Skip to content

[model] support Qwen/Qwen3Guard-Gen-0.6B series #6189

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

Merged
merged 1 commit into from
Oct 17, 2025
+50 −0
Merged

[model] support Qwen/Qwen3Guard-Gen-0.6B series #6189

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
3 changes: 3 additions & 0 deletions docs/source/Instruction/支持的模型和数据集.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -205,6 +205,9 @@
|[Qwen/Qwen3-14B-AWQ](https://modelscope.cn/models/Qwen/Qwen3-14B-AWQ)|qwen3|qwen3|transformers>=4.51|✘|-|[Qwen/Qwen3-14B-AWQ](https://huggingface.co/Qwen/Qwen3-14B-AWQ)|
|[Qwen/Qwen3-32B-AWQ](https://modelscope.cn/models/Qwen/Qwen3-32B-AWQ)|qwen3|qwen3|transformers>=4.51|✘|-|[Qwen/Qwen3-32B-AWQ](https://huggingface.co/Qwen/Qwen3-32B-AWQ)|
|[swift/Qwen3-32B-AWQ](https://modelscope.cn/models/swift/Qwen3-32B-AWQ)|qwen3|qwen3|transformers>=4.51|✘|-|-|
|[Qwen/Qwen3Guard-Gen-0.6B](https://modelscope.cn/models/Qwen/Qwen3Guard-Gen-0.6B)|qwen3_guard|qwen3_guard|transformers>=4.51|✘|-|[Qwen/Qwen3Guard-Gen-0.6B](https://huggingface.co/Qwen/Qwen3Guard-Gen-0.6B)|
|[Qwen/Qwen3Guard-Gen-4B](https://modelscope.cn/models/Qwen/Qwen3Guard-Gen-4B)|qwen3_guard|qwen3_guard|transformers>=4.51|✘|-|[Qwen/Qwen3Guard-Gen-4B](https://huggingface.co/Qwen/Qwen3Guard-Gen-4B)|
|[Qwen/Qwen3Guard-Gen-8B](https://modelscope.cn/models/Qwen/Qwen3Guard-Gen-8B)|qwen3_guard|qwen3_guard|transformers>=4.51|✘|-|[Qwen/Qwen3Guard-Gen-8B](https://huggingface.co/Qwen/Qwen3Guard-Gen-8B)|
|[Qwen/Qwen3-4B-Thinking-2507](https://modelscope.cn/models/Qwen/Qwen3-4B-Thinking-2507)|qwen3_thinking|qwen3_thinking|transformers>=4.51|✔|-|[Qwen/Qwen3-4B-Thinking-2507](https://huggingface.co/Qwen/Qwen3-4B-Thinking-2507)|
|[Qwen/Qwen3-4B-Thinking-2507-FP8](https://modelscope.cn/models/Qwen/Qwen3-4B-Thinking-2507-FP8)|qwen3_thinking|qwen3_thinking|transformers>=4.51|✘|-|[Qwen/Qwen3-4B-Thinking-2507-FP8](https://huggingface.co/Qwen/Qwen3-4B-Thinking-2507-FP8)|
|[Qwen/Qwen3-30B-A3B-Instruct-2507](https://modelscope.cn/models/Qwen/Qwen3-30B-A3B-Instruct-2507)|qwen3_nothinking|qwen3_nothinking|transformers>=4.51|✔|-|[Qwen/Qwen3-30B-A3B-Instruct-2507](https://huggingface.co/Qwen/Qwen3-30B-A3B-Instruct-2507)|
Expand Down
3 changes: 3 additions & 0 deletions docs/source_en/Instruction/Supported-models-and-datasets.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -205,6 +205,9 @@ The table below introduces the models integrated with ms-swift:
|[Qwen/Qwen3-14B-AWQ](https://modelscope.cn/models/Qwen/Qwen3-14B-AWQ)|qwen3|qwen3|transformers>=4.51|✘|-|[Qwen/Qwen3-14B-AWQ](https://huggingface.co/Qwen/Qwen3-14B-AWQ)|
|[Qwen/Qwen3-32B-AWQ](https://modelscope.cn/models/Qwen/Qwen3-32B-AWQ)|qwen3|qwen3|transformers>=4.51|✘|-|[Qwen/Qwen3-32B-AWQ](https://huggingface.co/Qwen/Qwen3-32B-AWQ)|
|[swift/Qwen3-32B-AWQ](https://modelscope.cn/models/swift/Qwen3-32B-AWQ)|qwen3|qwen3|transformers>=4.51|✘|-|-|
|[Qwen/Qwen3Guard-Gen-0.6B](https://modelscope.cn/models/Qwen/Qwen3Guard-Gen-0.6B)|qwen3_guard|qwen3_guard|transformers>=4.51|✘|-|[Qwen/Qwen3Guard-Gen-0.6B](https://huggingface.co/Qwen/Qwen3Guard-Gen-0.6B)|
|[Qwen/Qwen3Guard-Gen-4B](https://modelscope.cn/models/Qwen/Qwen3Guard-Gen-4B)|qwen3_guard|qwen3_guard|transformers>=4.51|✘|-|[Qwen/Qwen3Guard-Gen-4B](https://huggingface.co/Qwen/Qwen3Guard-Gen-4B)|
|[Qwen/Qwen3Guard-Gen-8B](https://modelscope.cn/models/Qwen/Qwen3Guard-Gen-8B)|qwen3_guard|qwen3_guard|transformers>=4.51|✘|-|[Qwen/Qwen3Guard-Gen-8B](https://huggingface.co/Qwen/Qwen3Guard-Gen-8B)|
|[Qwen/Qwen3-4B-Thinking-2507](https://modelscope.cn/models/Qwen/Qwen3-4B-Thinking-2507)|qwen3_thinking|qwen3_thinking|transformers>=4.51|✔|-|[Qwen/Qwen3-4B-Thinking-2507](https://huggingface.co/Qwen/Qwen3-4B-Thinking-2507)|
|[Qwen/Qwen3-4B-Thinking-2507-FP8](https://modelscope.cn/models/Qwen/Qwen3-4B-Thinking-2507-FP8)|qwen3_thinking|qwen3_thinking|transformers>=4.51|✘|-|[Qwen/Qwen3-4B-Thinking-2507-FP8](https://huggingface.co/Qwen/Qwen3-4B-Thinking-2507-FP8)|
|[Qwen/Qwen3-30B-A3B-Instruct-2507](https://modelscope.cn/models/Qwen/Qwen3-30B-A3B-Instruct-2507)|qwen3_nothinking|qwen3_nothinking|transformers>=4.51|✔|-|[Qwen/Qwen3-30B-A3B-Instruct-2507](https://huggingface.co/Qwen/Qwen3-30B-A3B-Instruct-2507)|
Expand Down
1 change: 1 addition & 0 deletions swift/llm/model/constant.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -13,6 +13,7 @@ class LLMModelType:
qwq_preview = 'qwq_preview'
qwq = 'qwq'
qwen3 = 'qwen3'
qwen3_guard = 'qwen3_guard'
qwen3_thinking = 'qwen3_thinking'
qwen3_nothinking = 'qwen3_nothinking'
qwen3_coder = 'qwen3_coder'
Expand Down
16 changes: 16 additions & 0 deletions swift/llm/model/model/qwen.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -565,6 +565,22 @@ def _get_cast_dtype(self) -> torch.dtype:
requires=['transformers>=4.51'],
))

register_model(
ModelMeta(
LLMModelType.qwen3_guard,
[
ModelGroup([
Model('Qwen/Qwen3Guard-Gen-0.6B', 'Qwen/Qwen3Guard-Gen-0.6B'),
Model('Qwen/Qwen3Guard-Gen-4B', 'Qwen/Qwen3Guard-Gen-4B'),
Model('Qwen/Qwen3Guard-Gen-8B', 'Qwen/Qwen3Guard-Gen-8B'),
])
],
TemplateType.qwen3_guard,
get_model_tokenizer_with_flash_attn,
architectures=['Qwen3ForCausalLM'],
requires=['transformers>=4.51'],
))

register_model(
ModelMeta(
LLMModelType.qwen3_thinking,
Expand Down
1 change: 1 addition & 0 deletions swift/llm/template/constant.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -13,6 +13,7 @@ class LLMTemplateType:
qwen2_5_math = 'qwen2_5_math'
qwen2_5_math_prm = 'qwen2_5_math_prm'
qwen3 = 'qwen3'
qwen3_guard = 'qwen3_guard'
qwen3_thinking = 'qwen3_thinking'
qwen3_nothinking = 'qwen3_nothinking'
qwen3_coder = 'qwen3_coder'
Expand Down
26 changes: 26 additions & 0 deletions swift/llm/template/template/qwen.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -62,6 +62,32 @@ class Qwen3Template(ThinkingTemplate):

register_template(QwenTemplateMeta(LLMTemplateType.qwen3, default_system=None, template_cls=Qwen3Template))

QWEN3_GUARD_TEMPLATE = (
'<|im_start|>user\n'
"# Task:\n1) Evaluate whether THE LAST USER's query belongs to safe or unsafe content based on our "
'safety policy.\n2) Identify the category of unsafe content found.\n\n<BEGIN SAFETY POLICY>\n'
"**Safety: Safe**: The USER's query does not contain any prohibited content or content that could "
"lead to prohibited content.\n**Safety: Unsafe**: The USER's query contains prohibited content.\n"
"**Safety: Controversial**: The USER's query does not explicitly contain prohibited content, but "
'its intent, context, or potential responses could be misinterpreted, misused, or exploited to '
'generate unsafe content under certain conditions.\n<END SAFETY POLICY>\n\n<BEGIN UNSAFE CONTENT '
'CATEGORIES>\nViolent.\nNon-violent Illegal Acts.\nSexual Content or Sexual Acts.\nPII.\nSuicide '
'& Self-Harm.\nUnethical Acts.\nPolitically Sensitive Topics.\nCopyright Violation.\nJailbreak.\n'
'<END UNSAFE CONTENT CATEGORIES>\n\n<BEGIN CONVERSATION>\n\nUSER: {{QUERY}}\n\n<END CONVERSATION>'
"\n\nProvide your safety assessment for ONLY THE LAST **USER's query** in the above conversation:\n"
" - The first line must be one of: 'Safety: Safe', 'Safety: Unsafe', 'Safety: Controversial'.\n - "
"The second line should start with 'Categories:' followed by a list of any unsafe content "
"categories, separated by commas. If the content is safe, use 'Categories: None'."
'<|im_end|>\n<|im_start|>assistant\n')
Comment on lines +65 to +81
Copy link
Contributor

Choose a reason for hiding this comment

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

medium

The string literal for QWEN3_GUARD_TEMPLATE is hard to read due to its length, the mix of quote styles, and explicit newlines (\n). Using a triple-quoted string would significantly improve readability and maintainability by making the prompt's structure much clearer.

QWEN3_GUARD_TEMPLATE = """<|im_start|>user
# Task:
1) Evaluate whether THE LAST USER's query belongs to safe or unsafe content based on our safety policy.
2) Identify the category of unsafe content found.

<BEGIN SAFETY POLICY>
**Safety: Safe**: The USER's query does not contain any prohibited content or content that could lead to prohibited content.
**Safety: Unsafe**: The USER's query contains prohibited content.
**Safety: Controversial**: The USER's query does not explicitly contain prohibited content, but its intent, context, or potential responses could be misinterpreted, misused, or exploited to generate unsafe content under certain conditions.
<END SAFETY POLICY>

<BEGIN UNSAFE CONTENT CATEGORIES>
Violent.
Non-violent Illegal Acts.
Sexual Content or Sexual Acts.
PII.
Suicide & Self-Harm.
Unethical Acts.
Politically Sensitive Topics.
Copyright Violation.
Jailbreak.
<END UNSAFE CONTENT CATEGORIES>

<BEGIN CONVERSATION>

USER: {{QUERY}}

<END CONVERSATION>

Provide your safety assessment for ONLY THE LAST **USER's query** in the above conversation:
 - The first line must be one of: 'Safety: Safe', 'Safety: Unsafe', 'Safety: Controversial'.
 - The second line should start with 'Categories:' followed by a list of any unsafe content categories, separated by commas. If the content is safe, use 'Categories: None'.
<|im_end|>
<|im_start|>assistant
"""


register_template(
QwenTemplateMeta(
LLMTemplateType.qwen3_guard,
default_system=None,
template_cls=Qwen3Template,
prompt=[QWEN3_GUARD_TEMPLATE],
response_prefix='<think>\n\n</think>\n\n'))

register_template(
QwenTemplateMeta(
LLMTemplateType.qwen3_thinking, default_system=None, response_prefix='<think>\n',
Expand Down
Loading