@tduncan / Nvidia-Optimization-Stack-VLLM Mirror · read-only / commits / 8b6725b0cf
Fork 0

A high-throughput and memory-efficient inference and serving engine for LLMs

Mirrored from https://github.com/tduncan-lum/Nvidia-Optimization-Stack---VLLM.git

Synced Nov 16, 07:00 UTC

[Misc] Update benchmark to support image_url file or http (#10287)

Signed-off-by: rbbang <anjaehyun87@gmail.com>

Jaehyun An committed Nov 16, 2024 at 19:15 UTC 8b6725b0cf4ee5f363218f4bc341970c80297ccf
1 file changed +13
benchmarks/benchmark_serving.py
+13
@@ -251,6 +251,19 @@ def sample_hf_requests(
251 "url": f"data:image/jpeg;base64,{image_base64}"
252 },
253 }
254 + elif "image" in data and isinstance(data["image"], str):
255 + if (data["image"].startswith("http://") or \
256 + data["image"].startswith("file://")):
257 + image_url = data["image"]
258 + else:
259 + image_url = f"file://{data['image']}"
260 +
261 + mm_content = {
262 + "type": "image_url",
263 + "image_url": {
264 + "url": image_url
265 + },
266 + }
267 else:
268 mm_content = None
269