Skip to content

Add LoRA support for AsyncGRPO #5610

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
jonahsamost wants to merge 3 commits into
base: main
Choose a base branch
from
Open

Add LoRA support for AsyncGRPO #5610

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
2c7fdf8
working lora + gemma4
Apr 20, 2026
e158804
training
Apr 21, 2026
9c5daf1
bugbot
Apr 21, 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
38 changes: 38 additions & 0 deletions trl/chat_template_utils.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -305,6 +305,40 @@ def clone_chat_template(
},
}

gemma4_schema = {
"type": "object",
"properties": {
"role": {"const": "assistant"},
"thinking": {"type": "string"},
"content": {"type": "string"},
"tool_calls": {
"x-regex-iterator": r"<\|tool_call>(.*?)<tool_call\|>",
"type": "array",
"items": {
"type": "object",
"properties": {
"type": {"const": "function"},
"function": {
"type": "object",
"x-regex": r"call\:(?P<name>\w+)(?P<arguments>\{.*\})",
"properties": {
"name": {
"type": "string",
},
"arguments": {
"type": "object",
"x-parser": "gemma4-tool-call",
"additionalProperties": {},
},
},
},
},
},
},
},
"x-regex": r"(\<\|channel\>thought\n(?P<thinking>.*?)\<channel\|\>)?(?P<content>(?:(?!\<\|tool_call\>).)+)?(?P<tool_calls>\<\|tool_call\>.*\<tool_call\|\>)?",
}


deepseekv3_chat_template = (_CHAT_TEMPLATES_DIR / "deepseekv3.jinja").read_text()

Expand All @@ -328,6 +362,8 @@ def clone_chat_template(

qwen3_5_chat_template_4b_and_above = (_CHAT_TEMPLATES_DIR / "qwen3_5_4b_and_above.jinja").read_text()

gemma4_chat_template = (_CHAT_TEMPLATES_DIR / "gemma4.jinja").read_text()


ProcessingClassT = TypeVar("ProcessingClassT", PreTrainedTokenizer, ProcessorMixin)

Expand Down Expand Up @@ -382,6 +418,8 @@ def add_response_schema(processing_class: ProcessingClassT) -> ProcessingClassT:
tokenizer.response_schema = qwen3_schema
elif chat_template in [qwen3_5_chat_template_2b_and_below, qwen3_5_chat_template_4b_and_above]:
tokenizer.response_schema = qwen3_5_schema
elif chat_template in [gemma4_chat_template]:
tokenizer.response_schema = gemma4_schema
else:
raise ValueError(
"Unrecognized chat template, failed to add response schema. Please manually set the response schema on "
Expand Down
Loading