https://open-source-silicon.dev logo
Join Slack
Powered by
Hi everyone, so I am trying to use autotuner with ...
# openroad
a
Hi everyone, so I am trying to use autotuner with openroad, but i am getting a deprecationWarning for "ray". So I wanted to ask is there a specific version of ray that I can specify to make autotuner work? any advice is much appreciated thank you. Added the screenshot for your reference.
v
@Vitor Viana Bandeira can you plz look into this?
v
@Akshit Bhatia The supported version is 
2.9.3
as per the 
requirements.txt
file. Also, did you follow the steps in our docs? https://github.com/The-OpenROAD-Project/OpenROAD-flow-scripts/tree/master/tools/AutoTuner#setting-up-autotuner
a
Yes I followed the git, and am still getting the above error. Do you have any suggestions how I cam make it work?
v
can you double check the version installed with 
pip3 list | grep ray
?
a
I get this as the output "memray 1.14.0 ray 2.38.0"
v
so you have some issue with your local environment that is preventing AT to get the correct version
did you run 
./tools/AutoTuner/setup.sh
on your current shell or only once?
a
I ran that only once
v
you need to run this on every shell before calling the distributed.py, could you try again?
a
So I ran the command "./tools/AutoTuner/setup.sh" and then in the folder "./tools/AutoTuner/src/autotuner" I ran the command "python3 distributed.py --design gcd --platform sky130hd \ --config ../../../../flow/designs/sky130hd/gcd/autotuner.json \ tune --samples 5", and it is still giving me the same DeprecationWarning.
Any advice on why is it not working still?
v
did you check to see if the version was propagated correctly after running setup.sh again?
maybe try sourcing the setup.sh file
a
i checked the version after the setup and it's still the same
v
did you try 
source setup.sh
? and check again the version. as long as your environment does not pick up the correct version AT will no work
worst case run the commands from 
setup.sh
by hand, if even that does not work you have something peculiar with your setup
a
yes the source setup.sh worked. Thank you so much for your help.
🙌 1
Also, I wanted to ask one more question, in order to see the GUI, we have to run this command "tensorboard --logdir=./<logpath>", but how to find the logpath? abd do we have to run this command before we run the command "python3 distributed.py --design gcd --platform sky130hd \ --config ../../../../flow/designs/sky130hd/gcd/autotuner.json \ tune --samples 5"?
as the GCP tutorial is not there on git, it is not very clear how to use the GUI for monitoring
v
you will ../../../../flow/logs/sky130hd/gcd/<autotuner logs> name will be created here
a
Ok thank you so much
12 Views
Open in Slack
PreviousNext
Powered by