Skip to content

[Fix] Fix weight_loader property assignment for qwen3-next FP8 models #21662

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
Fridge003 merged 2 commits into from
Mar 30, 2026
Merged

[Fix] Fix weight_loader property assignment for qwen3-next FP8 models #21662

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
74db2cd
[Fix] Fix weight_loader assignment for qwen3-next FP8 models
Fridge003 Mar 30, 2026
7b9612b
[Fix] Handle both ModelWeightParameter and plain params in weight_loa…
Fridge003 Mar 30, 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
21 changes: 17 additions & 4 deletions python/sglang/srt/models/qwen3_next.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -135,11 +135,11 @@ def __init__(

# Override weight_loader for packed checkpoint format.
# Must capture original_loader BEFORE overwriting.
self.in_proj_qkvz.weight.weight_loader = self._make_packed_weight_loader(
self.in_proj_qkvz
self._override_weight_loader(
self.in_proj_qkvz, self._make_packed_weight_loader(self.in_proj_qkvz)
)
self.in_proj_ba.weight.weight_loader = self._make_packed_weight_loader(
self.in_proj_ba
self._override_weight_loader(
self.in_proj_ba, self._make_packed_weight_loader(self.in_proj_ba)
)

# Conv1d weight loader setup
Expand Down Expand Up @@ -216,6 +216,19 @@ def __init__(
dt_bias=self.dt_bias,
)

@staticmethod
def _override_weight_loader(module, new_loader):
"""Override weight_loader on a module's weight parameter.

ModelWeightParameter exposes weight_loader as a read-only property
backed by _weight_loader, while plain parameters store it as a
regular attribute. This helper handles both cases."""
param = module.weight
if hasattr(param, "_weight_loader"):
param._weight_loader = new_loader
else:
param.weight_loader = new_loader

@staticmethod
def _make_packed_weight_loader(module):
"""Create a weight_loader that does contiguous TP slicing for fused
Expand Down
Loading