🔥Qwen-Image 4.8x speedup w/ Hybrid Acceleration for low VRAM GPUs

We have release a hybrid acceleration example for qwen-image, feel free to take a try (Cache + Context Parallelism + FP8 Weight Only + Torch Compile), for example: (Also check qwen-image-fast)

# Baseline (NVIDIA L20 48GiB, ~120s w/ CPU Offload)
python3 run_qwen_image_cp.py --height 1024 --width 1024

# + (DBCache + TaylorSeer) 
# + Context Parallelism (Ulysses)
# + FP8 Weight Only (not require offload anymore) 
# + Torch Compile (NVIDIA L20x2, ~25s, ~4.8x speedup)
torchrun --nproc_per_node=2 run_qwen_image_cp.py \
         --height 1024 --width 1024 \
         --parallel-type ulysses --quantize \
         --cache --Fn 1 --rdt 0.12 --mcc 2 --taylorseer \
         --compile

• Baseline: 120s

• Hybrid Acceleration: 25s