ros/LLaMA-Factory
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

[model] add qwen3 nothink (#8869)

Browse Source
This commit is contained in:
Yaowei Zheng
2025-08-11 23:17:32 +08:00
committed by GitHub
parent 893edb26d0
commit e695fdfa70
6 changed files with 47 additions and 18 deletions
Download Patch File Download Diff File Expand all files Collapse all files

22
src/llamafactory/data/template.py
View File

@@ -96,7 +96,7 @@ class Template:
def add_thought(self, content: str = "") -> str:
r"""Add empty thought to assistant message."""
return f"{self.thought_words[0]}\n\n{self.thought_words[1]}\n\n" + content
return f"{self.thought_words[0]}{self.thought_words[1]}" + content
def remove_thought(self, content: str) -> str:
r"""Remove thought from assistant message."""
@@ -518,7 +518,7 @@ def register_template(
format_prefix=format_prefix or default_prefix_formatter,
default_system=default_system,
stop_words=stop_words or [],
thought_words=thought_words or ("<think>", "</think>"),
thought_words=thought_words or ("<think>\n", "\n</think>\n\n"),
efficient_eos=efficient_eos,
replace_eos=replace_eos,
replace_jinja_template=replace_jinja_template,
@@ -579,7 +579,7 @@ def parse_template(tokenizer: "PreTrainedTokenizer") -> "Template":
format_prefix=EmptyFormatter(slots=[prefix]) if prefix else EmptyFormatter(),
default_system=default_system,
stop_words=[],
thought_words=("<think>", "</think>"),
thought_words=("<think>\n", "\n</think>\n\n"),
efficient_eos=False,
replace_eos=False,
replace_jinja_template=False,
@@ -1750,6 +1750,22 @@ register_template(
)
# copied from qwen template
register_template(
name="qwen3_nothink",
format_user=StringFormatter(slots=["<|im_start|>user\n{{content}}<|im_end|>\n<|im_start|>assistant\n"]),
format_assistant=StringFormatter(slots=["{{content}}<|im_end|>\n"]),
format_system=StringFormatter(slots=["<|im_start|>system\n{{content}}<|im_end|>\n"]),
format_function=FunctionFormatter(slots=["{{content}}<|im_end|>\n"], tool_format="qwen"),
format_observation=StringFormatter(
slots=["<|im_start|>user\n<tool_response>\n{{content}}\n</tool_response><|im_end|>\n<|im_start|>assistant\n"]
),
format_tools=ToolFormatter(tool_format="qwen"),
stop_words=["<|im_end|>"],
replace_eos=True,
)
# copied from chatml template
register_template(
name="qwen2_audio",