423A35C7/LLaMA-Factory
1
0
Fork 0
mirror of https://github.com/hiyouga/LLaMA-Factory.git synced 2025-12-17 12:20:37 +08:00
Code Issues Packages Projects Releases Wiki Activity

support multiimage inference

Browse Source
This commit is contained in:
hiyouga
2024-11-01 07:25:20 +00:00
parent 641d0dab08
commit e80a481927
7 changed files with 103 additions and 63 deletions
Download Patch File Download Diff File Expand all files Collapse all files

36
src/llamafactory/chat/vllm_engine.py
View File

@@ -101,14 +101,14 @@ class VllmEngine(BaseEngine):
messages: Sequence[Dict[str, str]],
system: Optional[str] = None,
tools: Optional[str] = None,
image: Optional["ImageInput"] = None,
video: Optional["VideoInput"] = None,
images: Optional[Sequence["ImageInput"]] = None,
videos: Optional[Sequence["VideoInput"]] = None,
**input_kwargs,
) -> AsyncIterator["RequestOutput"]:
request_id = f"chatcmpl-{uuid.uuid4().hex}"
if image is not None:
if images is not None:
if IMAGE_PLACEHOLDER not in messages[0]["content"]:
messages[0]["content"] = IMAGE_PLACEHOLDER + messages[0]["content"]
messages[0]["content"] = IMAGE_PLACEHOLDER * len(images) + messages[0]["content"]
paired_messages = messages + [{"role": "assistant", "content": ""}]
system = system or self.generating_args["default_system"]
@@ -157,14 +157,18 @@ class VllmEngine(BaseEngine):
skip_special_tokens=True,
)
if image is not None: # add image features
if not isinstance(image, (str, ImageObject)):
raise ValueError(f"Expected image input is a path or PIL.Image, but got {type(image)}.")
if images is not None: # add image features
image_data = []
for image in images:
if not isinstance(image, (str, ImageObject)):
raise ValueError(f"Expected image input is a path or PIL.Image, but got {type(image)}.")
if isinstance(image, str):
image = Image.open(image).convert("RGB")
if isinstance(image, str):
image = Image.open(image).convert("RGB")
multi_modal_data = {"image": image}
image_data.append(image)
multi_modal_data = {"image": image_data}
else:
multi_modal_data = None
@@ -182,12 +186,12 @@ class VllmEngine(BaseEngine):
messages: Sequence[Dict[str, str]],
system: Optional[str] = None,
tools: Optional[str] = None,
image: Optional["ImageInput"] = None,
video: Optional["VideoInput"] = None,
images: Optional[Sequence["ImageInput"]] = None,
videos: Optional[Sequence["VideoInput"]] = None,
**input_kwargs,
) -> List["Response"]:
final_output = None
generator = await self._generate(messages, system, tools, image, video, **input_kwargs)
generator = await self._generate(messages, system, tools, images, videos, **input_kwargs)
async for request_output in generator:
final_output = request_output
@@ -210,12 +214,12 @@ class VllmEngine(BaseEngine):
messages: Sequence[Dict[str, str]],
system: Optional[str] = None,
tools: Optional[str] = None,
image: Optional["ImageInput"] = None,
video: Optional["VideoInput"] = None,
images: Optional[Sequence["ImageInput"]] = None,
videos: Optional[Sequence["VideoInput"]] = None,
**input_kwargs,
) -> AsyncGenerator[str, None]:
generated_text = ""
generator = await self._generate(messages, system, tools, image, video, **input_kwargs)
generator = await self._generate(messages, system, tools, images, videos, **input_kwargs)
async for result in generator:
delta_text = result.outputs[0].text[len(generated_text) :]
generated_text = result.outputs[0].text