[KV Cache Injection] Causal Mask implementation for OPT and CodeGen #1677
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
bfineran
previously approved these changes
Jul 25, 2023
* initial implementation; testing now * fix a small blunder * cleanup --------- Co-authored-by: bogunowicz@arrival.com <bogunowicz@arrival.com>
bfineran
previously approved these changes
Jul 25, 2023
* initial implementation; testing now * fix a small blunder * cleanup * initial implementation * on to testing with deepsparse --------- Co-authored-by: bogunowicz@arrival.com <bogunowicz@arrival.com>
dbogunowicz
force-pushed
the
feature/damian/refactor_injection
branch
from
7737994
to
224412f
Compare
bfineran
previously approved these changes
Jul 27, 2023
| @@ -84,7 +81,7 @@ class Config: | |||
|
|
|||
| OPT_CONFIG = KeyValueCacheConfig( | |||
| model_name="opt", | |||
| positions_adjustment_transform=PositionsAdjustmentOPT, | |||
| additional_transforms=AdditionalTransformsOPT, | |||
There was a problem hiding this comment.
would be great to support a list here eventually instead in case we don't want to squash many steps into a single transform or enable more code sharing between model specific additional transforms outside of inheritance
dbogunowicz
changed the title
bfineran
approved these changes
Jul 27, 2023
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Outdated suggestions cannot be applied.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.
Introducing a gentle refactoring of
position_adjustmenttransformations toadditional_transformstransformations. The aim is to make thepositions_adjustmenttransformation more general since this transformation will now also include causal mask injection.Note: This PR contains the CodeGen causal mask support: #1676 as well as OPT causal mask support: #1688
Note2: Also added more verbose logging to the kv cache injection process:
e.g. for OPT:
and for CodeGen