[Feature][MTP]support mtp in v1_scheduler mode #3695
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
|
Thanks for your contribution! |
carryyu
reviewed
Aug 29, 2025
| self.model_inputs["block_tables"][idx : idx + 1, :encoder_block_num] = np.array( | ||
| request.block_tables, dtype="int32" | ||
| ) | ||
| # if self.model_inputs["is_block_step"][idx]: # has tasks to continue to decode |
Collaborator
Author
There was a problem hiding this comment.
这里保留注释的位置，是和陈坚在 Target Model 的 insert 有点区别，但我不确定是否和 PD 分离（未验证）或者 EP 有关系，希望先保留着；等后续验证完别的功能，再删除~
Collaborator
There was a problem hiding this comment.
这里和 PD 分离没有关系，就是正常确认 worker 需不需要做下一个step 的判定逻辑。（防止空输入，也防止有输入却没有正常 step)
Codecov Report✅ All modified and coverable lines are covered by tests. Additional details and impacted files@@ Coverage Diff @@
## develop #3695 +/- ##
=========================================
Coverage ? 8.42%
=========================================
Files ? 6
Lines ? 95
Branches ? 8
=========================================
Hits ? 8
Misses ? 87
Partials ? 0
Flags with carried forward coverage won't be shown. Click here to find out more. ☔ View full report in Codecov by Sentry. 🚀 New features to boost your workflow:
|
rainyfly
previously approved these changes
Sep 1, 2025
| self.model_inputs["block_tables"][idx : idx + 1, :encoder_block_num] = np.array( | ||
| request.block_tables, dtype="int32" | ||
| ) | ||
| # if self.model_inputs["is_block_step"][idx]: # has tasks to continue to decode |
Collaborator
There was a problem hiding this comment.
这里和 PD 分离没有关系，就是正常确认 worker 需不需要做下一个step 的判定逻辑。（防止空输入，也防止有输入却没有正常 step)
DDDivano
previously approved these changes
Sep 2, 2025
heavengate
approved these changes
Sep 2, 2025
freeliuzc
force-pushed
the
support_v1
branch
from
2c415cb to
ce99dbe
Compare
DDDivano
approved these changes
Sep 4, 2025
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
7 participants
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.
- 预取 2*MaxDraftToken + 2 位置
- kvcacheSchedule 放到 proposer run 后面
- 适应性修改多个 Kernel，新增 speculate_schedule_cache 管理投机解码 block；
- 修改 draft_model_preprocess，同时支持 V0/V1，管理 MTP 的所有 block 状态，并优化 Kernel 逻辑
- 修改 recover_decode_task，同时支持投机解码与非投机解码