-
Notifications
You must be signed in to change notification settings - Fork 230
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
make num samples work for directed prompt-based sequence generation #715
Conversation
@helloguo you can use the performance reporting and prompt-based sequence generation with support for --num-samples for benchmarking your ARM desktop performance improvements in pytorch/pytorch#125611 |
@@ -743,28 +746,33 @@ def callback(x): | |||
tokens_generated = y.size(0) - prompt_length | |||
tokens_sec = tokens_generated / t | |||
aggregate_metrics["tokens_per_sec"].append(tokens_sec) | |||
logging.debug( | |||
print( |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Is this change intentional? Like we want to always print this
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Yes, performance info is not logging. This is the whole reason we run this, we can't just shove it into a log. Mind you, we do dump the entire IR on people when we export from ET.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
It feels like this is a regression from a previous decision we've made, and logging.info
should have been sufficient there.. (or setup logger in main routine
make num samples work for directed prompt-based sequence generation