biondizzle/vllm
1
0
Fork 0
Code Issues Pull Requests Actions 2 Packages Projects Releases Wiki Activity

Convert examples to ruff-format (#18400)

Browse Source 
Signed-off-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>
This commit is contained in:
Harry Mellor
2025-05-26 17:57:54 +01:00
committed by GitHub
parent e7523c2e03
commit 27bebcd897
83 changed files with 2529 additions and 2405 deletions
Download Patch File Download Diff File Expand all files Collapse all files

71
examples/offline_inference/vision_language_embedding.py
View File

@@ -6,6 +6,7 @@ the correct prompt format on vision language models for multimodal embedding.
For most models, the prompt format should follow corresponding examples
on HuggingFace model repository.
"""
from argparse import Namespace
from dataclasses import asdict
from typing import Literal, NamedTuple, Optional, TypedDict, Union, get_args
@@ -44,19 +45,17 @@ class ModelRequestData(NamedTuple):
def run_e5_v(query: Query) -> ModelRequestData:
llama3_template = '<|start_header_id|>user<|end_header_id|>\n\n{}<|eot_id|><|start_header_id|>assistant<|end_header_id|>\n\n \n' # noqa: E501
llama3_template = "<|start_header_id|>user<|end_header_id|>\n\n{}<|eot_id|><|start_header_id|>assistant<|end_header_id|>\n\n \n" # noqa: E501
if query["modality"] == "text":
text = query["text"]
prompt = llama3_template.format(
f"{text}\nSummary above sentence in one word: ")
prompt = llama3_template.format(f"{text}\nSummary above sentence in one word: ")
image = None
elif query["modality"] == "image":
prompt = llama3_template.format(
"<image>\nSummary above image in one word: ")
prompt = llama3_template.format("<image>\nSummary above image in one word: ")
image = query["image"]
else:
modality = query['modality']
modality = query["modality"]
raise ValueError(f"Unsupported query modality: '{modality}'")
engine_args = EngineArgs(
@@ -83,10 +82,12 @@ def run_vlm2vec(query: Query) -> ModelRequestData:
image = query["image"]
elif query["modality"] == "text+image":
text = query["text"]
prompt = f"<|image_1|> Represent the given image with the following question: {text}" # noqa: E501
prompt = (
f"<|image_1|> Represent the given image with the following question: {text}" # noqa: E501
)
image = query["image"]
else:
modality = query['modality']
modality = query["modality"]
raise ValueError(f"Unsupported query modality: '{modality}'")
engine_args = EngineArgs(
@@ -136,7 +137,8 @@ def run_encode(model: str, modality: QueryModality, seed: Optional[int]):
# Disable other modalities to save memory
default_limits = {"image": 0, "video": 0, "audio": 0}
req_data.engine_args.limit_mm_per_prompt = default_limits | dict(
req_data.engine_args.limit_mm_per_prompt or {})
req_data.engine_args.limit_mm_per_prompt or {}
)
engine_args = asdict(req_data.engine_args) | {"seed": seed}
llm = LLM(**engine_args)
@@ -145,10 +147,12 @@ def run_encode(model: str, modality: QueryModality, seed: Optional[int]):
if req_data.image is not None:
mm_data["image"] = req_data.image
outputs = llm.embed({
"prompt": req_data.prompt,
"multi_modal_data": mm_data,
})
outputs = llm.embed(
{
"prompt": req_data.prompt,
"multi_modal_data": mm_data,
}
)
print("-" * 50)
for output in outputs:
@@ -164,23 +168,30 @@ model_example_map = {
def parse_args():
parser = FlexibleArgumentParser(
description='Demo on using vLLM for offline inference with '
'vision language models for multimodal embedding')
parser.add_argument('--model-name',
'-m',
type=str,
default="vlm2vec",
choices=model_example_map.keys(),
help='The name of the embedding model.')
parser.add_argument('--modality',
type=str,
default="image",
choices=get_args(QueryModality),
help='Modality of the input.')
parser.add_argument("--seed",
type=int,
default=None,
help="Set the seed when initializing `vllm.LLM`.")
description="Demo on using vLLM for offline inference with "
"vision language models for multimodal embedding"
)
parser.add_argument(
"--model-name",
"-m",
type=str,
default="vlm2vec",
choices=model_example_map.keys(),
help="The name of the embedding model.",
)
parser.add_argument(
"--modality",
type=str,
default="image",
choices=get_args(QueryModality),
help="Modality of the input.",
)
parser.add_argument(
"--seed",
type=int,
default=None,
help="Set the seed when initializing `vllm.LLM`.",
)
return parser.parse_args()