tao-train-segformer

Compare original and translation side by side

🇺🇸

Original

English

🇨🇳

Translation

Chinese

SegFormer

SegFormer for semantic segmentation. Lightweight transformer-based architecture with hierarchical feature extraction. Efficient for real-time segmentation tasks.

Set model.backbone.pretrained_backbone_path for backbone weights.

For TAO Deploy TensorRT actions (

gen_trt_engine

, TensorRT

evaluate

, and TensorRT

inference

), read

references/tao-deploy-segformer.md

first. Deploy spec templates live in this skill's

references/

folder with the

spec_template_deploy_*.yaml

prefix.

用于语义分割的SegFormer。基于Transformer的轻量级架构，具备分层特征提取能力，适用于实时分割任务。

设置model.backbone.pretrained_backbone_path以指定骨干网络权重。

对于TAO Deploy TensorRT操作（

gen_trt_engine

、TensorRT

evaluate

和TensorRT

inference

），请先阅读

references/tao-deploy-segformer.md

。部署配置模板存放在本skill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启用状态在

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/最大范围、枚举值、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skill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

；否则默认使用

auto

。当

automl_policy: auto

、

automl_enabled: true

且

schemas/train.schema.json

和

references/spec_template_train.yaml

已打包时，默认将训练操作通过

tao-skill-bank:tao-run-automl

路由，并传入该模型的

skill_dir

。保留数据集、配置、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skill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: segmentation
Formats: unet
Monitoring metric: val_miou

数据集类型： segmentation
格式： unet
监控指标： val_miou

Per-Action Dataset Requirements

各操作的数据集要求

Action	Spec Key	Source	List?
evaluate	dataset.segment.root_dir	eval_dataset	No
export	dataset.segment.root_dir	train_datasets	No
inference	dataset.segment.root_dir	eval_dataset	No
quantize	dataset.segment.root_dir	train_datasets	No
quantize	dataset.segment.quant_calibration_dataset.images_dir	train_datasets	No
train	dataset.segment.root_dir	train_datasets	No

操作	配置键	数据源	是否为列表？
evaluate	dataset.segment.root_dir	eval_dataset	否
export	dataset.segment.root_dir	train_datasets	否
inference	dataset.segment.root_dir	eval_dataset	否
quantize	dataset.segment.root_dir	train_datasets	否
quantize	dataset.segment.quant_calibration_dataset.images_dir	train_datasets	否
train	dataset.segment.root_dir	train_datasets	否

Typical Spec Overrides

典型配置覆盖

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

{
    "train.num_gpus": 1,
    "train.num_epochs": 10,
    "train.checkpoint_interval": 10,
    "train.validation_interval": 10,
    "dataset.segment.batch_size": 4,
    "dataset.segment.root_dir": f"{S3_TRAIN}",
}

evaluate (mandatory data sources):

python

{
    "evaluate.batch_size": 4,
    "dataset.segment.root_dir": f"{S3_EVAL}",
}

gen_trt_engine:

python

{
    "gen_trt_engine.tensorrt.data_type": "fp16",
}

inference (mandatory data sources):

python

{
    "dataset.segment.batch_size": 1,
    "dataset.segment.root_dir": f"{S3_EVAL}",
}

export (mandatory data sources):

python

{
    "dataset.segment.root_dir": f"{S3_TRAIN}",
}

quantize (mandatory data sources):

python

{
    "dataset.segment.root_dir": f"{S3_TRAIN}",
    "dataset.segment.quant_calibration_dataset.images_dir": f"{S3_TRAIN}",
}

数据源覆盖对每个操作都是必填项 —— 代理必须根据上述各操作数据集要求表构建数据源路径，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

{
    "train.num_gpus": 1,
    "train.num_epochs": 10,
    "train.checkpoint_interval": 10,
    "train.validation_interval": 10,
    "dataset.segment.batch_size": 4,
    "dataset.segment.root_dir": f"{S3_TRAIN}",
}

evaluate（必填数据源）：

python

{
    "evaluate.batch_size": 4,
    "dataset.segment.root_dir": f"{S3_EVAL}",
}

gen_trt_engine：

python

{
    "gen_trt_engine.tensorrt.data_type": "fp16",
}

inference（必填数据源）：

python

{
    "dataset.segment.batch_size": 1,
    "dataset.segment.root_dir": f"{S3_EVAL}",
}

export（必填数据源）：

python

{
    "dataset.segment.root_dir": f"{S3_TRAIN}",
}

quantize（必填数据源）：

python

{
    "dataset.segment.root_dir": f"{S3_TRAIN}",
    "dataset.segment.quant_calibration_dataset.images_dir": f"{S3_TRAIN}",
}

Eval Dataset

评估数据集

Optional. Validation data is typically part of the root_dir structure.

可选。验证数据通常包含在root_dir结构中。

Important Parameters

重要参数

dataset.segment.num_classes: Number of segmentation classes. Default 2 (binary). Must match the number of classes in your mask annotations.
model.backbone.type: Default fan_small_12_p4_hybrid. Supported includes FAN variants, SegFormer MIT variants, and others.
dataset.segment.root_dir: Root directory of the segmentation dataset.
dataset.segment.img_size: Input image size. Default 256. Increase for finer segmentation at the cost of memory.
train.optim.lr: Learning rate. Default 6e-5.
model.freeze_backbone: Whether to freeze the backbone during training. Useful for fine-tuning with limited data.
dataset.segment.batch_size: Per-GPU batch size. Default 8.

dataset.segment.num_classes：分割类别数量。默认值为2（二分类）。必须与掩码标注中的类别数量匹配。
model.backbone.type：默认值为fan_small_12_p4_hybrid。支持的类型包括FAN变体、SegFormer MIT变体等。
dataset.segment.root_dir：分割数据集的根目录。
dataset.segment.img_size：输入图像尺寸。默认值为256。增大该值可获得更精细的分割效果，但会占用更多内存。
train.optim.lr：学习率。默认值为6e-5。
model.freeze_backbone：训练期间是否冻结骨干网络。在数据有限的情况下进行微调时非常有用。
dataset.segment.batch_size：单GPU批次大小。默认值为8。

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.sync_batchnorm`	Sync BN across GPUs	configurable
`train.use_distributed_sampler`	Use distributed sampler	configurable

Multi-GPU strategy:
```
ddp_find_unused_parameters_true
```
No fsdp support

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生成工作进程）。

配置键	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.sync_batchnorm`	在GPU间同步BN层	可配置
`train.use_distributed_sampler`	使用分布式采样器	可配置

多GPU策略：
```
ddp_find_unused_parameters_true
```
不支持fsdp

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 1 GPU(s), recommended 2 GPU(s). 16GB+ (V100 or A100) VRAM per GPU. SegFormer is relatively lightweight. Default img_size=256 is memory-friendly. Increase img_size for higher resolution at the cost of memory and speed.

最少1块GPU，推荐2块GPU。每块GPU需16GB以上显存（V100或A100）。SegFormer相对轻量。默认img_size=256对内存友好。增大img_size可获得更高分辨率，但会消耗更多内存并降低速度。

Error Patterns

错误类型

CUDA out of memory: Reduce batch_size or img_size. SegFormer memory scales quadratically with image size.

num_classes mismatch: Ensure dataset.segment.num_classes matches the actual number of classes in your mask annotations.

CUDA内存不足：减小batch_size或img_size。SegFormer的内存占用与图像尺寸呈二次方关系。

num_classes不匹配：确保dataset.segment.num_classes与掩码标注中的实际类别数量一致。

Spec Param / Parent Model Inference

配置参数 / 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

segformer.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
quantize	`encryption_key`	`key`	encryption key
quantize	`quantize.model_path`	`parent_model`	model file inferred from the parent job results folder
quantize	`results_dir`	`output_dir`	current job results directory
train	`encryption_key`	`key`	encryption key
train	`model.backbone.pretrained_backbone_path`	`ptm_if_no_resume_model`	PTM when no resume checkpoint exists
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

segformer.config.json

的推理映射：

操作	配置字段	推理函数	含义
evaluate	`encryption_key`	`key`	加密密钥
evaluate	`evaluate.checkpoint`	`parent_model`	从父任务结果文件夹中推断出的模型文件
evaluate	`evaluate.trt_engine`	`parent_model`	从父任务结果文件夹中推断出的模型文件
evaluate	`results_dir`	`output_dir`	当前任务结果目录
export	`encryption_key`	`key`	加密密钥
export	`export.checkpoint`	`parent_model`	从父任务结果文件夹中推断出的模型文件
export	`export.onnx_file`	`create_onnx_file`	输出ONNX路径
export	`results_dir`	`output_dir`	当前任务结果目录
gen_trt_engine	`encryption_key`	`key`	加密密钥
gen_trt_engine	`gen_trt_engine.onnx_file`	`parent_model`	从父任务结果文件夹中推断出的模型文件
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`	当前任务结果目录
inference	`encryption_key`	`key`	加密密钥
inference	`inference.checkpoint`	`parent_model`	从父任务结果文件夹中推断出的模型文件
inference	`inference.trt_engine`	`parent_model`	从父任务结果文件夹中推断出的模型文件
inference	`results_dir`	`output_dir`	当前任务结果目录
quantize	`encryption_key`	`key`	加密密钥
quantize	`quantize.model_path`	`parent_model`	从父任务结果文件夹中推断出的模型文件
quantize	`results_dir`	`output_dir`	当前任务结果目录
train	`encryption_key`	`key`	加密密钥
train	`model.backbone.pretrained_backbone_path`	`ptm_if_no_resume_model`	无恢复检查点时使用的预训练模型
train	`results_dir`	`output_dir`	当前任务结果目录
train	`train.pretrained_model_path`	`ptm_if_no_resume_model`	无恢复检查点时使用的预训练模型
train	`train.resume_training_checkpoint_path`	`resume_model`	从当前任务结果文件夹中推断出的模型文件

对于

parent_model

或

parent_model_folder

，传入上游训练/导出/AutoML子任务ID作为

parent_job_id

。SDK会列出父任务结果文件夹，筛选检查点工件，并返回选中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-segformer — SegFormer deploy workflow for TensorRT engine generation, TensorRT evaluation, and TensorRT inference using TAO Deploy.

tao-deploy-segformer —— 用于SegFormer的部署工作流，支持使用TAO Deploy生成TensorRT引擎、进行TensorRT评估和TensorRT推理。