All measurements timing out with measure_programs.py

[sbird-kick]

Hi,

I have a Ubuntu 20.04 system with an rtx 2080ti GPU.

I have successfully built tvm (from this repo) and created the "build-tvm" enviroment. I am trying to gather new measurements on for my GPU. I downloaded the gpu dataset and linked it (ln -s dataset_gpu dataset) as asked.

However, when I try to run the measure_programs.py script, all the measurements time out.

(tvm-build) (anaconda3) root@fd589e5b9041:~/tenset/scripts# python3 measure_programs.py --target nv
idia/geforce-rtx-2080-ti --batch-size 128 
Load all tasks...
dataset/network_info
===== task: 0    programs: 0/607 =====
Get 128 programs to measure:
.T.T.T.T.T.T.T.T.T.T.T.T.T.T.T.T.T.T.T.T.T.T.T.T........................*T*T*T*T*T*T*T*T*T*T*T*T

I have tried to increase the timeout from 5 to 15 but it has not made a difference and giving the generic cuda as the target instead.

Does anyone have any idea what could be going on?