tao-train-mask-auto-encoder

Compare original and translation side by side

🇺🇸

Original

English

🇨🇳

Translation

Chinese

MAE

MAE (Masked Autoencoder) for self-supervised pretraining and fine-tuning. Masks random patches and reconstructs them to learn visual representations. Supports pretrain and finetune stages.

Set train.pretrained_model_path for pretrained MAE weights when fine-tuning.

For TAO Deploy TensorRT actions (

gen_trt_engine

), read

references/tao-deploy-mask-auto-encoder.md

first. Deploy spec templates live in this skill's

references/

folder with the

spec_template_deploy_*.yaml

prefix.

MAE（Masked Autoencoder）用于自监督预训练与微调。通过掩码随机图像块并重构它们来学习视觉表征，支持预训练和微调阶段。

微调时，请设置train.pretrained_model_path以指定预训练MAE权重。

对于TAO Deploy TensorRT操作（

gen_trt_engine

），请先阅读

references/tao-deploy-mask-auto-encoder.md

。部署规格模板存放在本skill的

references/

文件夹下，前缀为

spec_template_deploy_*.yaml

。

Dataclass Schemas

数据类模式（Dataclass Schemas）

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

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、枚举值、选项权重、数学条件、依赖关系以及常用参数。运行时不要依赖

~/tao-core

；维护人员在打包skill库前会重新生成模式/模板。

Train Action Policy

训练操作策略（Train Action Policy）

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来确定运行覆盖配置。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

已打包时，默认将训练操作路由至

tao-skill-bank:tao-run-automl

，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

以及部署流程）仍在本模型skill中处理。每次运行的

automl_policy

覆盖配置不会修改模型元数据。

Training Requirements

训练要求

Dataset type: image_classification
Formats: ssl
Accepted dataset intents: training, evaluation, testing
Monitoring metric: train_loss

数据集类型： image_classification
格式： ssl
接受的数据集用途： training、evaluation、testing
监控指标： train_loss

Per-Action Dataset Requirements

各操作的数据集要求

Action	Spec Key	Source	Files	List?
train	dataset.train_data_sources	train_datasets	images_train.tar.gz	No
train	dataset.val_data_sources	eval_dataset	images_val.tar.gz	No
evaluate	dataset.val_data_sources	eval_dataset	images_val.tar.gz	No
inference	dataset.test_data_sources	inference_dataset	images_test.tar.gz	No

操作	规格键（Spec Key）	来源	文件	是否为列表？
train	dataset.train_data_sources	train_datasets	images_train.tar.gz	否
train	dataset.val_data_sources	eval_dataset	images_val.tar.gz	否
evaluate	dataset.val_data_sources	eval_dataset	images_val.tar.gz	否
inference	dataset.test_data_sources	inference_dataset	images_test.tar.gz	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"
S3_EVAL = "s3://bucket/data/eval"

train (mandatory data sources):

python

{
    "dataset.train_data_sources": f"{S3_TRAIN}/images_train.tar.gz",
    "dataset.val_data_sources": f"{S3_EVAL}/images_val.tar.gz",
    "train.num_epochs": 10,
    "train.optim.lr": 2e-4,
}

evaluate (mandatory data sources):

python

{
    "dataset.val_data_sources": f"{S3_EVAL}/images_val.tar.gz",
}

inference (mandatory data sources):

python

{
    "dataset.test_data_sources": f"{S3_EVAL}/images_test.tar.gz",
}

数据源覆盖配置对每个操作都是必填项——agent必须根据上述“各操作的数据集要求”表格构建数据源路径，并将其包含在

spec_overrides

中。

python

S3_TRAIN = "s3://bucket/data/train"
S3_EVAL = "s3://bucket/data/eval"

train（必填数据源）：

python

{
    "dataset.train_data_sources": f"{S3_TRAIN}/images_train.tar.gz",
    "dataset.val_data_sources": f"{S3_EVAL}/images_val.tar.gz",
    "train.num_epochs": 10,
    "train.optim.lr": 2e-4,
}

evaluate（必填数据源）：

python

{
    "dataset.val_data_sources": f"{S3_EVAL}/images_val.tar.gz",
}

inference（必填数据源）：

python

{
    "dataset.test_data_sources": f"{S3_EVAL}/images_test.tar.gz",
}

Eval Dataset

评估数据集

Optional. Pretraining does not need eval data. Fine-tuning optionally uses val set.

可选。预训练不需要评估数据，微调可选择性使用验证集。

Important Parameters

重要参数

train.stage: Training stage. Options: pretrain, finetune. Pretrain learns representations via masking. Finetune adds a classification head.
model.arch: Architecture. Default convnextv2_base. Wide range of options including ConvNeXt, Hiera, ViT variants.
model.num_classes: Number of classes for fine-tuning. Default 1000 (ImageNet). Only relevant in finetune stage.
model.mask_ratio: Fraction of patches to mask during pretraining. Typically 0.75.
model.norm_pix_loss: Whether to normalize pixel values in reconstruction loss.
train.optim.lr: Learning rate. Default 2e-4.
dataset.augmentation: Augmentation settings including mixup, cutmix for fine-tuning.

train.stage：训练阶段。选项：pretrain、finetune。预训练通过掩码学习表征，微调则添加分类头。
model.arch：网络架构。默认convnextv2_base。支持多种选项，包括ConvNeXt、Hiera、ViT变体。
model.num_classes：微调时的类别数量。默认1000（ImageNet），仅在微调阶段有效。
model.mask_ratio：预训练时掩码的图像块比例，通常为0.75。
model.norm_pix_loss：是否在重构损失中对像素值进行归一化。
train.optim.lr：学习率，默认2e-4。
dataset.augmentation：数据增强设置，包括微调时的mixup、cutmix。

Multi-GPU / Multi-Node

多GPU/多节点

Launch method: Lightning-managed (single

python

process, Lightning spawns workers).

Spec Key	Description	Default
`train.num_gpus`	Number of GPUs	1
`train.gpu_ids`	GPU device indices	[0]
`train.num_nodes`	Number of nodes	1
`train.distributed_strategy`	`ddp` or `fsdp`	`ddp`

```
ddp
```
uses
```
find_unused_parameters=True
```
```
fsdp
```
forces FP16
Multi-GPU strongly recommended for pretraining (large batch sizes needed)

Multi-node env vars (set by orchestrator):

WORLD_SIZE

NODE_RANK

MASTER_ADDR

MASTER_PORT

NUM_GPU_PER_NODE

启动方式： Lightning管理（单个

python

进程，Lightning生成工作线程）。

规格键（Spec Key）	描述	默认值
`train.num_gpus`	GPU数量	1
`train.gpu_ids`	GPU设备索引	[0]
`train.num_nodes`	节点数量	1
`train.distributed_strategy`	`ddp` 或 `fsdp`	`ddp`

```
ddp
```
使用
```
find_unused_parameters=True
```
```
fsdp
```
强制使用FP16
预训练强烈推荐使用多GPU（需要大批次大小）

多节点环境变量（由编排器设置）：

WORLD_SIZE

、

NODE_RANK

、

MASTER_ADDR

、

MASTER_PORT

、

NUM_GPU_PER_NODE

。

Hardware

硬件要求

Minimum 2 GPU(s), recommended 8 GPU(s). 24GB+ (A100 recommended) VRAM per GPU. MAE pretraining benefits from large batch sizes across many GPUs. Fine-tuning is more modest in resource requirements.

最低2块GPU，推荐8块GPU。每块GPU需24GB及以上显存（推荐A100）。MAE预训练受益于多GPU上的大批次大小，微调对资源的要求相对较低。

Error Patterns

错误模式

Stage mismatch: Ensure train.stage matches your intent (pretrain vs finetune). Fine-tuning without a pretrained_model_path trains from scratch.

num_classes mismatch (finetune only): Ensure model.num_classes matches your dataset class count when fine-tuning.

阶段不匹配：确保train.stage与你的意图一致（预训练vs微调）。若微调时未设置pretrained_model_path，则会从头开始训练。

类别数量不匹配（仅微调）：微调时确保model.num_classes与数据集的类别数量一致。

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

mae.config.json

Action	Spec Field	Inference Function	Meaning
evaluate	`encryption_key`	`key`	encryption key
evaluate	`evaluate.checkpoint`	`parent_model`	model file inferred from the parent job results folder
evaluate	`evaluate.trt_engine`	`parent_model`	model file inferred from the parent job results folder
evaluate	`results_dir`	`output_dir`	current job results directory
export	`encryption_key`	`key`	encryption key
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	`results_dir`	`output_dir`	current job results directory
gen_trt_engine	`encryption_key`	`key`	encryption key
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.trt_engine`	`create_engine_file`	output TensorRT engine path
gen_trt_engine	`results_dir`	`output_dir`	current job results directory
inference	`encryption_key`	`key`	encryption key
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	`results_dir`	`output_dir`	current job results directory
train	`encryption_key`	`key`	encryption key
train	`results_dir`	`output_dir`	current job results directory
train	`train.pretrained_model_path`	`ptm_if_no_resume_model`	PTM when no resume checkpoint exists
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本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

mae.config.json

的推理映射：

操作	规格字段（Spec Field）	推理函数	含义
evaluate	`encryption_key`	`key`	加密密钥
evaluate	`evaluate.checkpoint`	`parent_model`	从父作业结果文件夹推断出的模型文件
evaluate	`evaluate.trt_engine`	`parent_model`	从父作业结果文件夹推断出的模型文件
evaluate	`results_dir`	`output_dir`	当前作业结果目录
export	`encryption_key`	`key`	加密密钥
export	`export.checkpoint`	`parent_model`	从父作业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	`results_dir`	`output_dir`	当前作业结果目录
gen_trt_engine	`encryption_key`	`key`	加密密钥
gen_trt_engine	`gen_trt_engine.onnx_file`	`parent_model`	从父作业结果文件夹推断出的模型文件
gen_trt_engine	`gen_trt_engine.trt_engine`	`create_engine_file`	输出TensorRT引擎路径
gen_trt_engine	`results_dir`	`output_dir`	当前作业结果目录
inference	`encryption_key`	`key`	加密密钥
inference	`inference.checkpoint`	`parent_model`	从父作业结果文件夹推断出的模型文件
inference	`inference.trt_engine`	`parent_model`	从父作业结果文件夹推断出的模型文件
inference	`results_dir`	`output_dir`	当前作业结果目录
train	`encryption_key`	`key`	加密密钥
train	`results_dir`	`output_dir`	当前作业结果目录
train	`train.pretrained_model_path`	`ptm_if_no_resume_model`	无恢复检查点时的预训练模型（PTM）
train	`train.resume_training_checkpoint_path`	`resume_model`	从当前作业结果文件夹推断出的模型文件

对于

parent_model

或

parent_model_folder

，需传入上游训练/导出/AutoML子作业ID作为

parent_job_id

。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来猜测检查点路径。

Deployment

部署

tao-deploy-mask-auto-encoder — MAE deploy workflow for TensorRT engine generation using TAO Deploy.

tao-deploy-mask-auto-encoder —— 用于通过TAO Deploy生成TensorRT引擎的MAE部署工作流。