chenchenghao/sglang
1
0
Fork 0
Code Issues Pull Requests Actions 11 Packages Projects Releases Wiki Activity

[Diffusion] add mova and hunyuanvideo to perf skills (#20563)

Browse Source
This commit is contained in:
Xiaoyu Zhang
2026-03-14 13:49:50 +08:00
committed by GitHub
parent 99a3b25c9b
commit f9e4221b71
2 changed files with 68 additions and 3 deletions
Download Patch File Download Diff File Expand all files Collapse all files

28
python/sglang/multimodal_gen/.claude/skills/diffusion-kernel/diffusion-benchmark-and-profile.md
View File

@@ -55,6 +55,8 @@ wget -O "${ASSET_DIR}/cat.png" \
https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/diffusers/cat.png
wget -O "${ASSET_DIR}/astronaut.jpg" \
https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/diffusers/astronaut.jpg
wget -O "${ASSET_DIR}/mova_single_person.jpg" \
https://github.com/OpenMOSS/MOVA/raw/main/assets/single_person.jpg
```
---
@@ -164,6 +166,32 @@ sglang generate \
--enable-torch-compile --warmup
```
### HunyuanVideo (848×480, 65 frames, 30 steps)
```bash
sglang generate \
--model-path=hunyuanvideo-community/HunyuanVideo \
--text-encoder-cpu-offload --pin-cpu-memory \
--prompt="A cat and a dog baking a cake together in a kitchen. The cat is carefully measuring flour, while the dog is stirring the batter with a wooden spoon. The kitchen is cozy, with sunlight streaming through the window." \
--save-output --num-frames=65 --width=848 --height=480 \
--num-inference-steps=30 \
--warmup --enable-torch-compile
```
### MOVA-720p (4 GPUs, 193 frames, 24 steps)
```bash
# Select four idle GPUs first:
# export CUDA_VISIBLE_DEVICES=$(python3 "$ENV_PY" print-idle-gpus --count 4)
sglang generate \
--model-path=OpenMOSS-Team/MOVA-720p \
--prompt="A man in a blue blazer and glasses speaks in a formal indoor setting, framed by wooden furniture and a filled bookshelf. Quiet room acoustics underscore his measured tone as he delivers his remarks. At one point, he says, \"I would also believe that this advance in AI recently wasn’t unexpected.\"" \
--image-path="${ASSET_DIR}/mova_single_person.jpg" \
--adjust-frames=false \
--num-gpus=4 --ring-degree=1 --ulysses-degree=4 \
--num-frames=193 --fps=24 \
--num-inference-steps=24 \
--enable-torch-compile --save-output --warmup
```
**Key metrics** (all models): denoise latency ★, end-to-end latency, peak GPU memory.
---

43
python/sglang/multimodal_gen/.claude/skills/diffusion-kernel/scripts/bench_diffusion_denoise.py
View File

@@ -17,7 +17,7 @@ Usage:
# Side-by-side comparison
python3 python/sglang/multimodal_gen/.claude/skills/diffusion-kernel/scripts/bench_diffusion_denoise.py --model flux --compare
# All 7 models, comparison
# All 9 models, comparison
python3 python/sglang/multimodal_gen/.claude/skills/diffusion-kernel/scripts/bench_diffusion_denoise.py --all --compare
Input images required for image-guided models:
@@ -26,6 +26,8 @@ Input images required for image-guided models:
https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/diffusers/cat.png
wget -O "${ASSET_DIR}/astronaut.jpg" \
https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/diffusers/astronaut.jpg
wget -O "${ASSET_DIR}/mova_single_person.jpg" \
https://github.com/OpenMOSS/MOVA/raw/main/assets/single_person.jpg
"""
import argparse
@@ -182,11 +184,44 @@ MODELS = {
"false",
],
},
# 8. hunyuanvideo-community/HunyuanVideo — Text-to-Video, 848×480, 65 frames, 30 steps
"hunyuanvideo": {
"path": "hunyuanvideo-community/HunyuanVideo",
"prompt": "A cat and a dog baking a cake together in a kitchen. The cat is carefully measuring flour, while the dog is stirring the batter with a wooden spoon. The kitchen is cozy, with sunlight streaming through the window.",
"extra_args": [
"--text-encoder-cpu-offload",
"--pin-cpu-memory",
"--num-frames=65",
"--width=848",
"--height=480",
"--num-inference-steps=30",
],
},
# 9. OpenMOSS-Team/MOVA-720p — Image-to-Video, 4 GPUs, 193 frames, 24 steps
# Requires: <repo>/inputs/diffusion_benchmark/figs/mova_single_person.jpg
"mova-720p": {
"path": "OpenMOSS-Team/MOVA-720p",
"prompt": 'A man in a blue blazer and glasses speaks in a formal indoor setting, framed by wooden furniture and a filled bookshelf. Quiet room acoustics underscore his measured tone as he delivers his remarks. At one point, he says, "I would also believe that this advance in AI recently was not unexpected."',
"image_path": str(ASSET_DIR / "mova_single_person.jpg"),
"extra_args": [
"--adjust-frames=false",
"--num-gpus=4",
"--ring-degree=1",
"--ulysses-degree=4",
"--num-frames=193",
"--fps=24",
"--num-inference-steps=24",
],
},
}
def required_gpus_for_model(model_key: str) -> int:
return 8 if model_key == "wan-t2v" else 1
if model_key == "wan-t2v":
return 8
if model_key == "mova-720p":
return 4
return 1
def build_sglang_cmd(
@@ -209,10 +244,12 @@ def build_sglang_cmd(
"generate",
f"--model-path={cfg['path']}",
f"--prompt={cfg['prompt']}",
f"--seed={seed}",
"--log-level=info",
]
if seed is not None:
cmd.append(f"--seed={seed}")
if "negative_prompt" in cfg:
cmd.append(f"--negative-prompt={cfg['negative_prompt']}")