tao-train-ocrnet

Compare original and translation side by side

🇺🇸

Original

English

🇨🇳

Translation

Chinese

OCRNet

OCRNet for scene text recognition. Recognizes text content from cropped text region images. Supports CTC and attention-based decoders.

Set train.pretrained_model_path for pretrained OCR weights.

For TAO Deploy TensorRT actions (

gen_trt_engine

, TensorRT

evaluate

, and TensorRT

inference

), read

references/tao-deploy-ocrnet.md

first. Deploy spec templates live in this skill's

references/

folder with the

spec_template_deploy_*.yaml

prefix.

用于场景文本识别的OCRNet。可识别裁剪后文本区域图像中的文本内容，支持CTC和基于注意力的解码器。

设置train.pretrained_model_path以指定预训练OCR权重。

对于TAO Deploy TensorRT操作（

gen_trt_engine

、TensorRT

evaluate

和TensorRT

inference

），请先阅读

references/tao-deploy-ocrnet.md

。部署规格模板位于该技能的

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

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、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

，并从显式的

automl_policy

值或用户的工作流请求中解析运行覆盖配置。将“turn off AutoML”、“disable AutoML”、“no HPO”或“plain training”等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，默认通过该模型的

skill_dir

将训练操作路由至

tao-skill-bank:tao-run-automl

。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；在模式缺失的情况下，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）仍在该模型技能中执行。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: ocrnet
Formats: default
Monitoring metric: val_acc

数据集类型： ocrnet
格式： default
监控指标： val_acc

Per-Action Dataset Requirements

各操作的数据集要求

Action	Spec Key	Source	Files	List?
dataset_convert	dataset_convert.input_img_dir	id		No
dataset_convert	dataset_convert.gt_file	id		No
evaluate	dataset.character_list_file	eval_dataset	character_list	No
evaluate	evaluate.test_dataset_dir	eval_dataset	results/{dataset_convert_job_id}/dataset_convert/lmdb	No
export	dataset.character_list_file	eval_dataset	character_list	No
gen_trt_engine	gen_trt_engine.tensorrt.calibration.cal_image_dir	calibration_dataset		Yes
inference	dataset.character_list_file	eval_dataset	character_list	No
inference	inference.inference_dataset_dir	eval_dataset	test.tar.gz	No
prune	dataset.character_list_file	eval_dataset	character_list	No
quantize	dataset.train_dataset_dir	train_datasets	results/{dataset_convert_job_id}/dataset_convert/lmdb	Yes
quantize	dataset.val_dataset_dir	eval_dataset	results/{dataset_convert_job_id}/dataset_convert/lmdb	No
quantize	dataset.character_list_file	eval_dataset	character_list	No
quantize	dataset.quant_calibration_dataset.images_dir	train_datasets	train.tar.gz	No
retrain	dataset.train_dataset_dir	train_datasets	results/{dataset_convert_job_id}/dataset_convert/lmdb	Yes
retrain	dataset.val_dataset_dir	eval_dataset	results/{dataset_convert_job_id}/dataset_convert/lmdb	No
retrain	dataset.character_list_file	eval_dataset	character_list	No
train	dataset.train_dataset_dir	train_datasets	results/{dataset_convert_job_id}/dataset_convert/lmdb	Yes
train	dataset.val_dataset_dir	eval_dataset	results/{dataset_convert_job_id}/dataset_convert/lmdb	No
train	dataset.character_list_file	eval_dataset	character_list	No

操作	规格键	来源	文件	是否为列表？
dataset_convert	dataset_convert.input_img_dir	id		否
dataset_convert	dataset_convert.gt_file	id		否
evaluate	dataset.character_list_file	eval_dataset	character_list	否
evaluate	evaluate.test_dataset_dir	eval_dataset	results/{dataset_convert_job_id}/dataset_convert/lmdb	否
export	dataset.character_list_file	eval_dataset	character_list	否
gen_trt_engine	gen_trt_engine.tensorrt.calibration.cal_image_dir	calibration_dataset		是
inference	dataset.character_list_file	eval_dataset	character_list	否
inference	inference.inference_dataset_dir	eval_dataset	test.tar.gz	否
prune	dataset.character_list_file	eval_dataset	character_list	否
quantize	dataset.train_dataset_dir	train_datasets	results/{dataset_convert_job_id}/dataset_convert/lmdb	是
quantize	dataset.val_dataset_dir	eval_dataset	results/{dataset_convert_job_id}/dataset_convert/lmdb	否
quantize	dataset.character_list_file	eval_dataset	character_list	否
quantize	dataset.quant_calibration_dataset.images_dir	train_datasets	train.tar.gz	否
retrain	dataset.train_dataset_dir	train_datasets	results/{dataset_convert_job_id}/dataset_convert/lmdb	是
retrain	dataset.val_dataset_dir	eval_dataset	results/{dataset_convert_job_id}/dataset_convert/lmdb	否
retrain	dataset.character_list_file	eval_dataset	character_list	否
train	dataset.train_dataset_dir	train_datasets	results/{dataset_convert_job_id}/dataset_convert/lmdb	是
train	dataset.val_dataset_dir	eval_dataset	results/{dataset_convert_job_id}/dataset_convert/lmdb	否
train	dataset.character_list_file	eval_dataset	character_list	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

{
    "train.num_epochs": 30,
    "train.checkpoint_interval": 10,
    "train.validation_interval": 10,
    "train.num_gpus": 1,
    "dataset.batch_size": 16,
    "dataset.train_dataset_dir": [f"{S3_TRAIN}/results/{dataset_convert_job_id}/dataset_convert/lmdb"],
    "dataset.val_dataset_dir": f"{S3_EVAL}/results/{dataset_convert_job_id}/dataset_convert/lmdb",
    "dataset.character_list_file": f"{S3_EVAL}/character_list",
}

gen_trt_engine (mandatory data sources):

python

{
    "gen_trt_engine.tensorrt.data_type": "fp16",
    "gen_trt_engine.tensorrt.calibration.cal_image_dir": [f"{S3_TRAIN}"],
}

evaluate (mandatory data sources):

python

{
    "dataset.character_list_file": f"{S3_EVAL}/character_list",
    "evaluate.test_dataset_dir": f"{S3_EVAL}/results/{dataset_convert_job_id}/dataset_convert/lmdb",
}

export (mandatory data sources):

python

{
    "dataset.character_list_file": f"{S3_EVAL}/character_list",
}

inference (mandatory data sources):

python

{
    "dataset.character_list_file": f"{S3_EVAL}/character_list",
    "inference.inference_dataset_dir": f"{S3_EVAL}/test.tar.gz",
}

prune (mandatory data sources):

python

{
    "dataset.character_list_file": f"{S3_EVAL}/character_list",
}

quantize (mandatory data sources):

python

{
    "dataset.train_dataset_dir": [f"{S3_TRAIN}/results/{dataset_convert_job_id}/dataset_convert/lmdb"],
    "dataset.val_dataset_dir": f"{S3_EVAL}/results/{dataset_convert_job_id}/dataset_convert/lmdb",
    "dataset.character_list_file": f"{S3_EVAL}/character_list",
    "dataset.quant_calibration_dataset.images_dir": f"{S3_TRAIN}/train.tar.gz",
}

retrain (mandatory data sources):

python

{
    "dataset.train_dataset_dir": [f"{S3_TRAIN}/results/{dataset_convert_job_id}/dataset_convert/lmdb"],
    "dataset.val_dataset_dir": f"{S3_EVAL}/results/{dataset_convert_job_id}/dataset_convert/lmdb",
    "dataset.character_list_file": f"{S3_EVAL}/character_list",
}

数据源覆盖配置对每个操作都是必需的——Agent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（必需数据源）：

python

{
    "train.num_epochs": 30,
    "train.checkpoint_interval": 10,
    "train.validation_interval": 10,
    "train.num_gpus": 1,
    "dataset.batch_size": 16,
    "dataset.train_dataset_dir": [f"{S3_TRAIN}/results/{dataset_convert_job_id}/dataset_convert/lmdb"],
    "dataset.val_dataset_dir": f"{S3_EVAL}/results/{dataset_convert_job_id}/dataset_convert/lmdb",
    "dataset.character_list_file": f"{S3_EVAL}/character_list",
}

gen_trt_engine（必需数据源）：

python

{
    "gen_trt_engine.tensorrt.data_type": "fp16",
    "gen_trt_engine.tensorrt.calibration.cal_image_dir": [f"{S3_TRAIN}"],
}

evaluate（必需数据源）：

python

{
    "dataset.character_list_file": f"{S3_EVAL}/character_list",
    "evaluate.test_dataset_dir": f"{S3_EVAL}/results/{dataset_convert_job_id}/dataset_convert/lmdb",
}

export（必需数据源）：

python

{
    "dataset.character_list_file": f"{S3_EVAL}/character_list",
}

inference（必需数据源）：

python

{
    "dataset.character_list_file": f"{S3_EVAL}/character_list",
    "inference.inference_dataset_dir": f"{S3_EVAL}/test.tar.gz",
}

prune（必需数据源）：

python

{
    "dataset.character_list_file": f"{S3_EVAL}/character_list",
}

quantize（必需数据源）：

python

{
    "dataset.train_dataset_dir": [f"{S3_TRAIN}/results/{dataset_convert_job_id}/dataset_convert/lmdb"],
    "dataset.val_dataset_dir": f"{S3_EVAL}/results/{dataset_convert_job_id}/dataset_convert/lmdb",
    "dataset.character_list_file": f"{S3_EVAL}/character_list",
    "dataset.quant_calibration_dataset.images_dir": f"{S3_TRAIN}/train.tar.gz",
}

retrain（必需数据源）：

python

{
    "dataset.train_dataset_dir": [f"{S3_TRAIN}/results/{dataset_convert_job_id}/dataset_convert/lmdb"],
    "dataset.val_dataset_dir": f"{S3_EVAL}/results/{dataset_convert_job_id}/dataset_convert/lmdb",
    "dataset.character_list_file": f"{S3_EVAL}/character_list",
}

Eval Dataset

评估数据集

Optional. Test data provided as separate tarball.

可选。测试数据以单独压缩包形式提供。

Important Parameters

重要参数

dataset.character_list_file: Path to character list defining the supported character set. This determines the output vocabulary size.
model.backbone: Default ResNet.
model.prediction: Decoder type. CTC or Attn (attention-based).
train.optim.lr: Learning rate. Default 1.0 (Adadelta optimizer). High default is specific to Adadelta.
dataset.batch_size: Per-GPU batch size. Default 16.

dataset.character_list_file：定义支持字符集的字符列表路径。该参数决定输出词汇量大小。
model.backbone：默认值为ResNet。
model.prediction：解码器类型，可选CTC或Attn（基于注意力）。
train.optim.lr：学习率，默认值为1.0（Adadelta优化器）。较高的默认值是Adadelta优化器的特性。
dataset.batch_size：单GPU批次大小，默认值为16。

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.distributed_strategy`	Strategy name	`auto`

Strategy:
```
auto
```
for single-GPU, reads
```
train.distributed_strategy
```
from config when multi-GPU
No explicit
```
num_nodes
```
in train script — single-node oriented
Lightweight model, single GPU typically sufficient

启动方式： Lightning管理（单个

python

进程，Lightning生成工作进程）。

规格键	描述	默认值
`train.num_gpus`	GPU数量	1
`train.gpu_ids`	GPU设备索引	[0]
`train.distributed_strategy`	策略名称	`auto`

策略：单GPU时使用
```
auto
```
，多GPU时从配置中读取
```
train.distributed_strategy
```
训练脚本中无明确的
```
num_nodes
```
参数——面向单节点场景
模型轻量化，通常单GPU足够

Hardware

硬件要求

Minimum 1 GPU(s), recommended 1 GPU(s). 8GB+ VRAM per GPU. OCR text recognition is lightweight. Single GPU is typically sufficient.

最低1块GPU，推荐1块GPU。每块GPU需8GB以上显存。OCR文本识别属于轻量化任务，通常单GPU即可满足需求。

Error Patterns

常见错误模式

dataset_convert required: If using raw images + gt files, run dataset_convert first to produce LMDB format.

Character list mismatch: All characters in training data must be present in the character_list file.

需先执行dataset_convert：如果使用原始图像+标注文件，需先运行dataset_convert生成LMDB格式数据。

字符列表不匹配：训练数据中的所有字符必须存在于character_list文件中。

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

ocrnet.config.json

Action	Spec Field	Inference Function	Meaning
dataset_convert	`results_dir`	`output_dir`	current job results directory
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	`model.pruned_graph_path`	`pruned_model`	parent pruned model
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.tensorrt.calibration.cal_cache_file`	`create_cal_cache`	calibration cache path
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	`model.pruned_graph_path`	`pruned_model`	parent pruned model
inference	`results_dir`	`output_dir`	current job results directory
prune	`encryption_key`	`key`	encryption key
prune	`prune.checkpoint`	`parent_model`	model file inferred from the parent job results folder
prune	`prune.pruned_file`	`create_pth_file`	output PTH path
prune	`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
retrain	`encryption_key`	`key`	encryption key
retrain	`model.pruned_graph_path`	`parent_model`	model file inferred from the parent job results folder
retrain	`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

ocrnet.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`	当前作业结果目录
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	`model.pruned_graph_path`	`pruned_model`	父剪枝模型
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.tensorrt.calibration.cal_cache_file`	`create_cal_cache`	校准缓存路径
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	`model.pruned_graph_path`	`pruned_model`	父剪枝模型
inference	`results_dir`	`output_dir`	当前作业结果目录
prune	`encryption_key`	`key`	加密密钥
prune	`prune.checkpoint`	`parent_model`	从父作业结果文件夹推断出的模型文件
prune	`prune.pruned_file`	`create_pth_file`	输出PTH路径
prune	`results_dir`	`output_dir`	当前作业结果目录
quantize	`encryption_key`	`key`	加密密钥
quantize	`quantize.model_path`	`parent_model`	从父作业结果文件夹推断出的模型文件
quantize	`results_dir`	`output_dir`	当前作业结果目录
retrain	`encryption_key`	`key`	加密密钥
retrain	`model.pruned_graph_path`	`parent_model`	从父作业结果文件夹推断出的模型文件
retrain	`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`	无恢复检查点时使用的预训练模型
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

，将上游训练/导出/AutoML子作业ID作为

parent_job_id

传递。SDK会列出父结果文件夹，过滤检查点工件，并返回所选模型文件或文件夹。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-ocrnet — OCRNet deploy workflow for TensorRT engine generation, TensorRT evaluation, and TensorRT inference using TAO Deploy.

tao-deploy-ocrnet — 用于TensorRT引擎生成、TensorRT评估和TensorRT推理的OCRNet部署工作流，基于TAO Deploy实现。