[0.9.1][MTP V1]MTP model adapt torchair #1840
Merged
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
NeverRaR
suggested changes
Jul 22, 2025
JC-ut0
force-pushed
the
v0.9.1-dev
branch
3 times, most recently
from
c201ddb to
aec1751
Compare
JC-ut0
force-pushed
the
v0.9.1-dev
branch
2 times, most recently
from
ec6ac82 to
bb12ac5
Compare
Signed-off-by: xuyexiong <[email protected]>
ganyi1996ppo
approved these changes
Jul 23, 2025
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Outdated suggestions cannot be applied.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.
What this PR does / why we need it?
MTP model adapt torchair graph mode
The MTP model only utilizes the torchair graph during the decode phase. Our padding strategy involves running a fixed number of 1+MTP tokens per batch, regardless of whether the main model accepts or rejects them. When generating the MTP hidden state, we only take the last index of the accepted tokens for each batch.
Does this PR introduce any user-facing change?
How was this patch tested?
Tested in DP4/TP4, TP16 DP1, TP8 DP2, and Prefilling Decoding Disaggregation