minke.yu/sglang
1
0
Fork 0
Code Issues Pull Requests Actions 13 Packages Projects Releases Wiki Activity

[Feature] Support file:// URL format for multimodal inputs (#14490)

Browse Source 
Co-authored-by: Yuhao Yang <47235274+yhyang201@users.noreply.github.com>
This commit is contained in:
Praneth Paruchuri
2026-01-31 23:44:44 -08:00
committed by GitHub
co-authored by Yuhao Yang
parent 71babdef51
commit 9bb1260558
1 changed files with 15 additions and 2 deletions
Download Patch File Download Diff File Expand all files Collapse all files
python/sglang/srt/utils/common.py
+15 -2
View File
@@ -70,7 +70,7 @@ from typing import (
Union, Union,
) )
from unittest import SkipTest from unittest import SkipTest
from urllib.parse import urlparse from urllib.parse import unquote, urlparse
import numpy as np import numpy as np
import orjson import orjson
@@ -860,6 +860,9 @@ def load_audio(
audio_file = BytesIO(response.content) audio_file = BytesIO(response.content)
response.close() response.close()
audio, original_sr = sf.read(audio_file) audio, original_sr = sf.read(audio_file)
elif audio_file.startswith("file://"):
audio_file = unquote(urlparse(audio_file).path)
audio, original_sr = sf.read(audio_file)
elif isinstance(audio_file, str): elif isinstance(audio_file, str):
audio, original_sr = sf.read(audio_file) audio, original_sr = sf.read(audio_file)
else: else:
@@ -905,6 +908,9 @@ def load_image(
image.load() # Force loading to avoid issues after closing the stream image.load() # Force loading to avoid issues after closing the stream
finally: finally:
response.close() response.close()
elif image_file.startswith("file://"):
image_file = unquote(urlparse(image_file).path)
image = Image.open(image_file)
elif image_file.lower().endswith(("png", "jpg", "jpeg", "webp", "gif")): elif image_file.lower().endswith(("png", "jpg", "jpeg", "webp", "gif")):
image = Image.open(image_file) image = Image.open(image_file)
elif image_file.startswith("data:"): elif image_file.startswith("data:"):
@@ -925,6 +931,10 @@ def get_image_bytes(image_file: Union[str, bytes]):
timeout = int(os.getenv("REQUEST_TIMEOUT", "3")) timeout = int(os.getenv("REQUEST_TIMEOUT", "3"))
response = requests.get(image_file, timeout=timeout) response = requests.get(image_file, timeout=timeout)
return response.content return response.content
elif image_file.startswith("file://"):
image_file = unquote(urlparse(image_file).path)
with open(image_file, "rb") as f:
return f.read()
elif image_file.lower().endswith(("png", "jpg", "jpeg", "webp", "gif")): elif image_file.lower().endswith(("png", "jpg", "jpeg", "webp", "gif")):
with open(image_file, "rb") as f: with open(image_file, "rb") as f:
return f.read() return f.read()
@@ -974,8 +984,11 @@ def load_video(video_file: Union[str, bytes], use_gpu: bool = True):
tmp_file.write(video_bytes) tmp_file.write(video_bytes)
tmp_file.close() tmp_file.close()
vr = VideoReader(tmp_file.name, ctx=ctx) vr = VideoReader(tmp_file.name, ctx=ctx)
elif video_file.startswith("file://"):
video_file = unquote(urlparse(video_file).path)
vr = VideoReader(video_file, ctx=ctx)
# `urlparse` supports file:// paths, and so does VideoReader # `urlparse` supports file:// paths, and so does VideoReader
elif os.path.isfile(urlparse(video_file).path): elif os.path.isfile(unquote(urlparse(video_file).path)):
vr = VideoReader(video_file, ctx=ctx) vr = VideoReader(video_file, ctx=ctx)
else: else:
video_bytes = pybase64.b64decode(video_file, validate=True) video_bytes = pybase64.b64decode(video_file, validate=True)