-
Notifications
You must be signed in to change notification settings - Fork 6
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
Lower aten.linalg_vector_norm
to ttnn.moreh_norm
#686
base: main
Are you sure you want to change the base?
Conversation
I consistently observe PCC errors for reduction along non-last dimension ``` FAILED tests/lowering/reduction/test_vector_norm.py::test_vector_norm[input_shape1-2-2] - AssertionError: 0.9976093101814988 FAILED tests/lowering/reduction/test_vector_norm.py::test_vector_norm[input_shape2-2-1] - AssertionError: 0.8991108908766973 ```
try: | ||
ndims = len(x.meta["val"].size()) | ||
dim = [d if d >= 0 else d + ndims for d in dim] | ||
except: |
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.
what exception does this cause?
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.
If the shape of x
is not known at compile time, x.meta["val"].size()
fails.
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.
maybe then better to handle inside the target wrapper where values are already known?
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 change it to more graceful get_shape
Is there a ticket? |
I doubt if this makes a difference in Python though
Ticket
Problem description
Compile
aten.linalg_vector_norm
tottnn.moreh_norm
, the PyTorch FX wrapper forttnn.operations.moreh.norm
.I consistently observe PCC errors for reduction along non-last dimension. I am reporting this as a kernel issue:
ttnn.moreh_norm
for non-last dim tt-metal#16335What's changed