Tao Train Centerpose

NVIDIAOfficialCreated by NVIDIA Corporation

Skill
Tool integration
Data and AI
Apache-2.0
Medium risk

From the author

CenterPose for keypoint / pose estimation. Detects object centers and regresses keypoint locations for 6-DoF object pose estimation. Use when training, evaluating, exporting, or running inference for a TAO CenterPose model. Trigger phrases include "train CenterPose", "6-DoF object pose", "keypoint estimation", "object pose regression".

— the description by NVIDIA Corporation

AutoGineer summary

Written by AutoGineer from the skill’s own files, not by the author

The tao-train-centerpose skill provides a structured interface for training, evaluating, exporting, and running inference with CenterPose models. CenterPose is designed for keypoint and pose estimation, detecting object centers and regressing keypoint locations to determine 6-DoF object poses. The skill supports AutoML for training optimization and integrates with TAO Deploy for generating TensorRT engines and performing deployment-stage evaluation and inference.

What it does

Trains CenterPose models for 6-DoF object pose estimation.
Supports AutoML policies to automate hyperparameter optimization during training.
Evaluates model performance using 3DIoU and 2DMPE metrics.
Exports trained models to ONNX format for deployment.
Generates TensorRT engines for optimized inference.
Provides templates and schema validation for configuration files.

When to use it

When you need to estimate the 6-DoF pose of objects in images.
When training a new CenterPose model using custom datasets.
When optimizing a model for deployment using TensorRT.
When you need to evaluate model accuracy against test datasets.

Advantages

Automated hyperparameter tuning via AutoML integration.
Standardized workflow for training, evaluation, and deployment.
Consistent configuration management using schema-validated templates.
Support for multiple precision formats including FP32, FP16, and INT8.

Usage examples

“Train a CenterPose model with AutoML enabled.”: The agent initiates the training process using the CenterPose architecture, applying the default AutoML policy to optimize training parameters.
“Export my trained CenterPose model to ONNX.”: The agent triggers the export action, converting the model checkpoint into an ONNX file suitable for deployment.
“Generate a TensorRT engine for my exported CenterPose model.”: The agent uses the CenterPose deploy workflow to convert the ONNX model into a TensorRT engine for optimized execution.
“Evaluate my CenterPose model using the test dataset.”: The agent runs the evaluation action against the specified test data, reporting 3DIoU and 2DMPE metrics.

Install

Download the skill as a ZIP:
https://mp-staging.autogineer.com/agent-skills/nvidia/tao-train-centerpose/download

Unzip it into the cross-client skills folder of your project:

mkdir -p .agents/skills && unzip tao-train-centerpose.zip -d .agents/skills

Documentation by NVIDIA Corporation

CenterPose

Standalone install? If this session was not initialized by the TAO skill bank plugin, run the tao-setup skill first (host preflight, credentials, cross-skill discovery).

CenterPose for keypoint / pose estimation. Detects object centers and regresses keypoint locations. Used for 6-DoF object pose estimation.

Set model.backbone.pretrained_backbone_path.

For TAO Deploy TensorRT actions (gen_trt_engine, TensorRT evaluate, and TensorRT inference), use the deploy spec templates packaged in this skill's references/ folder with the spec_template_deploy_*.yaml prefix.

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 for an action requires schemas/<action>.schema.json and references/spec_template_<action>.yaml to exist and parse. Use the packaged selected-action 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.

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. Use automl_policy: on by default and only expose on / off in new launch prompts. Treat phrases like "turn off AutoML", "disable AutoML", "no HPO", or "plain training" as automl_policy: off for this run only. When automl_policy: on, 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.

Training Requirements

Dataset type: centerpose
Formats: default
Training monitoring metrics: val_3DIoU, val_2DMPE
Evaluate task metrics: test_3DIoU, test_2DMPE
AutoML metric contract: use test_3DIoU with maximize direction for the required evaluation-backed baseline, every recommendation through eval_fn, best-model selection, and final evaluation. val_3DIoU is emitted by training but not by the evaluate action, so it is only suitable for an explicitly accepted training-proxy run without an impact baseline.

Per-Action Dataset Requirements

Action	Spec Key	Source	Files	List?
evaluate	dataset.test_data	eval_dataset	test.tar.gz	No
gen_trt_engine	gen_trt_engine.tensorrt.calibration.cal_image_dir	calibration_dataset	train.tar.gz	Yes
inference	dataset.inference_data	inference_dataset	val.tar.gz	No
train	dataset.train_data	train_datasets	train.tar.gz	No
train	dataset.val_data	eval_dataset	val.tar.gz	No

Typical Spec Overrides

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.

TRAIN_DIR = "/path/to/extracted/train"
VAL_DIR = "/path/to/extracted/val"
TEST_DIR = "/path/to/extracted/test"
INFER_DIR = VAL_DIR
CAL_IMAGE_DIRS = ["/path/to/extracted/train/<sequence_or_image_dir>"]

train (mandatory data sources):

{
    "train.num_epochs": 30,
    "train.checkpoint_interval": 10,
    "train.validation_interval": 10,
    "train.num_gpus": 1,
    "dataset.category": "bike",
    "dataset.batch_size": 4,
    "dataset.train_data": TRAIN_DIR,
    "dataset.val_data": VAL_DIR,
}

evaluate (mandatory data sources):

{
    "dataset.category": "bike",
    "dataset.test_data": TEST_DIR,
}

inference (mandatory data sources):

{
    "dataset.category": "bike",
    "dataset.inference_data": INFER_DIR,
}

gen_trt_engine (mandatory data sources):

{
    "gen_trt_engine.tensorrt.calibration.cal_image_dir": CAL_IMAGE_DIRS,
}

Eval Dataset

Optional. Val and test datasets are provided as separate tarballs. Training writes val_3DIoU/val_2DMPE KPIs, while evaluate writes test_3DIoU/test_2DMPE; do not configure an evaluate-backed AutoML callback to extract the training-prefixed names.

Important Parameters

dataset.num_classes: Number of object categories. Default 1.
dataset.num_joints: Number of keypoints per object. Fixed at 8 (bbox keypoints). Valid range: exactly 8.
dataset.input_res: Input resolution. Fixed at 512. Output resolution fixed at 128.
dataset.category: Object category name. Default "cereal_box".
model.backbone.model_type: Default fan_small. Backbone options limited in schema.
train.optim.lr: Learning rate. Default 6e-5. MultiStep scheduler with lr_steps=[90, 120], lr_decay=0.1.
train.loss_config: Rich loss config with toggles: mse_loss, obj_scale, obj_scale_uncertainty, hps_uncertainty, reg_bbox, hm_hp. Weights: wh_weight=0.1, off_weight=1, hp_weight=1.
inference.use_pnp: Use PnP for 6-DoF pose. Default True. Requires camera intrinsics (focal_length_x/y, principle_point_x/y).
export.input_width: Export input size. Fixed at 512x512. opset_version=16.

Multi-GPU / Multi-Node

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]

Strategy: auto (Lightning picks the best strategy automatically)
No explicit num_nodes or distributed_strategy config — single-node only
No sync_batchnorm

Export / TRT Defaults

Export input: 512x512 (fixed), opset 16
TRT data types: FP32, FP16, INT8
TRT opt_batch_size: 4, max_batch_size: 8

Hardware

Minimum 1 GPU(s), recommended 2 GPU(s). 16GB+ VRAM per GPU. CenterPose is moderately memory-intensive depending on input resolution and number of keypoints.

Error Patterns

num_joints mismatch: Ensure dataset.num_joints matches the keypoint count in your annotations.

Extract S3 tarballs for local Docker: The starter-kit S3 data is packaged as train.tar.gz, val.tar.gz, and test.tar.gz, but the CenterPose TAO actions consume extracted folders. Extract each archive and set dataset.train_data, dataset.val_data, dataset.test_data, and dataset.inference_data to the extracted split directories.

Checkpoint handoff: CenterPose training writes concrete checkpoints such as model_epoch_000_step_00008.pth and a centerpose_model_latest.pth symlink. Use the SDK/model checkpoint resolver or the exact epoch/step checkpoint for evaluate, inference, export, and resume. Use the symlink only when the user explicitly asks for latest.

TAO Deploy postprocessor compatibility: Use the deploy image resolved from the skill's pinned deploy image or the selected platform. A successful gen_trt_engine run does not prove deploy evaluate or inference works; inspect those action exit codes and logs separately, especially for CenterPose postprocessor errors such as TypeError: only 0-dimensional arrays can be converted to Python scalars.

Spec Param / Parent Model Inference

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 centerpose.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.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	`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.resume_training_checkpoint_path`	`resume_model`	model file inferred from the current job results folder

For parent_model or 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.

Deployment

tao-deploy-centerpose

Files

ATTRIBUTION.mdAdded by AutoGineer930 B
BENCHMARK.md7.0 KiB
LICENSELicence file added by AutoGineer11.2 KiB
SKILL.md11.2 KiB
config/skillspector-baseline.yaml834 B
evals/evals.json786 B
references/skill_info.yaml3.4 KiB
references/spec_template_deploy_evaluate.yaml256 B
references/spec_template_deploy_gen_trt_engine.yaml453 B
references/spec_template_deploy_inference.yaml471 B
references/spec_template_evaluate.yaml2.4 KiB
references/spec_template_export.yaml2.5 KiB
references/spec_template_gen_trt_engine.yaml2.6 KiB
references/spec_template_inference.yaml2.6 KiB
references/spec_template_train.yaml2.2 KiB
references/tao-deploy-centerpose.md5.5 KiB
references/tao-deploy-centerpose.skill_info.yaml2.0 KiB
schemas/evaluate.schema.json35.2 KiB
schemas/export.schema.json35.4 KiB
schemas/gen_trt_engine.schema.json40.6 KiB
schemas/inference.schema.json37.9 KiB
schemas/manifest.json9.6 KiB
schemas/train.schema.json32.0 KiB
skill-card.md4.5 KiB
skill.oms.sig8.9 KiB

schemas/export.schema.json

{
  "automl_default_parameters": [
    "train.optim.lr_decay",
    "train.optim.lr"
  ],
  "automl_disabled_parameters": [
    "train.cudnn",
    "gen_trt_engine.tensorrt.calibration.cal_image_dir",
    "train.gpu_ids",
    "wandb.tags",
    "model.backbone",
    "dataset.std",
    "evaluate",
    "inference",
    "train.loss_config",
    "train",
    "gen_trt_engine",
    "gen_trt_engine.tensorrt.layers_precision",
    "dataset",
    "gen_trt_engine.tensorrt",
    "dataset._eig_val",
    "dataset.mean",
    "model",
    "train.optim.lr_steps",
    "dataset.flip_idx",
    "evaluate.gpu_ids",
    "train.optim",
    "gen_trt_engine.tensorrt.calibration",
    "dataset._eig_vec",
    "export",
    "wandb",
    "inference.gpu_ids"
  ],
  "default": {
    "dataset": {
      "_eig_val": [
        0.2141788,
        0.01817699,
        0.00341571
      ],
      "_eig_vec": [
        [
          -0.58752847,
          -0.69563484,
          0.41340352
        ],
        [
          -0.5832747,
          0.00994535,
          -0.81221408
        ],
        [
          -0.56089297,
          0.71832671,
          0.41158938
        ]
      ],
      "aug_rot": 0,
      "batch_size": 4,
      "category": "cereal_box",
      "center_3D": false,
      "dense_hp": false,
      "flip": 0.5,
      "flip_idx": [
        [
          1,
          5
        ],
        [
          3,
          7
        ],
        [
          2,
          6
        ],
        [
          4,
          8
        ]
      ],
      "hm_hp": true,
      "hps_uncertainty": false,
      "inference_data": "",
      "input_res": 512,
      "max_objs": 10,
      "mean": [
        0.40789654,
        0.44719302,
        0.47026115
      ],
      "mse_loss": false,
      "no_color_aug": false,
      "not_rand_crop": false,
      "num_classes": 1,
      "num_joints": 8,
      "num_symmetry": 1,
      "obj_scale": true,
      "obj_scale_uncertainty": false,
      "output_res": 128,
      "pin_memory": true,
      "reg_bbox": true,
      "reg_hp_offset": true,
      "reg_offset": true,
      "std": [
        0.28863828,
        0.27408164,
        0.27809835
      ],
      "test_data": "",
      "train_data": "",
      "use_absolute_scale": false,
      "val_data": "",
      "workers": 8
    },
    "encryption_key": "",
    "export": {
      "batch_size": -1,
      "checkpoint": "",
      "do_constant_folding": true,
      "gpu_id": 0,
      "input_channel": 3,
      "input_height": 512,
      "input_width": 512,
      "num_select": 100,
      "on_cpu": false,
      "onnx_file": "",
      "opset_version": 16,
      "results_dir": "",
      "verbose": false
    },
    "model": {
      "backbone": {
        "model_type": "fan_small",
        "pretrained_backbone_path": ""
      },
      "down_ratio": 4,
      "final_kernel": 1,
      "head_conv": 256,
      "last_level": 5,
      "out_channel": 0,
      "use_convGRU": true,
      "use_pretrained": false
    },
    "model_name": "",
    "results_dir": "",
    "train": {
      "checkpoint_interval": 1,
      "checkpoint_interval_unit": "epoch",
      "clip_grad_val": 100.0,
      "cudnn": {
        "benchmark": false,
        "deterministic": true
      },
      "gpu_ids": [
        0
      ],
      "is_dry_run": false,
      "loss_config": {
        "dense_hp": false,
        "dimension_ref": "",
        "hm_hp": true,
        "hm_hp_weight": 1,
        "hm_weight": 1,
        "hp_weight": 1,
        "hps_uncertainty": false,
        "mse_loss": false,
        "num_stacks": 1,
        "obj_scale": true,
        "obj_scale_uncertainty": false,
        "obj_scale_weight": 1,
        "off_weight": 1,
        "reg_bbox": true,
        "reg_hp_offset": true,
        "reg_loss": "l1",
        "reg_offset": true,
        "use_residual": false,
        "wh_weight": 0.1
      },
      "num_epochs": 10,
      "num_gpus": 1,
      "num_nodes": 1,
      "optim": {
        "lr": 6e-05,
        "lr_decay": 0.1,
        "lr_scheduler": "MultiStep",
        "lr_steps": [
          90,
          120
        ]
      },
      "precision": "fp32",
      "pretrained_model_path": "",
      "results_dir": "",
      "resume_training_checkpoint_path": "",
      "seed": 1234,
      "validation_interval": 1
    },
    "wandb": {
      "enable": true,
      "entity": "",
      "group": "",
      "name": "TAO Toolkit Training",
      "project": "TAO Toolkit",
      "reinit": false,
      "run_id": "",
      "save_code": false,
      "sync_tensorboard": false,
      "tags": [
        "tao-toolkit"
      ]
    }
  },
  "popular": {
    "evaluate": {
      "gpu_ids": [
        0
      ],
      "num_gpus": 1,
      "num_nodes": 1
    },
    "gen_trt_engine": {
      "batch_size": -1,
      "gpu_id": 0,
      "tensorrt": {
        "calibration": {
          "cal_batch_size": 1,
          "cal_batches": 1
        },
        "max_batch_size": 8,
        "min_batch_size": 1,
        "opt_batch_size": 4
      }
    },
    "inference": {
      "gpu_ids": [
        0
      ],
      "num_gpus": 1,
      "num_nodes": 1
    },
    "train": {
      "checkpoint_interval": 1,
      "gpu_ids": [
        0
      ],
      "num_epochs": 10,
      "num_gpus": 1,
      "num_nodes": 1,
      "validation_interval": 1
    }
  },
  "properties": {
    "automl_disabled_parameters": [
      "wandb",
      "dataset",
      "train",
      "model",
      "inference",
      "export",
      "evaluate",
      "gen_trt_engine"
    ],
    "dataset": {
      "automl_disabled_parameters": [
        "dataset.mean",
        "dataset.std",
        "dataset._eig_val",
        "dataset._eig_vec",
        "dataset.flip_idx"
      ],
      "automl_enabled": false,
      "default": {
        "_eig_val": [
          0.2141788,
          0.01817699,
          0.00341571
        ],
        "_eig_vec": [
          [
            -0.58752847,
            -0.69563484,
            0.41340352
          ],
          [
            -0.5832747,
            0.00994535,
            -0.81221408
          ],
          [
            -0.56089297,
            0.71832671,
            0.41158938
          ]
        ],
        "aug_rot": 0,
        "batch_size": 4,
        "category": "cereal_box",
        "center_3D": false,
        "dense_hp": false,
        "flip": 0.5,
        "flip_idx": [
          [
            1,
            5
          ],
          [
            3,
            7
          ],
          [
            2,
            6
          ],
          [
            4,
            8
          ]
        ],
        "hm_hp": true,
        "hps_uncertainty": false,
        "inference_data": "",
        "input_res": 512,
        "max_objs": 10,
        "mean": [
          0.40789654,
          0.44719302,
          0.47026115
        ],
        "mse_loss": false,
        "no_color_aug": false,
        "not_rand_crop": false,
        "num_classes": 1,
        "num_joints": 8,
        "num_symmetry": 1,
        "obj_scale": true,
        "obj_scale_uncertainty": false,
        "output_res": 128,
        "pin_memory": true,
        "reg_bbox": true,
        "reg_hp_offset": true,
        "reg_offset": true,
        "std": [
          0.28863828,
          0.27408164,
          0.27809835
        ],
        "test_data": "",
        "train_data": "",
        "use_absolute_scale": false,
        "val_data": "",
        "workers": 8
      },
      "description": "Configurable parameters to construct the dataset for a CenterPose experiment.",
      "properties": {
        "_eig_val": {
          "automl_enabled": false,
          "default": [
            0.2141788,
            0.01817699,
            0.00341571
          ],
          "description": "Eigenvalues for color data augmentation from CenterNet.",
          "type": "list"
        },
        "_eig_vec": {
          "automl_enabled": false,
          "default": [
            [
              -0.58752847,
              -0.69563484,
              0.41340352
            ],
            [
              -0.5832747,
              0.00994535,
              -0.81221408
            ],
            [
              -0.56089297,
              0.71832671,
              0.41158938
            ]
          ],
          "description": "Eigenvectors for color data augmentation from CenterNet.",
          "type": "list"
        },
        "aug_rot": {
          "default": 0,
          "description": "Rotation angle for data augmentation.",
          "maximum": Infinity,
          "minimum": 0,
          "title": "Rotation Angle",
          "type": "int"
        },
        "batch_size": {
          "default": 4,
          "description": "Batch size.",
          "maximum": Infinity,
          "minimum": 1,
          "title": "Batch Size",
          "type": "int"
        },
        "category": {
          "default": "cereal_box",
          "description": "Category of the object.",
          "title": "Category",
          "type": "string"
        },
        "center_3D": {
          "default": false,
          "description": "Use 3D center loss for the object.",
          "title": "3D Center",
          "type": "bool"
        },
        "dense_hp": {
          "default": false,
          "description": "Use dense heatmaps.",
          "title": "Dense Heatmaps",
          "type": "bool"
        },
        "flip": {
          "default": 0.5,
          "description": "Flip probability for data augmentation.",
          "maximum": 1.0,
          "minimum": 0.1,
          "title": "Flip Probability",
          "type": "float"
        },
        "flip_idx": {
          "automl_enabled": false,
          "default": [
            [
              1,
              5
            ],
            [
              3,
              7
            ],
            [
              2,
              6
            ],
            [
              4,
              8
            ]
          ],
          "description": "Flipping indices for keypoints.",
          "type": "list"
        },
        "hm_hp": {
          "default": true,
          "description": "Use heatmaps for keypoints.",
          "title": "Heatmaps for Keypoints",
          "type": "bool"
        },
        "hps_uncertainty": {
          "default": false,
          "description": "Use heatmaps uncertainty loss.",
          "title": "Heatmaps Uncertainty",
          "type": "bool"
        },
        "inference_data": {
          "default": "",
          "description": "Path to inference data.",
          "title": "Inference Data",
          "type": "string"
        },
        "input_res": {
          "default": 512,
          "description": "Input resolution.",
          "maximum": 512,
          "minimum": 512,
          "title": "Input Resolution",
          "type": "int"
        },
        "max_objs": {
          "default": 10,
          "description": "Maximum detected number of objects.",
          "maximum": Infinity,
          "minimum": 1,
          "title": "Maximum Detected Objects",
          "type": "int"
        },
        "mean": {
          "automl_enabled": false,
          "default": [
            0.40789654,
            0.44719302,
            0.47026115
          ],
          "description": "Mean values for normalization.",
          "title": "Mean",
          "type": "list"
        },
        "mse_loss": {
          "default": false,
          "description": "Use mean squared error loss.",
          "title": "Mean Squared Error Loss",
          "type": "bool"
        },
        "no_color_aug": {
          "default": false,
          "description": "No color augmentation.",
          "title": "No Color Augmentation",
          "type": "bool"
        },
        "not_rand_crop": {
          "default": false,
          "description": "No random cropping.",
          "title": "No Random Cropping",
          "type": "bool"
        },
        "num_classes": {
          "default": 1,
          "description": "Number of classes.",
          "maximum": Infinity,
          "minimum": 1,
          "title": "Number of Classes",
          "type": "int"
        },
        "num_joints": {
          "default": 8,
          "description": "Number of 3D bounding box keypoints.",
          "maximum": 8,
          "minimum": 8,
          "title": "Number of Keypoints",
          "type": "int"
        },
        "num_symmetry": {
          "default": 1,
          "description": "Number of the object symmetries.",
          "maximum": Infinity,
          "minimum": 1,
          "title": "Number of Symmetries",
          "type": "int"
        },
        "obj_scale": {
          "default": true,
          "description": "Use object scale loss.",
          "title": "Object Scale",
          "type": "bool"
        },
        "obj_scale_uncertainty": {
          "default": false,
          "description": "Use object scale uncertainty loss.",
          "title": "Object Scale Uncertainty",
          "type": "bool"
        },
        "output_res": {
          "default": 128,
          "description": "Output resolution.",
          "maximum": 128,
          "minimum": 128,
          "title": "Output Resolution",
          "type": "int"
        },
        "pin_memory": {
          "default": true,
          "description": "Pin memory.",
          "title": "Pin Memory",
          "type": "bool"
        },
        "reg_bbox": {
          "default": true,
          "description": "Use bounding box regression loss.",
          "title": "Bounding Box Regression",
          "type": "bool"
        },
        "reg_hp_offset": {
          "default": true,
          "description": "Use offset regression loss for keypoints.",
          "title": "Offset Regression for Keypoints",
          "type": "bool"
        },
        "reg_offset": {
          "default": true,
          "description": "Use offset regression loss.",
          "title": "Offset Regression",
          "type": "bool"
        },
        "std": {
          "automl_enabled": false,
          "default": [
            0.28863828,
            0.27408164,
            0.27809835
          ],
          "description": "Standard deviation values for normalization.",
          "title": "Standard Deviation",
          "type": "list"
        },
        "test_data": {
          "default": "",
          "description": "Path to testing data.",
          "title": "Testing Data",
          "type": "string"
        },
        "train_data": {
          "default": "",
          "description": "Path to training data.",
          "title": "Training Data",
          "type": "string"
        },
        "use_absolute_scale": {
          "default": false,
          "description": "Use absolute scale loss.",
          "title": "Absolute Scale",
          "type": "bool"
        },
        "val_data": {
          "default": "",
          "description": "Path to validation data.",
          "title": "Validation Data",
          "type": "string"
        },
        "workers": {
          "default": 8,
          "description": "Number of workers.",
          "maximum": Infinity,
          "minimum": 1,
          "title": "Workers",
          "type": "int"
        }
      },
      "type": "collection"
    },
    "encryption_key": {
      "default": "",
      "description": "Key for encrypting model checkpoints",
      "title": "Encryption key",
      "type": "string"
    },
    "export": {
      "automl_enabled": false,
      "default": {
        "batch_size": -1,
        "checkpoint": "",
        "do_constant_folding": true,
        "gpu_id": 0,
        "input_channel": 3,
        "input_height": 512,
        "input_width": 512,
        "num_select": 100,
        "on_cpu": false,
        "onnx_file": "",
        "opset_version": 16,
        "results_dir": "",
        "verbose": false
      },
      "description": "Configurable parameters to export the CenterPose ONNX model.",
      "properties": {
        "batch_size": {
          "default": -1,
          "description": "ONNX model batch size (-1: dynamic).",
          "title": "ONNX Batch Size",
          "type": "int"
        },
        "checkpoint": {
          "default": "",
          "description": "Path to the checkpoint.",
          "title": "Checkpoint",
          "type": "string"
        },
        "do_constant_folding": {
          "default": true,
          "description": "Do constant folding on ONNX model.",
          "title": "Constant Folding",
          "type": "bool"
        },
        "gpu_id": {
          "default": 0,
          "description": "GPU ID used for training.",
          "title": "GPU ID",
          "type": "int"
        },
        "input_channel": {
          "default": 3,
          "description": "Input channel.",
          "maximum": 3,
          "minimum": 3,
          "title": "Input Channel",
          "type": "int"
        },
        "input_height": {
          "default": 512,
          "description": "Input height.",
          "maximum": 512,
          "minimum": 512,
          "title": "Input Height",
          "type": "int"
        },
        "input_width": {
          "default": 512,
          "description": "Input width.",
          "maximum": 512,
          "minimum": 512,
          "title": "Input Width",
          "type": "int"
        },
        "num_select": {
          "default": 100,
          "description": "Number of selected objects.",
          "maximum": 100,
          "minimum": 100,
          "title": "Number of Selected Objects",
          "type": "int"
        },
        "on_cpu": {
          "default": false,
          "description": "Export the ONNX using CPU only.",
          "title": "ONNX_CPU",
          "type": "bool"
        },
        "onnx_file": {
          "default": "",
          "description": "Path to the ONNX file.",
          "title": "ONNX File",
          "type": "string"
        },
        "opset_version": {
          "default": 16,
          "description": "ONNX opset version.",
          "maximum": 16,
          "minimum": 16,
          "title": "Opset Version",
          "type": "int"
        },
        "results_dir": {
          "default": "",
          "description": "Results directory.",
          "title": "Results Directory",
          "type": "string"
        },
        "verbose": {
          "default": false,
          "description": "Verbose mode.",
          "title": "Verbose",
          "type": "bool"
        }
      },
      "type": "collection"
    },
    "model": {
      "automl_disabled_parameters": [
        "model.backbone"
      ],
      "automl_enabled": false,
      "default": {
        "backbone": {
          "model_type": "fan_small",
          "pretrained_backbone_path": ""
        },
        "down_ratio": 4,
        "final_kernel": 1,
        "head_conv": 256,
        "last_level": 5,
        "out_channel": 0,
        "use_convGRU": true,
        "use_pretrained": false
      },
      "description": "Configurable parameters to build the CenterPose model.",
      "properties": {
        "backbone": {
          "automl_enabled": false,
          "default": {
            "model_type": "fan_small",
            "pretrained_backbone_path": ""
          },
          "description": "Backbone model config.",
          "properties": {
            "model_type": {
              "default": "fan_small",
              "description": "Model type.",
              "title": "Model Type",
              "type": "string"
            },
            "pretrained_backbone_path": {
              "default": "",
              "description": "Path to the pretrained backbone model.",
              "title": "Pretrained Backbone Model",
              "type": "string"
            }
          },
          "title": "Backbone Model",
          "type": "collection"
        },
        "down_ratio": {
          "default": 4,
          "description": "Down ratio.",
          "maximum": 4,
          "minimum": 4,
          "title": "Down Ratio",
          "type": "int"
        },
        "final_kernel": {
          "default": 1,
          "description": "Final kernel size.",
          "maximum": 1,
          "minimum": 1,
          "title": "Final Kernel Size",
          "type": "int"
        },
        "head_conv": {
          "default": 256,
          "description": "Head convolution.",
          "maximum": 256,
          "minimum": 256,
          "title": "Head Convolution",
          "type": "int"
        },
        "last_level": {
          "default": 5,
          "description": "Last level.",
          "maximum": 5,
          "minimum": 5,
          "title": "Last Level",
          "type": "int"
        },
        "out_channel": {
          "default": 0,
          "description": "Output channel.",
          "maximum": 0,
          "minimum": 0,
          "title": "Output Channel",
          "type": "int"
        },
        "use_convGRU": {
          "default": true,
          "description": "Use convolutional GRU.",
          "title": "Convolutional GRU",
          "type": "bool"
        },
        "use_pretrained": {
          "default": false,
          "description": "Use pretrained model.",
          "title": "Pretrained Model",
          "type": "bool"
        }
      },
      "type": "collection"
    },
    "model_name": {
      "default": "",
      "description": "Name of model if invoking task via :code:`model_agnostic`",
      "title": "Model name",
      "type": "string"
    },
    "results_dir": {
      "default": "",
      "description": "\n        Path to where all the assets generated from a task are stored.\n        ",
      "title": "Results directory",
      "type": "string"
    },
    "train": {
      "automl_disabled_parameters": [
        "train.gpu_ids",
        "train.cudnn",
        "train.optim",
        "train.loss_config"
      ],
      "automl_enabled": false,
      "default": {
        "checkpoint_interval": 1,
        "checkpoint_interval_unit": "epoch",
        "clip_grad_val": 100.0,
        "cudnn": {
          "benchmark": false,
          "deterministic": true
        },
        "gpu_ids": [
          0
        ],
        "is_dry_run": false,
        "loss_config": {
          "dense_hp": false,
          "dimension_ref": "",
          "hm_hp": true,
          "hm_hp_weight": 1,
          "hm_weight": 1,
          "hp_weight": 1,
          "hps_uncertainty": false,
          "mse_loss": false,
          "num_stacks": 1,
          "obj_scale": true,
          "obj_scale_uncertainty": false,
          "obj_scale_weight": 1,
          "off_weight": 1,
          "reg_bbox": true,
          "reg_hp_offset": true,
          "reg_loss": "l1",
          "reg_offset": true,
          "use_residual": false,
          "wh_weight": 0.1
        },
        "num_epochs": 10,
        "num_gpus": 1,
        "num_nodes": 1,
        "optim": {
          "lr": 6e-05,
          "lr_decay": 0.1,
          "lr_scheduler": "MultiStep",
          "lr_steps": [
            90,
            120
          ]
        },
        "precision": "fp32",
        "pretrained_model_path": "",
        "results_dir": "",
        "resume_training_checkpoint_path": "",
        "seed": 1234,
        "validation_interval": 1
      },
      "description": "Configurable parameters to train the CenterPose model.",
      "popular": [
        "validation_interval",
        "checkpoint_interval",
        "num_gpus",
        "num_nodes",
        "num_epochs",
        "gpu_ids"
      ],
      "properties": {
        "checkpoint_interval": {
          "default": 1,
          "description": "The interval (in epochs) at which a checkpoint will be saved. Helps resume training.",
          "minimum": 1,
          "popular": true,
          "title": "Checkpoint interval",
          "type": "int"
        },
        "checkpoint_interval_unit": {
          "default": "epoch",
          "description": "The unit of the checkpoint interval.",
          "enum": [
            "epoch",
            "step"
          ],
          "title": "Checkpoint interval unit",
          "type": "categorical"
        },
        "clip_grad_val": {
          "default": 100.0,
          "description": "Gradient clipping value.",
          "maximum": Infinity,
          "minimum": 1.0,
          "title": "Gradient Clipping Value",
          "type": "float"
        },
        "cudnn": {
          "automl_enabled": false,
          "default": {
            "benchmark": false,
            "deterministic": true
          },
          "properties": {
            "benchmark": {
              "default": false,
              "type": "bool"
            },
            "deterministic": {
              "default": true,
              "type": "bool"
            }
          },
          "type": "collection"
        },
        "gpu_ids": {
          "automl_enabled": false,
          "default": [
            0
          ],
          "description": "\n        List of GPU IDs to run the training on. The length of this list\n        must be equal to the number of gpus in train.num_gpus.",
          "popular": true,
          "title": "GPU IDs",
          "type": "list"
        },
        "is_dry_run": {
          "default": false,
          "description": "Run a training iteration without model saving.",
          "title": "Dry Run",
          "type": "bool"
        },
        "loss_config": {
          "automl_enabled": false,
          "default": {
            "dense_hp": false,
            "dimension_ref": "",
            "hm_hp": true,
            "hm_hp_weight": 1,
            "hm_weight": 1,
            "hp_weight": 1,
            "hps_uncertainty": false,
            "mse_loss": false,
            "num_stacks": 1,
            "obj_scale": true,
            "obj_scale_uncertainty": false,
            "obj_scale_weight": 1,
            "off_weight": 1,
            "reg_bbox": true,
            "reg_hp_offset": true,
            "reg_loss": "l1",
            "reg_offset": true,
            "use_residual": false,
            "wh_weight": 0.1
          },
          "description": "Model loss configuration.",
          "properties": {
            "dense_hp": {
              "default": false,
              "description": "Use dense heatmaps.",
              "title": "Dense Heatmaps",
              "type": "bool"
            },
            "dimension_ref": {
              "default": "",
              "description": "Dimension reference.",
              "title": "Dimension Reference",
              "type": "string"
            },
            "hm_hp": {
              "default": true,
              "description": "Use heatmaps for keypoints.",
              "title": "Heatmaps for Keypoints",
              "type": "bool"
            },
            "hm_hp_weight": {
              "default": 1,
              "description": "Weight for heatmaps for keypoints.",
              "maximum": 1,
              "minimum": 1,
              "title": "Heatmaps for Keypoints Weight",
              "type": "int"
            },
            "hm_weight": {
              "default": 1,
              "description": "Weight for heatmaps.",
              "maximum": 1,
              "minimum": 1,
              "title": "Heatmaps Weight",
              "type": "int"
            },
            "hp_weight": {
              "default": 1,
              "description": "Weight for keypoints.",
              "maximum": 1,
              "minimum": 1,
              "title": "Keypoints Weight",
              "type": "int"
            },
            "hps_uncertainty": {
              "default": false,
              "description": "Use heatmaps uncertainty loss.",
              "title": "Heatmaps Uncertainty",
              "type": "bool"
            },
            "mse_loss": {
              "default": false,
              "description": "Use mean squared error loss.",
              "title": "Mean Squared Error Loss",
              "type": "bool"
            },
            "num_stacks": {
              "default": 1,
              "description": "Number of stacks.",
              "maximum": 1,
              "minimum": 1,
              "title": "Number of Stacks",
              "type": "int"
            },
            "obj_scale": {
              "default": true,
              "description": "Use object scale loss.",
              "title": "Object Scale",
              "type": "bool"
            },
            "obj_scale_uncertainty": {
              "default": false,
              "description": "Use object scale uncertainty loss.",
              "title": "Object Scale Uncertainty",
              "type": "bool"
            },
            "obj_scale_weight": {
              "default": 1,
              "description": "Weight for object scale loss.",
              "maximum": 1,
              "minimum": 1,
              "title": "Object Scale Weight",
              "type": "int"
            },
            "off_weight": {
              "default": 1,
              "description": "Weight for offset loss.",
              "maximum": 1,
              "minimum": 1,
              "title": "Offset Weight",
              "type": "int"
            },
            "reg_bbox": {
              "default": true,
              "description": "Use bounding box regression loss.",
              "title": "Bounding Box Regression",
              "type": "bool"
            },
            "reg_hp_offset": {
              "default": true,
              "description": "Use offset regression loss for keypoints.",
              "title": "Offset Regression for Keypoints",
              "type": "bool"
            },
            "reg_loss": {
              "default": "l1",
              "description": "Regression loss function.",
              "title": "Regression Loss Function",
              "type": "string"
            },
            "reg_offset": {
              "default": true,
              "description": "Use offset regression loss.",
              "title": "Offset Regression",
              "type": "bool"
            },
            "use_residual": {
              "default": false,
              "description": "Use residual loss.",
              "title": "Residual Loss",
              "type": "bool"
            },
            "wh_weight": {
              "default": 0.1,
              "description": "Weight for width and height loss.",
              "maximum": 0.1,
              "minimum": 0.1,
              "title": "Width and Height Weight",
              "type": "float"
            }
          },
          "title": "Loss Config",
          "type": "collection"
        },
        "num_epochs": {
          "default": 10,
          "description": "Number of epochs to run the training.",
          "maximum": Infinity,
          "minimum": 1,
          "popular": true,
          "title": "Number of epochs",
          "type": "int"
        },
        "num_gpus": {
          "default": 1,
          "description": "The number of GPUs to run the train job.",
          "minimum": 1,
          "popular": true,
          "title": "Number of GPUs",
          "type": "int"
        },
        "num_nodes": {
          "default": 1,
          "description": "Number of nodes to run the training on. If > 1, then multi-node is enabled.",
          "minimum": 1,
          "popular": true,
          "title": "Number of nodes",
          "type": "int"
        },
        "optim": {
          "automl_default_parameters": [
            "train.optim.lr",
            "train.optim.lr_decay"
          ],
          "automl_disabled_parameters": [
            "train.optim.lr_steps"
          ],
          "automl_enabled": false,
          "default": {
            "lr": 6e-05,
            "lr_decay": 0.1,
            "lr_scheduler": "MultiStep",
            "lr_steps": [
              90,
              120
            ]
          },
          "description": "Model optimizer configuration.",
          "properties": {
            "lr": {
              "automl_enabled": true,
              "default": 6e-05,
              "description": "Learning rate.",
              "maximum": Infinity,
              "minimum": 0.0,
              "title": "Learning Rate",
              "type": "float"
            },
            "lr_decay": {
              "automl_enabled": true,
              "default": 0.1,
              "description": "Learning rate decay.",
              "maximum": Infinity,
              "minimum": 0.0,
              "title": "Learning Rate Decay",
              "type": "float"
            },
            "lr_scheduler": {
              "default": "MultiStep",
              "description": "Learning rate scheduler.",
              "title": "Learning Rate Scheduler",
              "type": "string"
            },
            "lr_steps": {
              "automl_enabled": false,
              "default": [
                90,
                120
              ],
              "description": "Learning rate steps.",
              "title": "Learning Rate Steps",
              "type": "list"
            }
          },
          "title": "Optimizer Config",
          "type": "collection"
        },
        "precision": {
          "default": "fp32",
          "description": "Training precision.",
          "title": "Precision",
          "type": "string"
        },
        "pretrained_model_path": {
          "default": "",
          "description": "Path to the pretrained model.",
          "title": "Pretrained Model",
          "type": "string"
        },
        "results_dir": {
          "default": "",
          "description": "\n        Path to where all the assets generated from a task are stored.\n        ",
          "title": "Results directory",
          "type": "string"
        },
        "resume_training_checkpoint_path": {
          "default": "",
          "description": "Path to the checkpoint to resume training from.",
          "title": "Resume checkpoint path",
          "type": "string"
        },
        "seed": {
          "default": 1234,
          "description": "The seed for the initializer in PyTorch. If < 0, disable fixed seed.",
          "maximum": Infinity,
          "minimum": -1,
          "title": "Seed for randomization",
          "type": "int"
        },
        "validation_interval": {
          "default": 1,
          "description": "\n        The interval (in epochs) at which a evaluation\n        will be triggered on the validation dataset.",
          "minimum": 1,
          "popular": true,
          "title": "Validation interval",
          "type": "int"
        }
      },
      "type": "collection"
    },
    "wandb": {
      "automl_disabled_parameters": [
        "wandb.tags"
      ],
      "automl_enabled": false,
      "default": {
        "enable": true,
        "entity": "",
        "group": "",
        "name": "TAO Toolkit Training",
        "project": "TAO Toolkit",
        "reinit": false,
        "run_id": "",
        "save_code": false,
        "sync_tensorboard": false,
        "tags": [
          "tao-toolkit"
        ]
      },
      "properties": {
        "enable": {
          "default": true,
          "type": "bool"
        },
        "entity": {
          "default": "",
          "type": "string"
        },
        "group": {
          "default": "",
          "type": "string"
        },
        "name": {
          "default": "TAO Toolkit Training",
          "type": "string"
        },
        "project": {
          "default": "TAO Toolkit",
          "type": "string"
        },
        "reinit": {
          "default": false,
          "type": "bool"
        },
        "run_id": {
          "default": "",
          "type": "string"
        },
        "save_code": {
          "default": false,
          "type": "bool"
        },
        "sync_tensorboard": {
          "default": false,
          "type": "bool"
        },
        "tags": {
          "automl_enabled": false,
          "default": [
            "tao-toolkit"
          ],
          "type": "list"
        }
      },
      "type": "collection"
    }
  },
  "type": "object",
  "x_tao_schema": {
    "action": "export",
    "core_module": "centerpose",
    "model": "centerpose",
    "network_arch": "centerpose",
    "schema_action": "export",
    "schema_version": 1,
    "source": "tao-core dataclass config"
  }
}

Requirements and compatibility

Works with: Claude Code, Claude apps, OpenAI Codex, GitHub Copilot, Cursor, Gemini CLI, Kiro, OpenCode, Goose, Amp, Roo Code, JetBrains Junie, Factory Droid, Devin (ex-Windsurf), Mistral Vibe, OpenHands, Letta

“Works with” means the tool reads the open Agent Skills format; a skill written for one tool may use features others lack.

Needs: Declares allowed tools, Needs Python, Needs Node.js

The author says: Requires docker + nvidia-container-toolkit.

Security

What our scanner noticed· Medium risk

Large base64-encoded blobs

A static scan is not a guarantee. Skills are third-party content: review a skill before you let an agent use it.

Attribution and licence

Created by

NVIDIA Corporation

Publisher

NVIDIA (Official)

Source

NVIDIA/skills · skills/tao-train-centerpose · commit fce3cca

Licence

Apache License 2.0 (redistribution allowed)

LICENSE (added by AutoGineer from LICENSE-APACHE)
ATTRIBUTION.md (written by AutoGineer)

Synced

2026-11-08

AutoGineer is not affiliated with or endorsed by NVIDIA Corporation or NVIDIA.

Download ZIP

The ZIP holds the skill’s files, its licence files and an ATTRIBUTION.md. Download tao-train-centerpose.zip