Tao Train Action Recognition

NVIDIAOfficialCreated by NVIDIA Corporation

Skill
Tool integration
Data and AI
Apache-2.0
Medium risk

From the author

Action recognition from video sequences. Supports RGB, optical flow, and joint (multi-stream) input types for classifying temporal actions in video clips. Use when training, evaluating, exporting, or running inference on a TAO action-recognition model. Trigger phrases include "train action recognition", "video action classification", "RGB + optical flow action model", "TAO ActionRecognition".

— the description by NVIDIA Corporation

AutoGineer summary

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

The tao-train-action-recognition skill provides a structured interface for managing the lifecycle of action recognition models within the TAO Toolkit ecosystem. It supports training, evaluation, inference, and model export for video-based temporal action classification. The skill handles various input types, including RGB and optical flow, and integrates with AutoML workflows to optimize training parameters. It requires Docker and the NVIDIA container toolkit for execution, relying on specific configuration files to manage datasets, model backbones, and training hyperparameters.

What it does

Trains action recognition models using RGB, optical flow, or joint multi-stream inputs.
Evaluates model performance using metrics like accuracy and mean accuracy.
Exports trained models to ONNX format for deployment.
Runs inference on video clips to classify temporal actions.
Supports AutoML-enabled training workflows for parameter optimization.
Manages model checkpoints and configuration overrides for consistent training and deployment.

When to use it

When you need to classify temporal actions within video sequences.
When working with TAO Toolkit-based computer vision pipelines.
When you need to optimize model training parameters using AutoML.
When you need to export a trained action recognition model for production deployment.

Advantages

Supports multiple input modalities including RGB and optical flow.
Automated training workflows via AutoML integration.
Standardized configuration through YAML spec files.
Docker-native execution ensures environment consistency.

Usage examples

“Train an action recognition model using the provided training and validation datasets.”: The agent executes the training command within the TAO container using the specified train.yaml configuration, producing a trained model checkpoint.
“Evaluate my trained action recognition model on the test dataset.”: The agent runs the evaluation command, processing the test dataset and reporting accuracy and mean accuracy metrics.
“Export my action recognition model to ONNX format.”: The agent triggers the export process, converting the specified checkpoint into an ONNX file for deployment.
“Run inference on a video clip to classify actions.”: The agent uses the inference command and the provided model checkpoint to classify actions within the target video dataset.

Install

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

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

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

Documentation by NVIDIA Corporation

Action Recognition

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).

Action recognition from video sequences. Supports RGB, optical flow, and joint (multi-stream) input types for classifying temporal actions in video clips.

Set model.pretrained_model_path for pretrained backbone weights.

Quick Start (docker run)

Docker-native launch — no TAO SDK and no Python on the host. Use the local Docker/platform skill instead when it gives a stricter environment-specific command (non-root UID mapping, cache redirects, remote daemons).

TAO_PYT_IMAGE_DEFAULT=nvcr.io/nvidia/tao/tao-toolkit:7.3.0-pyt  # versions-key: images.tao_toolkit.pyt
TAO_PYT_IMAGE="${TAO_PYT_IMAGE:-$TAO_PYT_IMAGE_DEFAULT}"
RUN_ROOT="${RUN_ROOT:-$PWD}"
DOCKER_COMMON=(
  --rm --gpus all --shm-size=8g
  --shm-size=8g
  --ulimit memlock=-1
  --ulimit stack=67108864
  -v "$RUN_ROOT/data:/data:ro"
  -v "$RUN_ROOT/specs:/specs:ro"
  -v "$RUN_ROOT/results:/results"
)

Train:

docker run "${DOCKER_COMMON[@]}" "$TAO_PYT_IMAGE" \
  action_recognition train -e /specs/train.yaml

Evaluate:

docker run "${DOCKER_COMMON[@]}" "$TAO_PYT_IMAGE" \
  action_recognition evaluate -e /specs/evaluate.yaml

Inference:

docker run "${DOCKER_COMMON[@]}" "$TAO_PYT_IMAGE" \
  action_recognition inference -e /specs/inference.yaml

Export:

docker run "${DOCKER_COMMON[@]}" "$TAO_PYT_IMAGE" \
  action_recognition export -e /specs/export.yaml

Every action takes its spec with -e; results_dir is set in the spec or overridden on the command line. Mount any pretrained-weights directory the spec references, and keep every in-container path consistent across actions.

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: action_recognition
Formats: default
Training monitoring metrics: val_loss, val_acc
Evaluate task metrics: accuracy, m_accuracy
AutoML metric contract: for the required evaluation-backed baseline and final comparison, use accuracy with maximize direction and run evaluate through eval_fn for every recommendation. val_loss is suitable only for an explicitly accepted training-proxy run because evaluate does not emit it. Scratch runs with no starting checkpoint require a minimal default train job followed by evaluation of its exact epoch/step checkpoint for the baseline.

Per-Action Dataset Requirements

Action	Spec Key	Source	Files	List?
evaluate	evaluate.test_dataset_dir	train_datasets	test/ extracted from test.tar.gz	No
inference	inference.inference_dataset_dir	train_datasets	test/smile/ extracted from test/smile.tar.gz	No
train	dataset.train_dataset_dir	train_datasets	train/ extracted from train.tar.gz	No
train	dataset.val_dataset_dir	train_datasets	test/ extracted from test.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.

LOCAL_DATA = "/workspace/data/extracted"

If the source dataset is provided as the TAO sample archives train.tar.gz, test.tar.gz, or test/smile.tar.gz, download and extract them before launching the TAO container. The action-recognition entrypoints expect directory paths and fail with NotADirectoryError when these spec keys point at .tar.gz files.

train (mandatory data sources):

{
    "train.num_epochs": 30,
    "train.checkpoint_interval": 10,
    "train.validation_interval": 10,
    "train.num_gpus": 1,
    "dataset.label_map": {
        "catch": 0,
        "smile": 1
    },
    "dataset.batch_size": 2,
    "dataset.train_dataset_dir": f"{LOCAL_DATA}/train",
    "dataset.val_dataset_dir": f"{LOCAL_DATA}/test",
}

evaluate (mandatory data sources):

{
    "dataset.label_map": {
        "catch": 0,
        "smile": 1
    },
    "evaluate.test_dataset_dir": f"{LOCAL_DATA}/test",
}

inference (mandatory data sources):

{
    "dataset.label_map": {
        "catch": 0,
        "smile": 1
    },
    "inference.inference_dataset_dir": f"{LOCAL_DATA}/smile_infer/smile",
}

export (mandatory checkpoint + output path):

{
    "export.checkpoint": "<selected train checkpoint>",
    "export.onnx_file": "<results_dir>/action_recognition.onnx",
}

For direct local-docker chaining without the SDK resolver, select the concrete checkpoint produced by training, for example model_epoch_000_step_00005.pth, and pass that exact file to evaluate, inference, and export. Do not use the ar_model_latest.pth symlink unless the user explicitly requests latest-checkpoint behavior. For resume training, set train.resume_training_checkpoint_path to the exact epoch/step checkpoint being resumed.

Eval Dataset

Optional. Test dataset may be distributed as test.tar.gz separate from training; extract it and point the spec to the extracted test/ directory. TAO training emits val_loss and val_acc for the packaged sample data, while the evaluate action emits accuracy and m_accuracy. Use accuracy with maximize direction for the normal evaluation-backed AutoML workflow. Use val_loss with minimize direction only when the user explicitly accepts a training-only proxy without the required impact baseline.

Important Parameters

model.model_type: Input type: rgb, of (optical flow), or joint (multi-stream).
model.backbone: Default resnet_18. Used as the spatial feature extractor.
dataset.label_map: Dictionary mapping class names to indices.
model.rgb_seq_length: Number of frames per clip for RGB input.
model.of_seq_length: Number of frames for optical flow input.
train.optim.lr: Learning rate. Default 5e-4.

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 best strategy automatically)
No explicit num_nodes or distributed_strategy config — single-node oriented

Hardware

Minimum 1 GPU(s), recommended 2 GPU(s). 16GB+ VRAM per GPU. Memory depends on sequence length and input resolution. batch_size=2 is conservative for video data.

Error Patterns

Sequence length mismatch: Ensure video clips have enough frames for the configured rgb_seq_length or of_seq_length.

Evaluate/inference missing label map: Downstream actions rebuild the ActionRecognitionModel before loading the checkpoint, so they need the same dataset.label_map used during training. Include it with every evaluate or inference spec; otherwise model construction fails before the checkpoint can be validated.

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 action_recognition.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	`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
inference	`encryption_key`	`key`	encryption key
inference	`inference.checkpoint`	`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.of_pretrained_model_path`	`ptm_if_no_resume_model`	PTM when no resume checkpoint exists
train	`model.rgb_pretrained_model_path`	`ptm_if_no_resume_model`	PTM when no resume checkpoint exists
train	`results_dir`	`output_dir`	current job results directory
train	`train.resume_training_checkpoint_path`	`resume_model`	model file inferred from the current job results folder

For parent_model 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.

Files

ATTRIBUTION.mdAdded by AutoGineer954 B
BENCHMARK.md7.0 KiB
LICENSELicence file added by AutoGineer11.2 KiB
SKILL.md11.6 KiB
config/skillspector-baseline.yaml1.0 KiB
evals/evals.json816 B
references/skill_info.yaml2.7 KiB
references/spec_template_evaluate.yaml1.8 KiB
references/spec_template_export.yaml1.7 KiB
references/spec_template_inference.yaml1.8 KiB
references/spec_template_train.yaml1.6 KiB
schemas/evaluate.schema.json26.5 KiB
schemas/export.schema.json24.4 KiB
schemas/inference.schema.json26.5 KiB
schemas/manifest.json6.1 KiB
schemas/train.schema.json23.2 KiB
skill-card.md4.9 KiB
skill.oms.sig7.1 KiB

schemas/evaluate.schema.json

{
  "automl_default_parameters": [
    "model.dropout_ratio",
    "train.optim.lr"
  ],
  "automl_disabled_parameters": [
    "dataset.augmentation_config.of_input_mean",
    "train.cudnn",
    "train.gpu_ids",
    "wandb.tags",
    "dataset.augmentation_config",
    "evaluate",
    "inference",
    "train",
    "dataset.augmentation_config.of_input_std",
    "dataset.augmentation_config.scales",
    "dataset",
    "dataset.label_map",
    "model",
    "train.optim.lr_steps",
    "dataset.augmentation_config.rgb_input_mean",
    "evaluate.gpu_ids",
    "train.optim",
    "dataset.augmentation_config.rgb_input_std",
    "export",
    "wandb",
    "inference.gpu_ids"
  ],
  "default": {
    "dataset": {
      "augmentation_config": {
        "crop_smaller_edge": 256,
        "horizontal_flip_prob": 0.5,
        "of_input_mean": [
          0.5
        ],
        "of_input_std": [
          0.5
        ],
        "rgb_input_mean": [
          0.485,
          0.456,
          0.406
        ],
        "rgb_input_std": [
          0.229,
          0.224,
          0.225
        ],
        "scales": [
          1
        ],
        "train_crop_type": "random_crop",
        "val_center_crop": false
      },
      "batch_size": 2,
      "clips_per_video": 1,
      "train_dataset_dir": "",
      "val_dataset_dir": "",
      "workers": 8
    },
    "encryption_key": "",
    "evaluate": {
      "batch_size": 1,
      "checkpoint": "???",
      "gpu_ids": [
        0
      ],
      "num_gpus": 1,
      "num_nodes": 1,
      "results_dir": "",
      "test_dataset_dir": "???",
      "trt_engine": "",
      "video_eval_mode": "center",
      "video_num_segments": 10
    },
    "model": {
      "backbone": "resnet_18",
      "dropout_ratio": 0.5,
      "imagenet_pretrained": false,
      "input_height": 224,
      "input_type": "2d",
      "input_width": 224,
      "joint_pretrained_model_path": "",
      "model_type": "rgb",
      "num_fc": 64,
      "of_pretrained_model_path": "",
      "of_pretrained_num_classes": 5,
      "of_seq_length": 10,
      "rgb_pretrained_model_path": "",
      "rgb_pretrained_num_classes": 5,
      "rgb_seq_length": 3,
      "sample_rate": 1,
      "sample_strategy": "random_interval"
    },
    "model_name": "",
    "results_dir": "",
    "train": {
      "checkpoint_interval": 1,
      "checkpoint_interval_unit": "epoch",
      "clip_grad_norm": 0.0,
      "cudnn": {
        "benchmark": false,
        "deterministic": true
      },
      "gpu_ids": [
        0
      ],
      "num_epochs": 10,
      "num_gpus": 1,
      "num_nodes": 1,
      "optim": {
        "lr": 0.0005,
        "lr_decay": 0.1,
        "lr_monitor": "val_loss",
        "lr_scheduler": "MultiStep",
        "lr_steps": [
          15,
          25
        ],
        "min_lr": 0.0001,
        "momentum": 0.9,
        "patience": 1,
        "weight_decay": 0.0005
      },
      "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
    },
    "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",
      "model",
      "dataset",
      "train",
      "evaluate",
      "export",
      "inference"
    ],
    "dataset": {
      "automl_disabled_parameters": [
        "dataset.label_map",
        "dataset.augmentation_config"
      ],
      "automl_enabled": false,
      "default": {
        "augmentation_config": {
          "crop_smaller_edge": 256,
          "horizontal_flip_prob": 0.5,
          "of_input_mean": [
            0.5
          ],
          "of_input_std": [
            0.5
          ],
          "rgb_input_mean": [
            0.485,
            0.456,
            0.406
          ],
          "rgb_input_std": [
            0.229,
            0.224,
            0.225
          ],
          "scales": [
            1
          ],
          "train_crop_type": "random_crop",
          "val_center_crop": false
        },
        "batch_size": 2,
        "clips_per_video": 1,
        "train_dataset_dir": "",
        "val_dataset_dir": "",
        "workers": 8
      },
      "description": "Configurable parameters for the dataset.",
      "properties": {
        "augmentation_config": {
          "automl_disabled_parameters": [
            "dataset.augmentation_config.scales",
            "dataset.augmentation_config.rgb_input_mean",
            "dataset.augmentation_config.rgb_input_std",
            "dataset.augmentation_config.of_input_mean",
            "dataset.augmentation_config.of_input_std"
          ],
          "automl_enabled": false,
          "default": {
            "crop_smaller_edge": 256,
            "horizontal_flip_prob": 0.5,
            "of_input_mean": [
              0.5
            ],
            "of_input_std": [
              0.5
            ],
            "rgb_input_mean": [
              0.485,
              0.456,
              0.406
            ],
            "rgb_input_std": [
              0.229,
              0.224,
              0.225
            ],
            "scales": [
              1
            ],
            "train_crop_type": "random_crop",
            "val_center_crop": false
          },
          "description": "Configurable parameters for dataset augmentation.",
          "properties": {
            "crop_smaller_edge": {
              "default": 256,
              "description": "Smaller edge length of the center crop in validation.",
              "minimum": 1,
              "type": "int"
            },
            "horizontal_flip_prob": {
              "default": 0.5,
              "description": "Probability to apply horizontal flip to images.",
              "maximum": 1.0,
              "minimum": 0.0,
              "type": "float"
            },
            "of_input_mean": {
              "automl_enabled": false,
              "default": [
                0.5
              ],
              "description": "Mean value per channel to be substructed for optical flow input.",
              "type": "list"
            },
            "of_input_std": {
              "automl_enabled": false,
              "default": [
                0.5
              ],
              "description": "Std value to be divided for optical flow input.",
              "type": "list"
            },
            "rgb_input_mean": {
              "automl_enabled": false,
              "default": [
                0.485,
                0.456,
                0.406
              ],
              "description": "Mean value per channel to be substructed for RGB input.",
              "type": "list"
            },
            "rgb_input_std": {
              "automl_enabled": false,
              "default": [
                0.229,
                0.224,
                0.225
              ],
              "description": "Std value to be divided for RGB input.",
              "type": "list"
            },
            "scales": {
              "automl_enabled": false,
              "default": [
                1
              ],
              "description": "Scales list for multi_scale_crop.",
              "type": "list"
            },
            "train_crop_type": {
              "default": "random_crop",
              "description": "Crop type to crop image patches from the original input image.",
              "enum": [
                "random_crop",
                "multi_scale_crop",
                "no_crop"
              ],
              "type": "categorical"
            },
            "val_center_crop": {
              "default": false,
              "description": "Bool flag to apply center crop in validation.",
              "type": "bool"
            }
          },
          "type": "collection"
        },
        "batch_size": {
          "default": 2,
          "description": "Batch size of model input.",
          "minimum": 1,
          "type": "int"
        },
        "clips_per_video": {
          "default": 1,
          "description": "Number of clips sampled from single video.",
          "minimum": 1,
          "type": "int"
        },
        "label_map": {
          "automl_enabled": false,
          "description": "Dict mapping the class to class index",
          "type": "collection"
        },
        "train_dataset_dir": {
          "default": "",
          "description": "Absolute path to train dataset.",
          "type": "string"
        },
        "val_dataset_dir": {
          "default": "",
          "description": "Absolute path to validation dataset.",
          "type": "string"
        },
        "workers": {
          "default": 8,
          "description": "Number of workers to process data.",
          "minimum": 0,
          "type": "int"
        }
      },
      "type": "collection"
    },
    "encryption_key": {
      "default": "",
      "description": "Key for encrypting model checkpoints",
      "title": "Encryption key",
      "type": "string"
    },
    "evaluate": {
      "automl_disabled_parameters": [
        "evaluate.gpu_ids"
      ],
      "automl_enabled": false,
      "default": {
        "batch_size": 1,
        "checkpoint": "???",
        "gpu_ids": [
          0
        ],
        "num_gpus": 1,
        "num_nodes": 1,
        "results_dir": "",
        "test_dataset_dir": "???",
        "trt_engine": "",
        "video_eval_mode": "center",
        "video_num_segments": 10
      },
      "description": "Configurable parameters for an evaluation experiment.",
      "popular": [
        "num_gpus",
        "num_nodes",
        "gpu_ids"
      ],
      "properties": {
        "batch_size": {
          "default": 1,
          "description": "Batch size of data for evaluation.",
          "minimum": 1,
          "type": "int"
        },
        "checkpoint": {
          "default": "???",
          "description": "Path to the checkpoint used for evaluation.",
          "title": "Checkpoint path",
          "type": "string"
        },
        "gpu_ids": {
          "automl_enabled": false,
          "default": [
            0
          ],
          "description": "\n        List of GPU IDs to run the evaluation on. The length of this list\n        must be equal to the number of gpus in evaluate.num_gpus.",
          "popular": true,
          "title": "GPU IDs",
          "type": "list"
        },
        "num_gpus": {
          "default": 1,
          "description": "The number of GPUs to run the evaluation job.",
          "minimum": 1,
          "popular": true,
          "title": "Number of GPUs",
          "type": "int"
        },
        "num_nodes": {
          "default": 1,
          "description": "Number of nodes to run the evaluation on. If > 1, then multi-node is enabled.",
          "minimum": 1,
          "popular": true,
          "title": "Number of nodes",
          "type": "int"
        },
        "results_dir": {
          "default": "",
          "description": "\n        Path to where all the assets generated from a task are stored.\n        ",
          "title": "Results directory",
          "type": "string"
        },
        "test_dataset_dir": {
          "default": "???",
          "description": "The number of clips to do inference for single video.",
          "type": "string"
        },
        "trt_engine": {
          "default": "",
          "description": "Path to the TensorRT engine to be used for evaluation.\n                    This only works with :code:`tao-deploy`.",
          "title": "TensorRT Engine",
          "type": "string"
        },
        "video_eval_mode": {
          "default": "center",
          "description": "The video sampling mode for evaluation.",
          "enum": [
            "center",
            "conv",
            "all"
          ],
          "type": "categorical"
        },
        "video_num_segments": {
          "default": 10,
          "description": "The number of clips to do inference for single video.",
          "type": "int"
        }
      },
      "type": "collection"
    },
    "model": {
      "automl_default_parameters": [
        "model.dropout_ratio"
      ],
      "automl_enabled": false,
      "default": {
        "backbone": "resnet_18",
        "dropout_ratio": 0.5,
        "imagenet_pretrained": false,
        "input_height": 224,
        "input_type": "2d",
        "input_width": 224,
        "joint_pretrained_model_path": "",
        "model_type": "rgb",
        "num_fc": 64,
        "of_pretrained_model_path": "",
        "of_pretrained_num_classes": 5,
        "of_seq_length": 10,
        "rgb_pretrained_model_path": "",
        "rgb_pretrained_num_classes": 5,
        "rgb_seq_length": 3,
        "sample_rate": 1,
        "sample_strategy": "random_interval"
      },
      "description": "Configurable parameters for the model.",
      "properties": {
        "backbone": {
          "default": "resnet_18",
          "description": "The backbone of model architecture.",
          "enum": [
            "resnet_18",
            "resnet_34",
            "resnet_50",
            "resnet_101",
            "resnet_152",
            "i3d"
          ],
          "type": "categorical"
        },
        "dropout_ratio": {
          "automl_enabled": true,
          "default": 0.5,
          "description": "The dropout ratio for the model.",
          "maximum": 1.0,
          "minimum": 0.0,
          "type": "float"
        },
        "imagenet_pretrained": {
          "default": false,
          "description": "The bool flag to load imagenet pretrained weights.",
          "type": "bool"
        },
        "input_height": {
          "default": 224,
          "description": "The input height of the model.",
          "maximum": Infinity,
          "minimum": 1,
          "type": "int"
        },
        "input_type": {
          "default": "2d",
          "description": "The type of model input: [2d, 3d].",
          "enum": [
            "2d",
            "3d"
          ],
          "type": "categorical"
        },
        "input_width": {
          "default": 224,
          "description": "The input width of the model.",
          "maximum": Infinity,
          "minimum": 1,
          "type": "int"
        },
        "joint_pretrained_model_path": {
          "default": "",
          "description": "The pretrained weights for joint pretrained model.",
          "type": "string"
        },
        "model_type": {
          "default": "rgb",
          "description": "The type of model architecture: [rgb, of, joint].",
          "enum": [
            "rgb",
            "of",
            "joint"
          ],
          "type": "categorical"
        },
        "num_fc": {
          "default": 64,
          "description": "The number of hidden units in fully-connected layer.",
          "maximum": Infinity,
          "minimum": 1,
          "type": "int"
        },
        "of_pretrained_model_path": {
          "default": "",
          "description": "The pretrained weights for optical flow model.",
          "type": "string"
        },
        "of_pretrained_num_classes": {
          "default": 5,
          "description": "The classes number of the pretrained weights for optical flow model.",
          "maximum": Infinity,
          "minimum": 0,
          "type": "int"
        },
        "of_seq_length": {
          "default": 10,
          "description": "The optical flow sequence length.",
          "maximum": Infinity,
          "minimum": 1,
          "type": "int"
        },
        "rgb_pretrained_model_path": {
          "default": "",
          "description": "The pretrained weights for RGB model.",
          "type": "string"
        },
        "rgb_pretrained_num_classes": {
          "default": 5,
          "description": "The classes number of the pretrained weights for RGB model.",
          "maximum": Infinity,
          "minimum": 0,
          "type": "int"
        },
        "rgb_seq_length": {
          "default": 3,
          "description": "The RGB sequence length.",
          "maximum": Infinity,
          "minimum": 1,
          "type": "int"
        },
        "sample_rate": {
          "default": 1,
          "description": "The sample rate to sample frames from videos.",
          "type": "int"
        },
        "sample_strategy": {
          "default": "random_interval",
          "description": "The sample strategy to sample frames from videos.",
          "enum": [
            "random_interval",
            "consecutive"
          ],
          "type": "categorical"
        }
      },
      "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"
      ],
      "automl_enabled": false,
      "default": {
        "checkpoint_interval": 1,
        "checkpoint_interval_unit": "epoch",
        "clip_grad_norm": 0.0,
        "cudnn": {
          "benchmark": false,
          "deterministic": true
        },
        "gpu_ids": [
          0
        ],
        "num_epochs": 10,
        "num_gpus": 1,
        "num_nodes": 1,
        "optim": {
          "lr": 0.0005,
          "lr_decay": 0.1,
          "lr_monitor": "val_loss",
          "lr_scheduler": "MultiStep",
          "lr_steps": [
            15,
            25
          ],
          "min_lr": 0.0001,
          "momentum": 0.9,
          "patience": 1,
          "weight_decay": 0.0005
        },
        "results_dir": "",
        "resume_training_checkpoint_path": "",
        "seed": 1234,
        "validation_interval": 1
      },
      "description": "Configurable parameters for a train experiment.",
      "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_norm": {
          "default": 0.0,
          "description": "The L2 magnitude of graident to be clipped in the training.",
          "minimum": 0.0,
          "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"
        },
        "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"
          ],
          "automl_disabled_parameters": [
            "train.optim.lr_steps"
          ],
          "automl_enabled": false,
          "default": {
            "lr": 0.0005,
            "lr_decay": 0.1,
            "lr_monitor": "val_loss",
            "lr_scheduler": "MultiStep",
            "lr_steps": [
              15,
              25
            ],
            "min_lr": 0.0001,
            "momentum": 0.9,
            "patience": 1,
            "weight_decay": 0.0005
          },
          "description": "Configurable parameters for optimizer.",
          "properties": {
            "lr": {
              "automl_enabled": true,
              "default": 0.0005,
              "description": "Learning rate for training.",
              "maximum": Infinity,
              "minimum": 0.0,
              "type": "float"
            },
            "lr_decay": {
              "default": 0.1,
              "description": "Learning rate decay factor in learning rate scheduler.",
              "maximum": 1.0,
              "minimum": 0.0,
              "type": "float"
            },
            "lr_monitor": {
              "default": "val_loss",
              "description": "Learning rate monitor for AutoReduce learning rate scheduler.",
              "enum": [
                "val_loss",
                "train_loss"
              ],
              "type": "categorical"
            },
            "lr_scheduler": {
              "default": "MultiStep",
              "description": "Learning rate scheduler.",
              "enum": [
                "MultiStep",
                "AutoReduce"
              ],
              "type": "categorical"
            },
            "lr_steps": {
              "automl_enabled": false,
              "default": [
                15,
                25
              ],
              "description": "Steps to change learning rate in MultiStep scheduler.",
              "type": "list_2"
            },
            "min_lr": {
              "default": 0.0001,
              "description": "Minimum learning rate.",
              "maximum": Infinity,
              "minimum": 0.0,
              "type": "float"
            },
            "momentum": {
              "default": 0.9,
              "description": "Momentum coefficient for SGD.",
              "maximum": 1.0,
              "minimum": 0.0,
              "type": "float"
            },
            "patience": {
              "default": 1,
              "description": "Number of epochs for AutoReduce learning rate scheduler tolerance.",
              "type": "int"
            },
            "weight_decay": {
              "default": 0.0005,
              "description": "Weight decay coefficient for trainng.",
              "maximum": Infinity,
              "minimum": 0.0,
              "type": "float"
            }
          },
          "type": "collection"
        },
        "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": "evaluate",
    "core_module": "action_recognition",
    "model": "action-recognition",
    "network_arch": "action_recognition",
    "schema_action": "evaluate",
    "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 a command-line tool, 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-action-recognition · 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-action-recognition.zip