if you would like to have your name in an MLPerf inference submission here is your chance. You just need to have a decent CPU and about 50 GB disk space as the ML model is sparsified. Sorry, there is no time to explain what's happening inside -- but we can discuss that after the submission deadline which is Friday afternoon India time. Those who manage to run these commands won't regret the experience as you'll get to know many things about ML model performance. Join here for further queries. https://discord.gg/H6gGd9Kp If on Ubuntu: do this before running the script. Just change "apt" for other distros.

"sudo apt-get install -y wget git curl python3-dev python3-pip && python3 -m pip install cmind && source .profile && cm pull repo ctuning@mlcommons-ck"

Script to run: Can download this and do "bash run-pruned-bert.sh"

https://github.com/ctuning/mlcommons-ck/blob/master/cm-mlops/script/run-all-mlperf-models/run-pruned-bert.sh

For any queries: https://discord.gg/H6gGd9Kp

Successful participants will have the opportunity to do much cooler work in the next month. 

posted in From GO Admins Feb 21
by
874 views
3
Like
0
Love
0
Haha
0
Wow
1
Angry
0
Sad

5 Comments

5 Comments

Like

No update from anyone surprise

Like
1
Hi Arjun , Working guy (Experienced golang dev) here.
I just saw this, interested in running this.I have a thinkpad with linux mint,
I'll run this tmrw , i don't know if it will be still valid but i'm simply interested to know more about this just for knowledge.
Like

@ragnarocks Unfortunately the normal deadline is over but you can still run it for knowledge and for future submission - happens every 6 months. Also, if someone is interested in trying LLaMa2 model, the deadline is coming Friday. This is challenging as it involves doing coding and not just running. 

Like
I want to try the llama2 model, where to find the details
Like

@mikasa_ackermann Please join https://discord.gg/6su9FpSU and you can ask that. These are very much in flux and there are no proper READMEs. But the main challenge to run LLama2 is to have a good GPU like Nvidia RTX 4090. On CPU, it can take weeks or months to finish the run.