Skip to content

Conversation

@yaoyu-33
Copy link
Contributor

@yaoyu-33 yaoyu-33 commented Sep 23, 2025

  1. cpu only export
  2. fix checkpoint loading by reset model cfg's parallel
  3. delete extra state from state_dict for loading

@copy-pr-bot
Copy link

copy-pr-bot bot commented Sep 23, 2025

This pull request requires additional validation before any workflows can run on NVIDIA's runners.

Pull request vetters can view their responsibilities here.

Contributors can view more details about this message here.

@yaoyu-33
Copy link
Contributor Author

/ok to test 48979ac

Signed-off-by: yaoyu-33 <[email protected]>
@yaoyu-33
Copy link
Contributor Author

/ok to test 724e38f

Signed-off-by: yaoyu-33 <[email protected]>
@yaoyu-33
Copy link
Contributor Author

/ok to test f06b932

Signed-off-by: yaoyu-33 <[email protected]>
@yaoyu-33
Copy link
Contributor Author

/ok to test ffb73d2

@yaoyu-33 yaoyu-33 merged commit e023786 into main Sep 30, 2025
49 of 57 checks passed
@yaoyu-33 yaoyu-33 deleted the yuya/bridge-export-fix branch September 30, 2025 18:53
paul-gibbons pushed a commit to paul-gibbons/Megatron-Bridge that referenced this pull request Oct 29, 2025
* fix cpu init during export

Signed-off-by: yaoyu-33 <[email protected]>

* export env fix

Signed-off-by: yaoyu-33 <[email protected]>

* delete_extra_state for TE related during checkpoint loading for export

Signed-off-by: yaoyu-33 <[email protected]>

* paths fixes

Signed-off-by: yaoyu-33 <[email protected]>

* add override_provider option for checkpoint loading

Signed-off-by: yaoyu-33 <[email protected]>

* add unit test for override_provider option

Signed-off-by: yaoyu-33 <[email protected]>

* remove debug lines

Signed-off-by: yaoyu-33 <[email protected]>

* lint

Signed-off-by: yaoyu-33 <[email protected]>

* unit test fix

Signed-off-by: yaoyu-33 <[email protected]>

---------

Signed-off-by: yaoyu-33 <[email protected]>
Signed-off-by: Paul Gibbons <[email protected]>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

4 participants