[Hardware] Support platforms and plugin system (#774)

Signed-off-by: gcanlin <canlinguosdu@gmail.com>
This commit is contained in:
Canlin Guo authored and GitHub committed 2026-01-28 22:12:52 +08:00
1 parent b11d436e19
commit c9392dff01
103 files changed
+1770 -1218

No files matched your search

@@ -81,7 +81,7 @@ from vllm_omni.diffusion.data import DiffusionParallelConfig, logger
from vllm_omni.entrypoints.omni import Omni
from vllm_omni.inputs.data import OmniDiffusionSamplingParams
from vllm_omni.outputs import OmniRequestOutput
from vllm_omni.utils.platform_utils import detect_device_type, is_npu
from vllm_omni.platforms import current_omni_platform
def parse_args() -> argparse.Namespace:
@@ -280,6 +280,16 @@ def parse_args() -> argparse.Namespace:
action="store_true",
help="Disable torch.compile and force eager execution.",
)
parser.add_argument(
"--vae_use_slicing",
action="store_true",
help="Enable VAE slicing for memory optimization.",
)
parser.add_argument(
"--vae_use_tiling",
action="store_true",
help="Enable VAE tiling for memory optimization.",
)
parser.add_argument(
"--enable-cpu-offload",
action="store_true",
@@ -306,12 +316,8 @@ def main():
else:
input_image = input_images
device = detect_device_type()
generator = torch.Generator(device=device).manual_seed(args.seed)
generator = torch.Generator(device=current_omni_platform.device_type).manual_seed(args.seed)
# Enable VAE memory optimizations on NPU
vae_use_slicing = is_npu()
vae_use_tiling = is_npu()
parallel_config = DiffusionParallelConfig(
ulysses_degree=args.ulysses_degree,
ring_degree=args.ring_degree,
@@ -344,8 +350,8 @@ def main():
# Initialize Omni with appropriate pipeline
omni = Omni(
model=args.model,
vae_use_slicing=vae_use_slicing,
vae_use_tiling=vae_use_tiling,
vae_use_slicing=args.vae_use_slicing,
vae_use_tiling=args.vae_use_tiling,
cache_backend=args.cache_backend,
cache_config=cache_config,
parallel_config=parallel_config,
@@ -47,4 +47,8 @@ Key arguments:
- `--guidance_scale`: guidance scale for guidance-distilled models (default: 1.0, disabled). Unlike classifier-free guidance (--cfg_scale), guidance-distilled models take the guidance scale directly as an input parameter. Enabled when guidance_scale > 1. Ignored when not using guidance-distilled models.
- `--num_inference_steps`: diffusion sampling steps (more steps = higher quality, slower).
- `--output`: path to save the generated PNG.
- `--vae_use_slicing`: enable VAE slicing for memory optimization.
- `--vae_use_tiling`: enable VAE tiling for memory optimization.
- `--enable-cpu-offload`: enable CPU offloading for diffusion models.
> ℹ️ If you encounter OOM errors, try using `--vae_use_slicing` and `--vae_use_tiling` to reduce memory usage.