Skip to content

vllm_omni.platforms.npu

Modules:

Name Description
models
omni_connectors
platform
profiler
quant
worker

NPUOmniPlatform

Bases: OmniPlatform, NPUPlatform

NPU/Ascend implementation of OmniPlatform.

Inherits all NPU-specific implementations from vllm-ascend's NPUPlatform, and adds Omni-specific interfaces from OmniPlatform.

dist_backend class-attribute instance-attribute 

dist_backend: str = 'hccl'

create_autocast_context classmethod 

create_autocast_context(
    *, device_type, dtype, enabled=True
)

get_default_stage_config_path classmethod 

get_default_stage_config_path() -> str

get_device_count classmethod 

get_device_count() -> int

get_device_total_memory classmethod 

get_device_total_memory(device_id: int = 0) -> int

get_device_version classmethod 

get_device_version() -> str | None

get_diffusion_attn_backend_cls classmethod 

get_diffusion_attn_backend_cls(
    selected_backend: str | None, head_size: int
) -> str

get_diffusion_model_impl_qualname classmethod 

get_diffusion_model_impl_qualname(op_name: str) -> str

get_diffusion_packed_modules_mapping classmethod 

get_diffusion_packed_modules_mapping(
    model_class: type[Module],
) -> dict[str, list[str]] | None

get_free_memory classmethod 

get_free_memory(device: device | None = None) -> int

get_graph_wrapper_cls classmethod 

get_graph_wrapper_cls() -> type

get_omni_ar_worker_cls classmethod 

get_omni_ar_worker_cls() -> str

get_omni_generation_worker_cls classmethod 

get_omni_generation_worker_cls() -> str

get_profiler_cls classmethod 

get_profiler_cls() -> str

get_torch_device classmethod 

get_torch_device(local_rank: int | None = None) -> device

prepare_diffusion_op_runtime classmethod 

prepare_diffusion_op_runtime(
    op_name: str, **kwargs: Any
) -> None

set_device classmethod 

set_device(device: device) -> None

set_forward_context classmethod 

set_forward_context(
    attn_metadata,
    vllm_config,
    *,
    cudagraph_runtime_mode,
    batch_descriptor,
)

supports_torch_inductor classmethod 

supports_torch_inductor() -> bool

synchronize classmethod 

synchronize() -> None