-
Notifications
You must be signed in to change notification settings - Fork 3.4k
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 enable_model_summary
flag and deprecate weights_summary
#9699
Add enable_model_summary
flag and deprecate weights_summary
#9699
Conversation
enable_model_summary
flagenable_model_summary
flag and deprecate weights_summary
3a7328b
to
d83ed9d
Compare
075fb46
to
8de7fd5
Compare
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.
LG!. Just some minor comments.
877d0bf
to
4e87889
Compare
Codecov Report
@@ Coverage Diff @@
## master #9699 +/- ##
=======================================
- Coverage 93% 89% -4%
=======================================
Files 178 178
Lines 15653 15671 +18
=======================================
- Hits 14511 13920 -591
- Misses 1142 1751 +609 |
Do we need the |
This also came up in these places: Options so far:
@PyTorchLightning/core-contributors Poll: Option 1 -> 👍 |
I am strongly in favor of option 1. To expand on "consistent naming scheme" you mentioned, consider these Trainer arguments:
if I were reading these arguments and then saw one named IMO boolean argument names should be prefixed with a verb, such as |
Question: under what circumstances would users want to disable the model summary? Is it critical we have this flag? |
Maybe the summary is not correct for your model, maybe you are calling |
Hmm, if the summary is not correct that sounds like a bug, and we should operate under the assumption that we have no bugs. The other case seems pretty rare, but I don't have a lot of background knowledge here so don't feel strongly |
3a1c155
to
fe8d615
Compare
@williamFalcon makes sense, I agree user experience is very important and didn’t mean to suggest otherwise! |
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.
LG!
for more information, see https://pre-commit.ci
Co-authored-by: Adrian Wälchli <aedu.waelchli@gmail.com>
…tning-AI#9699) Co-authored-by: Adrian Wälchli <aedu.waelchli@gmail.com> Co-authored-by: Kaushik B <45285388+kaushikb11@users.noreply.github.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Co-authored-by: Kaushik B <kaushikbokka@gmail.com>
What does this PR do?
Fixes #9043
Adds
enable_model_summary
flag to matchenable_progress_bar
flag being added here: #9664This PR also deprecates the Trainer
weights_summary
flag and property which dictated the mode for display. Users are expected to customize this with the callback. This deprecation was required because we needed to support arbitrary module depths for summarization (int) while the current mode only supported a few fixed modes ("top" vs "full"). This customization is much easier with the ModelSummary callbacks @kaushikb11 introduced.Does your PR introduce any breaking changes? If yes, please list them.
Before submitting
PR review
Anyone in the community is welcome to review the PR.
Before you start reviewing make sure you have read Review guidelines. In short, see the following bullet-list:
Did you have fun?
Make sure you had fun coding 🙃