adjust HPU warmup: use dummy inputs with shape more close to real scenario #689
Conversation
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
|
@regisss , pls help review, thx! |
…nario to avoid wrong output from reranker model Signed-off-by: Liu, Kaixuan <[email protected]>
|
cc @Narsil |
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Outdated suggestions cannot be applied.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.
In original implementation, we use dummy inputs with shapes like [1,128], [1,256],[2,128],[2,256] to do warmup, aiming to generate recipe cache in warmup stage. And in real serving scenario, we padding the input_ids/attention_masks to shapes cached in warmup stage. However, we found precision issue for reranker models following tei docs . We think it may be because wrong graphs/recipe was used during replay stage. Hence we adjust the
create_warmup_batchfunction in this PR, to make the dummy inputs more close to real scenario, hence during warmup stage, these inputs will also be padded in python backend and will generate right recipe caching/graph, which will be the same with serving stage. We made several round experiments, and the wrong output issue disappears after this PR.