Skip to content

cp: [Model Support] Qwen3-Next (1068) into r0.2.0 #1126

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鈥檒l occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
ananthsub merged 1 commit into from
Oct 30, 2025
Merged

cp: [Model Support] Qwen3-Next (1068) into r0.2.0 #1126

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
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
2 changes: 2 additions & 0 deletions examples/conversion/compare_hf_and_megatron/compare.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -490,6 +490,7 @@ def _load_megatron_model(args):
model_provider.expert_model_parallel_size = ep
model_provider.expert_tensor_parallel_size = etp
model_provider.pipeline_dtype = torch.bfloat16
model_provider.finalize()
model_provider.initialize_model_parallel(seed=0)
megatron_model = bridge.load_megatron_model(
args.megatron_model_path,
Expand All @@ -510,6 +511,7 @@ def _load_megatron_model(args):
model_provider.expert_model_parallel_size = ep
model_provider.expert_tensor_parallel_size = etp
model_provider.pipeline_dtype = torch.bfloat16
model_provider.finalize()
model_provider.initialize_model_parallel(seed=0)
megatron_model = model_provider.provide_distributed_model(wrap_with_ddp=False)

Expand Down
2 changes: 2 additions & 0 deletions pyproject.toml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -66,6 +66,7 @@ classifiers = [
"Topic :: Utilities",
]
dependencies = [
"transformers==4.57.1",
"datasets",
"omegaconf>=2.3.0",
"tensorboard>=2.19.0",
Expand Down Expand Up @@ -98,6 +99,7 @@ no-build-isolation-package = [
"causal-conv1d",
"nv-grouped-gemm",
"flash_mla",
"flash-linear-attention",
]
prerelease = "allow"
override-dependencies = [
Expand Down
6 changes: 2 additions & 4 deletions src/megatron/bridge/models/conversion/auto_bridge.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -701,10 +701,8 @@ def to_megatron_provider(self, load_weights: bool = True, hf_path: str | Path |
provider: ModelProviderMixin = self._model_bridge.provider_bridge(self.hf_pretrained)

if load_weights:
# TODO(yuya): Temporary workaround: keep initialization enabled.
# When disabled, TP-related attributes are not added properly and TP breaks in Megatron-Core.
# We will fix this in the core soon; setting this to True may slow startup a bit.
provider.perform_initialization = True
# Skip weights initialization since we are going to load weights
provider.perform_initialization = False
if hf_path is None:
provider.register_pre_wrap_hook(
partial(self._model_bridge.load_weights_hf_to_megatron, self.hf_pretrained)
Expand Down
Loading