Quote from: papampi on November 08, 2017, 07:19:28 PM

Quote from: leenoox on November 08, 2017, 06:53:00 PM

Quote from: papampi on November 01, 2017, 06:20:57 PM

Quote from: hurvajs77 on November 01, 2017, 05:45:03 PM

@damNmad:

I found a small bug in your telegram configuration. Command which use for getting GPUs count is invalid when rig have installed over 9 cards...

Your command:

Code:

GPU_COUNT=$(nvidia-smi -L | tail -n 1| cut -c 5 |awk '{ SUM += $1+1} ; { print SUM }')

and result:

Code:

m1@rig-bafomet:~$ nvidia-smi -L | tail -n 1| cut -c 5 |awk '{ SUM += $1+1} ; { print SUM }'
2

There isn't any reason, why use awk. Easier way is use wc only.

So, fix:

Code:

GPU_COUNT=$(nvidia-smi -L | wc -l')

and result:

Code:

m1@rig-bafomet:~$ nvidia-smi -L | wc -l
11

I used to use same cmd, fixed it with this one for 1.4:

Code:

nvidia-smi --query-gpu=count --format=csv,noheader,nounits | tail -1

Here is even more optimized code, less cycles, no need to use pipe:

Code:

nvidia-smi -i 0 --query-gpu=count --format=csv,noheader,nounits

every one has GPU0 plugged in so we can query only one GPU to get the total number of GPU's instead of querying all GPU's to return the same number then pipe it trough tail Wink

Totally correct and thanks for the help 👍👍👍

I've never understood why in this query every GPU shows the total number of GPUs instead of just the number of GPUs.!!!