Support per-layer MTP configuration - #48264
Open
eladsegal wants to merge 5 commits into
Open
Conversation
Contributor
CI recapDashboard: View test results in Grafana |
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
What does this PR do?
Adds independent per-layer configuration for MTP layers.
Previously,
get_mtp_config()retained the main model'sper_layer_config. This could apply the main-model overrides to MTP layers unintentionally.With this PR, MTP layers discard the main model's per-layer overrides by default and can define their own through
mtp_per_layer_config. MTP masks are created from the corresponding layer config, preserving per-layer settings such assliding_window.This also fixes MTP attention-mask selection for Inkling. Previously, mask creation looked for
layer_typeon the genericMtpLayerwrapper, which does not define it, and therefore always fell back to full attention. As a result, Inkling did not use the correct masks for itshybridandhybrid_slidingMTP layers. Mask creation now selects the pattern fromlayer_typesin the MTP config.Who can review?
@ArthurZucker
@Cyrilvallez