EngineX-Hygon/sglang
5
0
Fork 0
Code Issues Pull Requests Actions 7 Projects Releases Wiki Activity

[CPU] Fix phi4-mm prompt issue in bench_serving (#9900)

Browse Source
This commit is contained in:
blzheng
2025-09-09 11:12:32 +08:00
committed by GitHub
parent ba066ca02f
commit 97fff98c68
1 changed files with 15 additions and 7 deletions
Download Patch File Download Diff File Expand all files Collapse all files

20
python/sglang/bench_serving.py
View File

@@ -995,17 +995,25 @@ def sample_mmmu_requests(
prompt = f"Question: {question}\n\nAnswer: " prompt = f"Question: {question}\n\nAnswer: "
if apply_chat_template: if apply_chat_template:
try: try:
prompt = tokenizer.apply_chat_template( is_phi4_multimodal = (
[ "phi-4-multimodal" in tokenizer.name_or_path.lower()
{ )
"role": "user", if is_phi4_multimodal:
"content": [ # <|endoftext10|> is the image token used in the phi-4-multimodal model.
content = prompt.replace("image 1", "<|endoftext10|>")
else:
content = [
{ {
"type": "image_url", "type": "image_url",
"image_url": {"url": image_data}, "image_url": {"url": image_data},
}, },
{"type": "text", "text": prompt}, {"type": "text", "text": prompt},
], ]
prompt = tokenizer.apply_chat_template(
[
{
"role": "user",
"content": content,
} }
], ],
add_generation_prompt=True, add_generation_prompt=True,