidia-smi
Wed Nov 6 13:06:27 2019
+-----------------------------------------------------------------------------+
| NVIDIA-SMI 410.48 Driver Version: 410.48 |
|-------------------------------+----------------------+----------------------+
| GPU Name Persistence-M| Bus-Id Disp.A | Volatile Uncorr. ECC |
| Fan Temp Perf Pwr:Usage/Cap| Memory-Usage | GPU-Util Compute M. |
|===============================+======================+======================|
| 0 Tesla V100-PCIE... Off | 00000000:3B:00.0 Off | 0 |
| N/A 31C P0 25W / 250W | 0MiB / 16130MiB | 0% Default |
+-------------------------------+----------------------+----------------------+
| 1 Tesla V100-PCIE... Off | 00000000:D8:00.0 Off | 0 |
| N/A 33C P0 25W / 250W | 0MiB / 16130MiB | 0% Default |
+-------------------------------+----------------------+----------------------+

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

[Toseph]

I'll try installing pytorch against the python install and see if that works and use the install guide linked above.

[Toseph]

Updating the machine to the latest drivers allowed tensorqtl to successfully run the test bench using the gpus. We noted that cuda 10.0 and cuda 10.1 both work against this driver and with tensorqtl.

| NVIDIA-SMI 418.87.01 Driver Version: 418.87.01 CUDA Version: 10.1 |

[francois-a]

Great, thanks for the update!