Skip to content

add qwen3 to megatron conversion #802

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

Closed
abdalgader-a wants to merge 1 commit into from
Closed

add qwen3 to megatron conversion #802

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
7 changes: 6 additions & 1 deletion nemo_rl/models/megatron/community_import.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -85,9 +85,14 @@ def export_model_from_megatron(
from nemo.tron.converter.qwen import HFQwen2Exporter

exporter_cls = HFQwen2Exporter

elif hf_config.model_type == "qwen3":
Copy link
Contributor

@ashors1 ashors1 Aug 1, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Suggested change
elif hf_config.model_type == "qwen3":
elif hf_config.model_type in ("qwen3", "qwen3_moe"):

Copy link
Contributor

@ashors1 ashors1 Aug 1, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Ah, actually I just tested this and it looks like this doesn't work with MoE models at the moment. Could we extend this to support MoE? If too much work, we could always add MoE support in a follow up PR

from nemo.tron.converter.qwen import HFQwen3Exporter

exporter_cls = HFQwen3Exporter
else:
raise ValueError(
f"Unknown model: {hf_model_name}. Currently, only Qwen2 and Llama are supported. "
f"Unknown model: {hf_model_name}. Currently, only Qwen2, Qwen3 and Llama are supported. "
"If you'd like to run with a different model, please raise an issue or consider adding your own converter."
)
print(f"Exporting model {hf_model_name} to {output_path}...")
Expand Down
Loading