NPUW: Switch to PYRAMID attention by default for prefill#33736
Open
dmatveev wants to merge 5 commits intoopenvinotoolkit:masterfrom
Open
NPUW: Switch to PYRAMID attention by default for prefill#33736dmatveev wants to merge 5 commits intoopenvinotoolkit:masterfrom
dmatveev wants to merge 5 commits intoopenvinotoolkit:masterfrom
Conversation
Will only work if the chunking is enabled: - PREFILL_HINT is DYNAMIC (default) - MAX_PROMPT_LEN > CHUNK_SIZE (both 1024 by default)
Currently we only route attention preferences to the npuw::CompiledModel when they're passed in the properties. This doesn't cover the case when e.g. PYRAMID hint gets enabled by default -- it is not present in the explicit props, so the `cfg` infrastructure should be used instead.
github-actions
bot
added
category: NPU
This file contains hidden or 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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Will only work if the chunking is enabled:
Also fixed a bug: Currently we only route attention preferences to the npuw::CompiledModel when they're passed in the properties. This doesn't cover the case when e.g. PYRAMID hint gets enabled by default -- it is not present in the explicit props, so the
cfginfrastructure should be used instead.