tao-train-pointpillars

Compare original and translation side by side

🇺🇸

Original

English

🇨🇳

Translation

Chinese

PointPillars

PointPillars for 3D object detection from LiDAR point clouds. Encodes point clouds into a pseudo-image via pillar-based representation, then applies 2D detection. Used in autonomous driving / robotics.

Typically trained from scratch. Provide train.resume_training_checkpoint_path to resume.

For TAO Deploy TensorRT actions (

gen_trt_engine

, TensorRT

evaluate

, and TensorRT

inference

), read

references/tao-deploy-pointpillars.md

first. Deploy spec templates live in this skill's

references/

folder with the

spec_template_deploy_*.yaml

prefix.

PointPillars用于基于LiDAR点云的3D目标检测。它通过基于柱体的表示方式将点云编码为伪图像，随后应用2D检测技术，适用于自动驾驶/机器人领域。

通常从头开始训练。提供train.resume_training_checkpoint_path参数可恢复训练。

对于TAO Deploy TensorRT操作（

gen_trt_engine

、TensorRT

evaluate

和TensorRT

inference

），请先阅读

references/tao-deploy-pointpillars.md

。部署规格模板位于本技能的

references/

文件夹中，前缀为

spec_template_deploy_*.yaml

。

Dataclass Schemas

数据类模式

Generated TAO Core schemas are packaged in

schemas/<action>.schema.json

, with

schemas/manifest.json

listing available actions. Each generated schema also emits

references/spec_template_<action>.yaml

from the schema top-level

default

field. AutoML enablement is declared at the model layer in

references/skill_info.yaml

via

automl_enabled

. Runnable AutoML still requires

schemas/train.schema.json

and

references/spec_template_train.yaml

to exist and parse. Use the packaged train schema for

automl_default_parameters

automl_disabled_parameters

, defaults, min/max bounds, enums, option weights, math conditions, dependencies, and popular parameters. Do not expect

~/tao-core

at runtime; maintainers regenerate schemas/templates before packaging the skill bank.

生成的TAO Core模式打包在

schemas/<action>.schema.json

中，

schemas/manifest.json

列出了可用操作。每个生成的模式还会从模式顶层的

default

字段生成

references/spec_template_<action>.yaml

。AutoML支持在

references/skill_info.yaml

中的模型层通过

automl_enabled

声明。可运行的AutoML仍要求

schemas/train.schema.json

和

references/spec_template_train.yaml

存在且可解析。使用打包的训练模式来配置

automl_default_parameters

、

automl_disabled_parameters

、默认值、最小/最大边界、枚举、选项权重、数学条件、依赖关系和常用参数。运行时不要依赖

~/tao-core

；维护人员在打包技能库前会重新生成模式/模板。

Train Action Policy

训练操作策略

This model is AutoML-enabled at the model layer. Before handling any train-stage request, read

references/skill_info.yaml

and resolve the run override from either an explicit

automl_policy

value or the user's workflow request. Treat phrases like "turn off AutoML", "disable AutoML", "no HPO", or "plain training" as

automl_policy: off

for this run only; otherwise default to

auto

. When

automl_policy: auto

automl_enabled: true

, and both

schemas/train.schema.json

and

references/spec_template_train.yaml

are packaged, route the train action through

tao-skill-bank:tao-run-automl

by default with this model's

skill_dir

. Preserve workflow/application overrides for datasets, specs, output directories, GPU/platform settings, parent checkpoints, and

automl_policy

. Use direct model training only when

automl_policy: off

or the packaged train schema/template is missing; in the missing-schema case, report that AutoML is enabled but not runnable for this model until schemas are generated.

Non-train actions such as

evaluate

inference

export

, and deploy flows stay in this model skill. The per-run

automl_policy

override does not change model metadata.

该模型在模型层支持AutoML。处理任何训练阶段请求前，请阅读

references/skill_info.yaml

，并通过显式的

automl_policy

值或用户的工作流请求解析运行覆盖配置。将“turn off AutoML”、“disable AutoML”、“no HPO”或“plain training”这类短语视为本次运行的

automl_policy: off

；否则默认设为

auto

。当

automl_policy: auto

、

automl_enabled: true

且

schemas/train.schema.json

和

references/spec_template_train.yaml

均已打包时，默认将训练操作通过

tao-skill-bank:tao-run-automl

路由，并传入该模型的

skill_dir

。保留数据集、规格、输出目录、GPU/平台设置、父检查点和

automl_policy

的工作流/应用覆盖配置。仅当

automl_policy: off

或打包的训练模式/模板缺失时，才使用直接模型训练；若模式缺失，需报告该模型已启用AutoML但无法运行，直到生成模式为止。

非训练操作（如

evaluate

、

inference

、

export

和部署流程）仍在本模型技能中执行。每次运行的

automl_policy

覆盖配置不会更改模型元数据。

Training Requirements

训练要求

Dataset type: pointpillars
Formats: default
Monitoring metric: loss

数据集类型: pointpillars
格式: default
监控指标: loss

Per-Action Dataset Requirements

各操作的数据集要求

Action	Spec Key	Source	Files	List?
dataset_convert	dataset.data_path	id		No
evaluate	dataset.data_path	train_datasets		No
evaluate	dataset.data_info_path	train_datasets	/results/{dataset_convert_job_id}/data_info/	No
export	dataset.data_path	train_datasets		No
export	dataset.data_info_path	train_datasets	/results/{dataset_convert_job_id}/data_info/	No
inference	dataset.data_path	train_datasets		No
inference	dataset.data_info_path	train_datasets	/results/{dataset_convert_job_id}/data_info/	No
prune	dataset.data_path	train_datasets		No
prune	dataset.data_info_path	train_datasets	/results/{dataset_convert_job_id}/data_info/	No
retrain	dataset.data_path	train_datasets		No
retrain	dataset.data_info_path	train_datasets	/results/{dataset_convert_job_id}/data_info/	No
train	dataset.data_path	train_datasets		No
train	dataset.data_info_path	train_datasets	/results/{dataset_convert_job_id}/data_info/	No

操作	规格键	来源	文件	是否为列表？
dataset_convert	dataset.data_path	id		否
evaluate	dataset.data_path	train_datasets		否
evaluate	dataset.data_info_path	train_datasets	/results/{dataset_convert_job_id}/data_info/	否
export	dataset.data_path	train_datasets		否
export	dataset.data_info_path	train_datasets	/results/{dataset_convert_job_id}/data_info/	否
inference	dataset.data_path	train_datasets		否
inference	dataset.data_info_path	train_datasets	/results/{dataset_convert_job_id}/data_info/	否
prune	dataset.data_path	train_datasets		否
prune	dataset.data_info_path	train_datasets	/results/{dataset_convert_job_id}/data_info/	否
retrain	dataset.data_path	train_datasets		否
retrain	dataset.data_info_path	train_datasets	/results/{dataset_convert_job_id}/data_info/	否
train	dataset.data_path	train_datasets		否
train	dataset.data_info_path	train_datasets	/results/{dataset_convert_job_id}/data_info/	否

Typical Spec Overrides

典型规格覆盖配置

Data source overrides are mandatory for every action — the agent MUST construct data source paths from the Per-Action Dataset Requirements table above and include them in

spec_overrides

python

S3_TRAIN = "s3://bucket/data/train"

train (mandatory data sources):

python

{
    "train.num_epochs": 30,
    "train.checkpoint_interval": 10,
    "train.validation_interval": 10,
    "train.num_gpus": 1,
    "dataset.data_path": f"{S3_TRAIN}",
    "dataset.data_info_path": f"{S3_TRAIN}//results/{dataset_convert_job_id}/data_info/",
}

evaluate (mandatory data sources):

python

{
    "dataset.data_path": f"{S3_TRAIN}",
    "dataset.data_info_path": f"{S3_TRAIN}//results/{dataset_convert_job_id}/data_info/",
}

export (mandatory data sources):

python

{
    "dataset.data_path": f"{S3_TRAIN}",
    "dataset.data_info_path": f"{S3_TRAIN}//results/{dataset_convert_job_id}/data_info/",
}

inference (mandatory data sources):

python

{
    "dataset.data_path": f"{S3_TRAIN}",
    "dataset.data_info_path": f"{S3_TRAIN}//results/{dataset_convert_job_id}/data_info/",
}

prune (mandatory data sources):

python

{
    "dataset.data_path": f"{S3_TRAIN}",
    "dataset.data_info_path": f"{S3_TRAIN}//results/{dataset_convert_job_id}/data_info/",
}

retrain (mandatory data sources):

python

{
    "dataset.data_path": f"{S3_TRAIN}",
    "dataset.data_info_path": f"{S3_TRAIN}//results/{dataset_convert_job_id}/data_info/",
}

数据源覆盖配置对每个操作都是必填项——代理必须根据上表的各操作数据集要求构建数据源路径，并将其包含在

spec_overrides

中。

python

S3_TRAIN = "s3://bucket/data/train"

train（必填数据源）:

python

{
    "train.num_epochs": 30,
    "train.checkpoint_interval": 10,
    "train.validation_interval": 10,
    "train.num_gpus": 1,
    "dataset.data_path": f"{S3_TRAIN}",
    "dataset.data_info_path": f"{S3_TRAIN}//results/{dataset_convert_job_id}/data_info/",
}

evaluate（必填数据源）:

python

{
    "dataset.data_path": f"{S3_TRAIN}",
    "dataset.data_info_path": f"{S3_TRAIN}//results/{dataset_convert_job_id}/data_info/",
}

export（必填数据源）:

python

{
    "dataset.data_path": f"{S3_TRAIN}",
    "dataset.data_info_path": f"{S3_TRAIN}//results/{dataset_convert_job_id}/data_info/",
}

inference（必填数据源）:

python

{
    "dataset.data_path": f"{S3_TRAIN}",
    "dataset.data_info_path": f"{S3_TRAIN}//results/{dataset_convert_job_id}/data_info/",
}

prune（必填数据源）:

python

{
    "dataset.data_path": f"{S3_TRAIN}",
    "dataset.data_info_path": f"{S3_TRAIN}//results/{dataset_convert_job_id}/data_info/",
}

retrain（必填数据源）:

python

{
    "dataset.data_path": f"{S3_TRAIN}",
    "dataset.data_info_path": f"{S3_TRAIN}//results/{dataset_convert_job_id}/data_info/",
}

Eval Dataset

评估数据集

Optional. Validation data (val.tar.gz) is separate from training. Used for mAP evaluation.

可选。验证数据（val.tar.gz）与训练数据分离，用于mAP评估。

Important Parameters

重要参数

train.num_epochs: Default 80 (much higher than other TAO models). PointPillars needs more epochs for convergence on 3D detection.
train.lr: Learning rate. Default 0.003 (adam_onecycle scheduler).
dataset.class_names: List of 3D object classes. Default 7 classes (KITTI-style). Modify to match your dataset.
dataset.data_path: Path to point cloud data directory.
dataset.data_info_path: Path to data info files from dataset_convert step.
dataset.point_cloud_range: Spatial extent of the point cloud to consider. Must match your sensor configuration.
model.dense_head.anchor_generator_config: Anchor configurations per class. Must be tuned for your object sizes and the point cloud range.

train.num_epochs: 默认值80（远高于其他TAO模型）。PointPillars在3D检测任务中需要更多轮次才能收敛。
train.lr: 学习率。默认值0.003（使用adam_onecycle调度器）。
dataset.class_names: 3D目标类别列表。默认7类（KITTI风格）。需根据数据集修改。
dataset.data_path: 点云数据目录路径。
dataset.data_info_path: 来自dataset_convert步骤的数据信息文件路径。
dataset.point_cloud_range: 要考虑的点云空间范围，必须与传感器配置匹配。
model.dense_head.anchor_generator_config: 每个类别的锚框配置，必须针对目标尺寸和点云范围进行调优。

Multi-GPU / Multi-Node

多GPU/多节点

Launch method:

torchrun

(LIGHTNING_EXCLUDED_NETWORK). Uses PyTorch native

DistributedDataParallel

(NOT Lightning Trainer).

Spec Key	Description	Default
`train.num_gpus`	Number of GPUs per node	1
`train.gpu_ids`	GPU device indices	[0]
`train.num_nodes`	Number of nodes	1

```
CUDA_VISIBLE_DEVICES
```
is explicitly set from
```
TAO_VISIBLE_DEVICES
```
Uses
```
nn.parallel.DistributedDataParallel
```
directly (not Lightning strategy)
```
NODE_RANK
```
is copied to
```
RANK
```
if
```
RANK
```
is unset

Multi-node env vars (set by orchestrator):

Variable	Purpose
`WORLD_SIZE`	Number of nodes
`NODE_RANK`	This node's rank
`MASTER_ADDR`	Rank-0 node IP
`MASTER_PORT`	Rank-0 port (default 29500)
`NUM_GPU_PER_NODE`	GPUs per node

启动方式:

torchrun

（LIGHTNING_EXCLUDED_NETWORK）。使用PyTorch原生的

DistributedDataParallel

（而非Lightning Trainer）。

规格键	描述	默认值
`train.num_gpus`	每个节点的GPU数量	1
`train.gpu_ids`	GPU设备索引	[0]
`train.num_nodes`	节点数量	1

```
CUDA_VISIBLE_DEVICES
```
由
```
TAO_VISIBLE_DEVICES
```
显式设置
直接使用
```
nn.parallel.DistributedDataParallel
```
（而非Lightning策略）
如果
```
RANK
```
未设置，将
```
NODE_RANK
```
复制到
```
RANK
```

多节点环境变量（由编排器设置）:

变量	用途
`WORLD_SIZE`	节点总数
`NODE_RANK`	当前节点的排名
`MASTER_ADDR`	排名为0的节点IP
`MASTER_PORT`	排名为0的节点端口（默认29500）
`NUM_GPU_PER_NODE`	每个节点的GPU数量

Hardware

硬件要求

Minimum 1 GPU(s), recommended 4 GPU(s). 16GB+ (V100 or A100) VRAM per GPU. PointPillars is relatively efficient for 3D detection. The main bottleneck is data I/O for large point cloud datasets.

最低1块GPU，推荐4块GPU。每块GPU需16GB及以上显存（V100或A100）。PointPillars在3D检测任务中效率相对较高，主要瓶颈是大型点云数据集的I/O。

Error Patterns

常见错误模式

dataset_convert required: Training will fail if data_info_path is not populated from a prior dataset_convert job. Always run convert first.

Point cloud range mismatch: If point_cloud_range does not match the actual sensor data extent, detections will be poor or empty.

Epoch numbering: PointPillars checkpoint epoch numbers may be offset by 1 from status.json reported epochs.

需先执行dataset_convert: 如果data_info_path未从之前的dataset_convert任务中获取，训练将失败。请始终先执行转换操作。

点云范围不匹配: 如果point_cloud_range与实际传感器数据范围不匹配，检测结果会很差或为空。

轮次编号: PointPillars的检查点轮次编号可能与status.json中报告的轮次编号相差1。

Spec Param / Parent Model Inference

规格参数/父模型推理

Model-specific inference mappings belong in this MD file, not in

config.json

. Generated runners should read this section and apply the mappings with SDK helpers before

create_job()

. This mirrors the old microservices

infer_params.py

flow.

Inference mappings from TAO Core

pointpillars.config.json

Action	Spec Field	Inference Function	Meaning
dataset_convert	`results_dir`	`output_dir`	current job results directory
evaluate	`evaluate.checkpoint`	`parent_model`	model file inferred from the parent job results folder
evaluate	`key`	`key`	encryption key
evaluate	`results_dir`	`output_dir`	current job results directory
export	`export.checkpoint`	`parent_model`	model file inferred from the parent job results folder
export	`export.onnx_file`	`create_onnx_file`	output ONNX path
export	`export.save_engine`	`create_engine_file`	output TensorRT engine path
export	`key`	`key`	encryption key
export	`results_dir`	`output_dir`	current job results directory
gen_trt_engine	`gen_trt_engine.onnx_file`	`parent_model`	model file inferred from the parent job results folder
gen_trt_engine	`gen_trt_engine.save_engine`	`create_engine_file`	output TensorRT engine path
gen_trt_engine	`key`	`key`	encryption key
gen_trt_engine	`results_dir`	`output_dir`	current job results directory
inference	`inference.checkpoint`	`parent_model`	model file inferred from the parent job results folder
inference	`inference.trt_engine`	`parent_model`	model file inferred from the parent job results folder
inference	`key`	`key`	encryption key
inference	`results_dir`	`output_dir`	current job results directory
prune	`key`	`key`	encryption key
prune	`prune.model`	`parent_model`	model file inferred from the parent job results folder
prune	`results_dir`	`output_dir`	current job results directory
retrain	`key`	`key`	encryption key
retrain	`results_dir`	`output_dir`	current job results directory
retrain	`train.pruned_model_path`	`parent_model`	model file inferred from the parent job results folder
train	`key`	`key`	encryption key
train	`model.pretrained_model_path`	`ptm_if_no_resume_model`	PTM when no resume checkpoint exists
train	`results_dir`	`output_dir`	current job results directory
train	`train.resume_training_checkpoint_path`	`resume_model`	model file inferred from the current job results folder

For

parent_model

parent_model_folder

, pass the upstream train/export/AutoML child job id as

parent_job_id

. The SDK lists the parent result folder, filters checkpoint artifacts, and returns the selected model file or folder. Do not add these mappings back to

config.json

and do not patch generated runner scripts to guess checkpoint paths.

模型特定的推理映射应放在此MD文件中，而非

config.json

。生成的运行器应读取本节内容，并在

create_job()

前使用SDK助手应用映射。这与旧微服务的

infer_params.py

流程一致。

来自TAO Core

pointpillars.config.json

的推理映射:

操作	规格字段	推理函数	含义
dataset_convert	`results_dir`	`output_dir`	当前任务结果目录
evaluate	`evaluate.checkpoint`	`parent_model`	从父任务结果文件夹推断出的模型文件
evaluate	`key`	`key`	加密密钥
evaluate	`results_dir`	`output_dir`	当前任务结果目录
export	`export.checkpoint`	`parent_model`	从父任务结果文件夹推断出的模型文件
export	`export.onnx_file`	`create_onnx_file`	输出ONNX路径
export	`export.save_engine`	`create_engine_file`	输出TensorRT引擎路径
export	`key`	`key`	加密密钥
export	`results_dir`	`output_dir`	当前任务结果目录
gen_trt_engine	`gen_trt_engine.onnx_file`	`parent_model`	从父任务结果文件夹推断出的模型文件
gen_trt_engine	`gen_trt_engine.save_engine`	`create_engine_file`	输出TensorRT引擎路径
gen_trt_engine	`key`	`key`	加密密钥
gen_trt_engine	`results_dir`	`output_dir`	当前任务结果目录
inference	`inference.checkpoint`	`parent_model`	从父任务结果文件夹推断出的模型文件
inference	`inference.trt_engine`	`parent_model`	从父任务结果文件夹推断出的模型文件
inference	`key`	`key`	加密密钥
inference	`results_dir`	`output_dir`	当前任务结果目录
prune	`key`	`key`	加密密钥
prune	`prune.model`	`parent_model`	从父任务结果文件夹推断出的模型文件
prune	`results_dir`	`output_dir`	当前任务结果目录
retrain	`key`	`key`	加密密钥
retrain	`results_dir`	`output_dir`	当前任务结果目录
retrain	`train.pruned_model_path`	`parent_model`	从父任务结果文件夹推断出的模型文件
train	`key`	`key`	加密密钥
train	`model.pretrained_model_path`	`ptm_if_no_resume_model`	无恢复检查点时使用预训练模型
train	`results_dir`	`output_dir`	当前任务结果目录
train	`train.resume_training_checkpoint_path`	`resume_model`	从当前任务结果文件夹推断出的模型文件

对于

parent_model

或

parent_model_folder

，将上游训练/导出/AutoML子任务ID作为

parent_job_id

传入。SDK会列出父结果文件夹，过滤检查点工件，并返回选定的模型文件或文件夹。请勿将这些映射添加回

config.json

，也不要修改生成的运行器脚本以猜测检查点路径。

Deployment

部署

tao-deploy-pointpillars — PointPillars deploy workflow for TensorRT engine generation, TensorRT evaluation, and TensorRT inference using TAO Deploy.

tao-deploy-pointpillars — 使用TAO Deploy进行PointPillars部署的工作流，包括TensorRT引擎生成、TensorRT评估和TensorRT推理。