渲染API

本章将介绍渲染相关的 API。

landmark.init_inference

landmark.init_inference(based_model_cls: Model, infer_model_cls: InferenceModule, model_config: str | Path | Config | Dict, inference_config: str | Path | Config | Dict)

Initialize the inference engine.

参数:

• based_model_cls (Model) – a model which is composed of the nerf components.

• infer_model_cls (InferenceModule) – an inference module, which is composed of a model, defines the preprocess, forward and postprocess functions.

• model_config (Union[str, Path, Config, Dict]) – used to define the model parameters.

• inference_config (Union[str, Path, Config, Dict]) – used to define the inference engine.

返回:

InferenceEngine

landmark.InferenceModule

class landmark.InferenceModule(model: Module | None = None)

Abstract inference module for inference engine adaptation

参数:

model (Optional[nn.Module]) – internal model which is composed of the nerf components

channel_last()

Set the model in channel last mode.

forward(*args, **kwargs)

Define the engine’s forward logic. By default, it uses the forward logic of internal model.

参数:

• args (list) – a list containing the input parameters of forward.

• kwargs (dict) – a dict containing input parameters of forward.

返回:

containing the output of the model forward.

返回类型:

Tuple[list, dict]

get_state_dict_from_ckpt(file_path: str | PathLike, map_location: str | Dict[str, str]) → Mapping[str, Any]

Get state_dict from the given file path.

参数:

• file_path (Union[str, os.PathLike]) – a string or os.PathLike object containing a file name.

• map_location (Union[str, Dict[str, str]]) – a string or a dict specifying how to remap storage locations.

返回:

the model’s state dict

返回类型:

Mapping[str, Any]

property inter_model

Returns the internal model.

load_from_state_dict(state_dict: Mapping[str, Any], strict: bool = False)

Copies parameters and buffers from state_dict into this module and its descendants. If strict is True, then the keys of state_dict must exactly match the keys returned by this module’s state_dict() function.

参数:

• state_dict (dict) – a dict containing parameters and persistent buffers.

• strict (bool, optional) – whether to strictly enforce that the keys in state_dict match the keys returned by this module’s key

property merge_config

Returns the offload merge config.

postprocess(*args, **kwargs)

Define the engine’s postprocess logic. By default, it returns the inputs.

参数:

• args (list) – a list containing the input parameters of postprocess.

• kwargs (dict) – a dict containing input parameters of postprocess.

返回:

containing the output of the model postprocess.

返回类型:

Tuple[list, dict]

preprocess(*args, **kwargs)

Define the engine’s preprocessing logic. By default, it returns the inputs.

参数:

• args (list) – a list containing the input parameters of preprocess.

• kwargs (dict) – a dict containing input parameters of preprocess.

返回:

containing the output of the model preprocess.

返回类型:

Tuple[list, dict]

property scene_manager

Returns the scene manager.

landmark.nerf_components.configs.config_parser.BaseConfig

class landmark.nerf_components.configs.config_parser.BaseConfig(config: dict | None = None)

This is a wrapper class for dict objects so that values of which can be accessed as attributes.

参数:

config (dict) – The dict object to be wrapped.

static from_file(filename: str, console_args=None, merge_configs=True)

Reads a python file and constructs a corresponding Config object.

参数:

filename (str) – Name of the file to construct the return object.

返回:

A Config object constructed with information in the file.

返回类型:

Config

抛出:

AssertionError – Raises an AssertionError if the file does not exist, or the file is not .py file

示例:

import os
from landmark.nerf_components.configs.config_parser import BaseConfig

octree_gs_render_config_path = os.path.join(
    os.path.dirname(os.path.abspath(__file__)),
    "benchmarks/nerf/octree_gs/confs/matrixcity_2block_render_ddp.py",
)
model_config = BaseConfig.from_file(octree_gs_render_config_path)
inference_config_dict = dict(
    parallel_config=dict(
        tp_size=1,
    ),
    runtime="Kernel",
    kernel_fusion=True,
    offload_config=dict(
        plane_split=[2, 1],
        local_plane_split=[1, 1],
        use_nccl=False,
    ),
)
inference_config = BaseConfig(inference_config_dict)