[Model] Fused mm preprocess normalisation on the Device (#50411)

Signed-off-by: wang.yuqi <yuqi.wang@daocloud.io>
Signed-off-by: wang.yuqi <noooop@126.com>
Co-authored-by: Cyrus Leung <cyrus.tl.leung@gmail.com>
This commit is contained in:
wang.yuqi
2026-08-06 03:11:14 -07:00
committed by GitHub
co-authored by Cyrus Leung
parent 9c22668436
commit 2fa490470d
13 changed files with 359 additions and 11 deletions
+9
View File
@@ -599,6 +599,7 @@ class EngineArgs:
mm_tensor_ipc: MMTensorIPC = MultiModalConfig.mm_tensor_ipc
mm_processor_device: MMProcessorDevice = "auto"
mm_ipc_gpu_memory_gb: float = MultiModalConfig.mm_ipc_gpu_memory_gb
mm_device_do_normalize: bool | None = MultiModalConfig.mm_device_do_normalize
# LoRA fields
enable_lora: bool = False
max_loras: int = LoRAConfig.max_loras
@@ -1383,6 +1384,13 @@ class EngineArgs:
"--mm-ipc-gpu-memory-gb",
**multimodal_kwargs["mm_ipc_gpu_memory_gb"],
)
multimodal_group.add_argument(
"--mm-device-do-normalize",
**{
**multimodal_kwargs["mm_device_do_normalize"],
"default": None,
},
)
# LoRA related configs
lora_kwargs = get_kwargs(LoRAConfig)
@@ -1769,6 +1777,7 @@ class EngineArgs:
video_pruning_method=self.video_pruning_method,
mm_tensor_ipc=self.mm_tensor_ipc,
mm_ipc_gpu_memory_gb=self.mm_ipc_gpu_memory_gb,
mm_device_do_normalize=self.mm_device_do_normalize,
mm_processor_device=self.mm_processor_device,
io_processor_plugin=self.io_processor_plugin,
renderer_num_workers=self.renderer_num_workers,