Skip to content
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

[Misc] Use NamedTuple in Multi-image example #8705

Merged
merged 3 commits into from
Sep 22, 2024
Merged

[Misc] Use NamedTuple in Multi-image example #8705

Changes from 2 commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
9d026e0
Use named tuple in multi-image example
alex-jw-brooks Sep 20, 2024
6db301f
Add load model return type annotations
alex-jw-brooks Sep 22, 2024
2bed891
Use typing NamedTuple instead of collections
alex-jw-brooks Sep 22, 2024
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
68 changes: 47 additions & 21 deletions examples/offline_inference_vision_language_multi_image.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -4,6 +4,7 @@
by the model.
"""
from argparse import Namespace
from collections import namedtuple
from typing import List

from transformers import AutoProcessor, AutoTokenizer
Expand All @@ -18,8 +19,12 @@
"https://upload.wikimedia.org/wikipedia/commons/7/77/002_The_lion_king_Snyggve_in_the_Serengeti_National_Park_Photo_by_Giles_Laurent.jpg",
]

ModelRequestData = namedtuple(
"ModelRequestData",
["llm", "prompt", "stop_token_ids", "image_data", "chat_template"])

def load_qwenvl_chat(question: str, image_urls: List[str]):

def load_qwenvl_chat(question: str, image_urls: List[str]) -> ModelRequestData:
model_name = "Qwen/Qwen-VL-Chat"
llm = LLM(
model=model_name,
Expand Down Expand Up @@ -48,10 +53,16 @@ def load_qwenvl_chat(question: str, image_urls: List[str]):

stop_tokens = ["<|endoftext|>", "<|im_start|>", "<|im_end|>"]
stop_token_ids = [tokenizer.convert_tokens_to_ids(i) for i in stop_tokens]
return llm, prompt, stop_token_ids, None, chat_template
return ModelRequestData(
llm=llm,
prompt=prompt,
stop_token_ids=stop_token_ids,
image_data=[fetch_image(url) for url in image_urls],
chat_template=chat_template,
)


def load_phi3v(question: str, image_urls: List[str]):
def load_phi3v(question: str, image_urls: List[str]) -> ModelRequestData:
llm = LLM(
model="microsoft/Phi-3.5-vision-instruct",
trust_remote_code=True,
Expand All @@ -62,10 +73,17 @@ def load_phi3v(question: str, image_urls: List[str]):
for i, _ in enumerate(image_urls, start=1))
prompt = f"<|user|>\n{placeholders}\n{question}<|end|>\n<|assistant|>\n"
stop_token_ids = None
return llm, prompt, stop_token_ids, None, None

return ModelRequestData(
llm=llm,
prompt=prompt,
stop_token_ids=stop_token_ids,
image_data=[fetch_image(url) for url in image_urls],
chat_template=None,
)


def load_internvl(question: str, image_urls: List[str]):
def load_internvl(question: str, image_urls: List[str]) -> ModelRequestData:
model_name = "OpenGVLab/InternVL2-2B"

llm = LLM(
Expand Down Expand Up @@ -93,10 +111,16 @@ def load_internvl(question: str, image_urls: List[str]):
stop_tokens = ["<|endoftext|>", "<|im_start|>", "<|im_end|>", "<|end|>"]
stop_token_ids = [tokenizer.convert_tokens_to_ids(i) for i in stop_tokens]

return llm, prompt, stop_token_ids, None, None
return ModelRequestData(
llm=llm,
prompt=prompt,
stop_token_ids=stop_token_ids,
image_data=[fetch_image(url) for url in image_urls],
chat_template=None,
)


def load_qwen2_vl(question, image_urls: List[str]):
def load_qwen2_vl(question, image_urls: List[str]) -> ModelRequestData:
try:
from qwen_vl_utils import process_vision_info
except ModuleNotFoundError:
Expand Down Expand Up @@ -143,7 +167,13 @@ def load_qwen2_vl(question, image_urls: List[str]):
else:
image_data, _ = process_vision_info(messages)

return llm, prompt, stop_token_ids, image_data, None
return ModelRequestData(
llm=llm,
prompt=prompt,
stop_token_ids=stop_token_ids,
image_data=image_data,
chat_template=None,
)


model_example_map = {
Expand All @@ -155,20 +185,17 @@ def load_qwen2_vl(question, image_urls: List[str]):


def run_generate(model, question: str, image_urls: List[str]):
llm, prompt, stop_token_ids, image_data, _ = model_example_map[model](
question, image_urls)
if image_data is None:
image_data = [fetch_image(url) for url in image_urls]
req_data = model_example_map[model](question, image_urls)

sampling_params = SamplingParams(temperature=0.0,
max_tokens=128,
stop_token_ids=stop_token_ids)
stop_token_ids=req_data.stop_token_ids)

outputs = llm.generate(
outputs = req_data.llm.generate(
{
"prompt": prompt,
"prompt": req_data.prompt,
"multi_modal_data": {
"image": image_data
"image": req_data.image_data
},
},
sampling_params=sampling_params)
Expand All @@ -179,13 +206,12 @@ def run_generate(model, question: str, image_urls: List[str]):


def run_chat(model: str, question: str, image_urls: List[str]):
llm, _, stop_token_ids, _, chat_template = model_example_map[model](
question, image_urls)
req_data = model_example_map[model](question, image_urls)

sampling_params = SamplingParams(temperature=0.0,
max_tokens=128,
stop_token_ids=stop_token_ids)
outputs = llm.chat(
stop_token_ids=req_data.stop_token_ids)
outputs = req_data.llm.chat(
[{
"role":
"user",
Expand All @@ -203,7 +229,7 @@ def run_chat(model: str, question: str, image_urls: List[str]):
],
}],
sampling_params=sampling_params,
chat_template=chat_template,
chat_template=req_data.chat_template,
)

for o in outputs:
Expand Down
Loading