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

[Model] Use merge_by_field_config for MM models (Llava family) (#26280)

Browse Source 
Signed-off-by: DarkLight1337 <tlleungac@connect.ust.hk>
This commit is contained in:
Cyrus Leung
2025-10-06 17:45:26 +08:00
committed by GitHub
parent 391612e78b
commit 19a00eb210
9 changed files with 155 additions and 229 deletions
Download Patch File Download Diff File Expand all files Collapse all files

218
examples/offline_inference/vision_language_multi_image.py
View File

@@ -371,6 +371,115 @@ def load_internvl(question: str, image_urls: list[str]) -> ModelRequestData:
)
def load_keye_vl(question: str, image_urls: list[str]) -> ModelRequestData:
model_name = "Kwai-Keye/Keye-VL-8B-Preview"
engine_args = EngineArgs(
model=model_name,
trust_remote_code=True,
max_model_len=8192,
max_num_seqs=5,
limit_mm_per_prompt={"image": len(image_urls)},
)
placeholders = [{"type": "image", "image": url} for url in image_urls]
messages = [
{
"role": "user",
"content": [
*placeholders,
{"type": "text", "text": question},
],
},
]
processor = AutoProcessor.from_pretrained(model_name, trust_remote_code=True)
prompt = processor.apply_chat_template(
messages, tokenize=False, add_generation_prompt=True
)
image_data = [fetch_image(url) for url in image_urls]
return ModelRequestData(
engine_args=engine_args,
prompt=prompt,
image_data=image_data,
)
def load_keye_vl1_5(question: str, image_urls: list[str]) -> ModelRequestData:
model_name = "Kwai-Keye/Keye-VL-1_5-8B"
engine_args = EngineArgs(
model=model_name,
trust_remote_code=True,
max_model_len=32768,
max_num_seqs=5,
limit_mm_per_prompt={"image": len(image_urls)},
)
placeholders = [{"type": "image", "image": url} for url in image_urls]
messages = [
{
"role": "user",
"content": [
*placeholders,
{"type": "text", "text": question},
],
},
]
processor = AutoProcessor.from_pretrained(model_name, trust_remote_code=True)
prompt = processor.apply_chat_template(
messages, tokenize=False, add_generation_prompt=True
)
image_data = [fetch_image(url) for url in image_urls]
return ModelRequestData(
engine_args=engine_args,
prompt=prompt,
image_data=image_data,
)
def load_kimi_vl(question: str, image_urls: list[str]) -> ModelRequestData:
model_name = "moonshotai/Kimi-VL-A3B-Instruct"
engine_args = EngineArgs(
model=model_name,
trust_remote_code=True,
max_model_len=4096,
max_num_seqs=4,
limit_mm_per_prompt={"image": len(image_urls)},
)
placeholders = [{"type": "image", "image": url} for url in image_urls]
messages = [
{
"role": "user",
"content": [
*placeholders,
{"type": "text", "text": question},
],
}
]
processor = AutoProcessor.from_pretrained(model_name, trust_remote_code=True)
prompt = processor.apply_chat_template(
messages, tokenize=False, add_generation_prompt=True
)
return ModelRequestData(
engine_args=engine_args,
prompt=prompt,
image_data=[fetch_image(url) for url in image_urls],
)
def load_llama4(question: str, image_urls: list[str]) -> ModelRequestData:
model_name = "meta-llama/Llama-4-Scout-17B-16E-Instruct"
@@ -505,115 +614,6 @@ def load_llava_onevision(question: str, image_urls: list[str]) -> ModelRequestDa
)
def load_keye_vl(question: str, image_urls: list[str]) -> ModelRequestData:
model_name = "Kwai-Keye/Keye-VL-8B-Preview"
engine_args = EngineArgs(
model=model_name,
trust_remote_code=True,
max_model_len=8192,
max_num_seqs=5,
limit_mm_per_prompt={"image": len(image_urls)},
)
placeholders = [{"type": "image", "image": url} for url in image_urls]
messages = [
{
"role": "user",
"content": [
*placeholders,
{"type": "text", "text": question},
],
},
]
processor = AutoProcessor.from_pretrained(model_name, trust_remote_code=True)
prompt = processor.apply_chat_template(
messages, tokenize=False, add_generation_prompt=True
)
image_data = [fetch_image(url) for url in image_urls]
return ModelRequestData(
engine_args=engine_args,
prompt=prompt,
image_data=image_data,
)
def load_keye_vl1_5(question: str, image_urls: list[str]) -> ModelRequestData:
model_name = "Kwai-Keye/Keye-VL-1_5-8B"
engine_args = EngineArgs(
model=model_name,
trust_remote_code=True,
max_model_len=32768,
max_num_seqs=5,
limit_mm_per_prompt={"image": len(image_urls)},
)
placeholders = [{"type": "image", "image": url} for url in image_urls]
messages = [
{
"role": "user",
"content": [
*placeholders,
{"type": "text", "text": question},
],
},
]
processor = AutoProcessor.from_pretrained(model_name, trust_remote_code=True)
prompt = processor.apply_chat_template(
messages, tokenize=False, add_generation_prompt=True
)
image_data = [fetch_image(url) for url in image_urls]
return ModelRequestData(
engine_args=engine_args,
prompt=prompt,
image_data=image_data,
)
def load_kimi_vl(question: str, image_urls: list[str]) -> ModelRequestData:
model_name = "moonshotai/Kimi-VL-A3B-Instruct"
engine_args = EngineArgs(
model=model_name,
trust_remote_code=True,
max_model_len=4096,
max_num_seqs=4,
limit_mm_per_prompt={"image": len(image_urls)},
)
placeholders = [{"type": "image", "image": url} for url in image_urls]
messages = [
{
"role": "user",
"content": [
*placeholders,
{"type": "text", "text": question},
],
}
]
processor = AutoProcessor.from_pretrained(model_name, trust_remote_code=True)
prompt = processor.apply_chat_template(
messages, tokenize=False, add_generation_prompt=True
)
return ModelRequestData(
engine_args=engine_args,
prompt=prompt,
image_data=[fetch_image(url) for url in image_urls],
)
def load_mistral3(question: str, image_urls: list[str]) -> ModelRequestData:
model_name = "mistralai/Mistral-Small-3.1-24B-Instruct-2503"