-
Notifications
You must be signed in to change notification settings - Fork 981
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
Add torch flop counter to profiler #693
Conversation
@gabrielilharco with this, might be worth putting fvcore import in a try/catch so it'll run natively with torch if fvcore isn't installed (minus act count), and then assert if fvcore profiler is selected and fvcore module not installed? |
Good call, will update |
Removing hierarchical flop counting for now until pytorch/pytorch#111809 is resolved. With #699 it now works for all supported models |
@gabrielilharco looks good, do you want to merge #693 #699 #701 to main and release? |
Sounds good! |
* Add torch flop counter * Moving fvcore to a try/catch * Fix undesired behavior of `transform.PreprocessCfg` * Remove cls_embed arg from forward/encode_image fns * Fix `model.get_model_preprocess_cfg` * Move CLIPA weights to UCSC-VLAA org * Remove outdated comment in pretrained.py * Fix arg not being passed to `image_transform` from `image_transform_v2` * remove hierarchical flop counting --------- Co-authored-by: Santiago Castro <[email protected]> Co-authored-by: Ross Wightman <[email protected]>
No description provided.