Skip to content

Gemma3 Provider & Bridge - #867

Merged
suiyoubi merged 23 commits into
mainfrom
aot/gemma3-provider
Oct 15, 2025
Merged

Gemma3 Provider & Bridge#867
suiyoubi merged 23 commits into
mainfrom
aot/gemma3-provider

Conversation

@suiyoubi

@suiyoubi suiyoubi commented Oct 3, 2025

Copy link
Copy Markdown
Contributor

part of #147

Signed-off-by: Ao Tang <aot@nvidia.com>
- Renamed `Gemma3ModelProvider` to `GemmaModelProvider` across multiple files for consistency.
- Updated the mapping logic in `GemmaModelBridge` to use a dictionary for better readability and maintainability.
- Adjusted imports to reflect the new provider name.

Signed-off-by: Ao Tang <aot@nvidia.com>
Signed-off-by: Ao Tang <aot@nvidia.com>
@copy-pr-bot

copy-pr-bot Bot commented Oct 3, 2025

Copy link
Copy Markdown

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.

Signed-off-by: Ao Tang <aot@nvidia.com>
Signed-off-by: Ao Tang <aot@nvidia.com>
Signed-off-by: Ao Tang <aot@nvidia.com>
Signed-off-by: Ao Tang <aot@nvidia.com>
Signed-off-by: Ao Tang <aot@nvidia.com>
@suiyoubi

suiyoubi commented Oct 3, 2025

Copy link
Copy Markdown
Contributor Author

/ok to test 12d96b5

Signed-off-by: Ao Tang <aot@nvidia.com>
@suiyoubi

suiyoubi commented Oct 3, 2025

Copy link
Copy Markdown
Contributor Author

/ok to test 088fe9a

f"layers.{local_layer_number}.",
f"layers.{global_layer_number}.",
)
if isinstance(layer_module, TransformerLayer):

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

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

why is this needed?

Copy link
Copy Markdown
Contributor Author

Choose a reason for hiding this comment

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

Non Mcore models does not neccessary have this attribute: e.g. The vision encoder for gemma3 (even tho it is not sharded)

Comment thread src/megatron/bridge/models/conversion/param_mapping.py Outdated
Comment thread src/megatron/bridge/models/conversion/param_mapping.py Outdated
@suiyoubi

Copy link
Copy Markdown
Contributor Author

/ok to test 148e3d0

Signed-off-by: Ao Tang <aot@nvidia.com>
@suiyoubi

Copy link
Copy Markdown
Contributor Author

/ok to test 019e81d

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.

3 participants