conv2d op bias fix #1353
Draft
conv2d op bias fix #1353
+10
−10
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
|
||||||||||||||||||||||||||||
|
||||||||||||||||||||||||||||||||||||
|
||||||||||||||||||||||||||||
|
||||||||||||||||||||||||||||||||||||
ashokkumarkannan1
force-pushed
the
akannan/conv2d_bias_issue
branch
from
bc97c0c to
f729548
Compare
|
||||||||||||||
|
||||||||||||||
1 similar comment
|
||||||||||||||
|
||||||||||||||
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Ticket
tenstorrent/tt-mlir#1533
Problem description
'ttnn.conv2d' op Bias must only have data on the final dimenstion
What's changed
Description
The conv2d bias will be fused during the post_initial_graph stage by the fuse_conv2d_bias pass where the bias edge which is added after the conv2d op will be connected to the conv2d port as part of the fusing the old edge attributes of type NCHW will be copied to the new edge as the tms introduce to the conv2d bias edge which causes the layout issue.
The approach to solving this issue involves removal of broadcast dims which is not necessary for the conv2d bias edge ,as we require the bias to be in the shape: (1, 1, 1, C_out) we will add the transpose to the decomposition of the conv2d for maintaining the channel last for the ttnn.