Skip to content

Extend LoRA for Gemma4 #3969

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

Open
RexBearIU wants to merge 1 commit into
base: main
Choose a base branch
from
+6 −0
Open

Extend LoRA for Gemma4 #3969

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
1 change: 1 addition & 0 deletions src/maxtext/configs/post_train/lora_module_path.yml
View file
Open in desktop
Copy link
Copy Markdown
Collaborator

@gagika gagika Jun 2, 2026

Choose a reason for hiding this comment

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

Are there any tests for Gemma4 Lora?

Copy link
Copy Markdown
Collaborator Author

@RexBearIU RexBearIU Jun 3, 2026

Choose a reason for hiding this comment

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

We ran the end-to-end LoRA training loop for Gemma 4 successfully without any issues.4 successfully without any issues. log

Original file line number Diff line number Diff line change
Expand Up @@ -21,6 +21,7 @@ mistral: "decoder/layers/.*(attention/(query|key|value|out)|mlp/(wi_0|wi_1|wo))"
deepseek2: "decoder/(dense_layers|moe_stack)/self_attention/(query|out|wkv_a|wkv_b)|decoder/(dense_layers|moe_stack)/(mlp|shared_experts)/(wi_0|wi_1|wo)"
gemma2: "decoder/layers/(self_attention_local|self_attention_global)/(query|key|value|out)|decoder/layers/(mlp_local|mlp_global)/(wi_0|wi_1|wo)"
gemma3: "decoder/layers/.*(self_attention/(query|key|value|out)|mlp/(wi_0|wi_1|wo|gate|up|down))"
gemma4: "decoder/(scanned_blocks|layers_remainder)/layers.*/.*(self_attention/(query|key|value|out)|mlp/.*(wi_0|wi_1|wo|shared_experts/(wi_0|wi_1|wo)))"
olmo3: "decoder/layers/.*(attention/(query|key|value|out)|mlp/(wi_0|wi_1|wo))"
Comment thread
RexBearIU marked this conversation as resolved.
gpt3: "decoder/layers/(self_attention/(qkv_proj|out)|mlp/(wi|wo))"

Expand Down
5 changes: 5 additions & 0 deletions src/maxtext/trainers/post_train/sft/train_sft.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -264,9 +264,14 @@ def setup_trainer_state(mt_config, goodput_recorder=None):
def train_model(mt_config, trainer, mesh):
"""Runs the SFT training loop in Tunix."""
with mesh, nn_partitioning.axis_rules(mt_config.logical_axis_rules):
# Disable NNX graph caching for MoE models (where experts > 1) to allow
# necessary dynamic metadata synchronization during forward passes (e.g., in jax.lax.scan).
enable_nnx_cache = mt_config.num_experts <= 1

trainer.train(
trainer.data_hooks.train_data_iterator,
trainer.data_hooks.eval_data_iterator,
cache_nnx_graph=enable_nnx_cache,
)
return trainer

Expand Down
Loading