Custom matmul attention #213
Closed
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
ngc92
force-pushed
the
custom-matmul-attention
branch
2 times, most recently
from
156fe48 to
3fb586c
Compare
ngc92
force-pushed
the
custom-matmul-attention
branch
from
64686b6 to
b43b84b
Compare
|
this was never intended to be actually merged, just to demonstrate how the custom matmul would look like. Now that it is full of conflicts, let's keep the number of open PRs to a reasonable level. |
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Outdated suggestions cannot be applied.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.
My own implementation of (lower-triangular) matrix multiplication.
It is not as efficient as CuBLAS, but since we only calculate half as many numbers, it is a net win.
Cannot get rid of
permuteyet, becausevis still needed later in the kernel. Will look into replacing the second matmul too, so we can capitalize on not needing permutes (might require changes to backward, too).