Skip to content

update bid to match each layers MTP source #23237

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
am17an merged 2 commits into from
May 18, 2026
Merged

update bid to match each layers MTP source #23237

Changes from 1 commit
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
7fe2eef
update bid to match each layers MTP source
May 17, 2026
734c6bd
Update conversion/qwen.py
bit-incarnas May 18, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
6 changes: 6 additions & 0 deletions conversion/qwen.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -600,6 +600,12 @@ def modify_tensors(self, data_torch: Tensor, name: str, bid: int | None) -> Iter
if name.find("layers.") != -1:
assert bid is not None
name = name.replace(f"mtp.layers.{bid}", f"model.layers.{bid + n_layer}")
# Keep bid coherent with the remapped name so downstream MoE
# expert-stacking (Qwen2MoeModel) caches and reads at the same
# layer slot. Without this, self._experts[0] gets populated with
# layer-48 names, then the stacker builds layer-0 lookup keys
# and KeyErrors.
Comment thread
bit-incarnas marked this conversation as resolved.
Outdated
bid = bid + n_layer
else:
remapper = {
"mtp.fc": "model.layers.{bid}.eh_proj",
Expand Down