-
Notifications
You must be signed in to change notification settings - Fork 355
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
Group norm bug fix #3014
Group norm bug fix #3014
Conversation
5f47fb1
to
e8d942f
Compare
937d342
to
8649c5b
Compare
…omposed ops instead of using INormalization Layer. Supported dynamic shape
8649c5b
to
03aca50
Compare
reshaped_input.shape, | ||
) | ||
|
||
eps_trt = get_trt_tensor(ctx, eps, f"{name}_eps", np.float32) |
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.
Why is this converted to np.float32?
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 is 1e-6 in most cases.
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.
float16 cannot handle that close to 0
f"{name}_div", | ||
sub_trt, | ||
sqrt_trt, | ||
) | ||
|
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.
Wanted to clarify if this div would require any mode, eg: trunc? Are the data types always compatible with the output types?
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.
I don't think so. The previous implementation from Evan did not include any mode.
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.
just FYI: we have impl.elementwise.floor_divide
and impl.elementwise.trunc_div
helper functions, if needed.
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.
LGTM
Description
Please include a summary of the change and which issue is fixed. Please also include relevant motivation and context. List any dependencies that are required for this change.
Fixes # (issue)
Type of change
Please delete options that are not relevant and/or add your own.
Checklist: