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

[webui] support escape html (#7190)

Browse Source 
Former-commit-id: cf9840374f171359c828b0d6f7a2aa9893c8f701
This commit is contained in:
hoshi-hiyouga
2025-03-06 16:52:21 +08:00
committed by GitHub
parent d2f845d70d
commit 2635794727
5 changed files with 92 additions and 11 deletions
Download Patch File Download Diff File Expand all files Collapse all files

12
src/llamafactory/chat/hf_engine.py
View File

@@ -126,6 +126,7 @@ class HuggingfaceEngine(BaseEngine):
num_return_sequences: int = input_kwargs.pop("num_return_sequences", 1)
repetition_penalty: Optional[float] = input_kwargs.pop("repetition_penalty", None)
length_penalty: Optional[float] = input_kwargs.pop("length_penalty", None)
skip_special_tokens: Optional[bool] = input_kwargs.pop("skip_special_tokens", None)
max_length: Optional[int] = input_kwargs.pop("max_length", None)
max_new_tokens: Optional[int] = input_kwargs.pop("max_new_tokens", None)
stop: Optional[Union[str, List[str]]] = input_kwargs.pop("stop", None)
@@ -145,6 +146,9 @@ class HuggingfaceEngine(BaseEngine):
if repetition_penalty is not None
else generating_args["repetition_penalty"],
length_penalty=length_penalty if length_penalty is not None else generating_args["length_penalty"],
skip_special_tokens=skip_special_tokens
if skip_special_tokens is not None
else generating_args["skip_special_tokens"],
eos_token_id=template.get_stop_token_ids(tokenizer),
pad_token_id=tokenizer.pad_token_id,
)
@@ -241,7 +245,9 @@ class HuggingfaceEngine(BaseEngine):
response_ids = generate_output[:, prompt_length:]
response = tokenizer.batch_decode(
response_ids, skip_special_tokens=generating_args["skip_special_tokens"], clean_up_tokenization_spaces=True
response_ids,
skip_special_tokens=getattr(gen_kwargs["generation_config"], "skip_special_tokens", True),
clean_up_tokenization_spaces=True,
)
results = []
for i in range(len(response)):
@@ -289,7 +295,9 @@ class HuggingfaceEngine(BaseEngine):
input_kwargs,
)
streamer = TextIteratorStreamer(
tokenizer, skip_prompt=True, skip_special_tokens=generating_args["skip_special_tokens"]
tokenizer,
skip_prompt=True,
skip_special_tokens=getattr(gen_kwargs["generation_config"], "skip_special_tokens", True),
)
gen_kwargs["streamer"] = streamer
thread = Thread(target=model.generate, kwargs=gen_kwargs, daemon=True)

5
src/llamafactory/chat/vllm_engine.py
View File

@@ -139,6 +139,7 @@ class VllmEngine(BaseEngine):
num_return_sequences: int = input_kwargs.pop("num_return_sequences", 1)
repetition_penalty: Optional[float] = input_kwargs.pop("repetition_penalty", None)
length_penalty: Optional[float] = input_kwargs.pop("length_penalty", None)
skip_special_tokens: Optional[bool] = input_kwargs.pop("skip_special_tokens", None)
max_length: Optional[int] = input_kwargs.pop("max_length", None)
max_new_tokens: Optional[int] = input_kwargs.pop("max_new_tokens", None)
stop: Optional[Union[str, List[str]]] = input_kwargs.pop("stop", None)
@@ -172,7 +173,9 @@ class VllmEngine(BaseEngine):
stop=stop,
stop_token_ids=self.template.get_stop_token_ids(self.tokenizer),
max_tokens=max_tokens,
skip_special_tokens=self.generating_args["skip_special_tokens"],
skip_special_tokens=skip_special_tokens
if skip_special_tokens is not None
else self.generating_args["skip_special_tokens"],
)
if images is not None: # add image features