Now liveThe Skillselion MCP - thousands of ranked skills, loaded into your agent mid-task. No install.Get it →
nvidia avatar

Tao Train Mask Auto Label

nvidia/skills

  • 1.4k installs
  • 2.8k repo stars
  • Updated August 4, 2026
  • nvidia/skills

tao-train-mask-auto-label is an agent skill that trains, evaluates, and runs inference for NVIDIA TAO MAL mask auto-label models for developers who need pseudo-label segmentation masks from minimal point or box annotatio

About

tao-train-mask-auto-label is an NVIDIA agent skill (version 0.1.0) for MAL (Mask Auto-Label) weakly supervised segmentation using a ViT-MAE backbone inside NVIDIA TAO Toolkit. The skill documents COCO-style annotation JSON with images.tar.gz datasets, mandatory spec overrides for train, evaluate, and inference, and defaults including vit-mae-base/16 architecture, learning_rate 1e-6, and crop_size 512 with mIoU monitoring. Developers reach for tao-train-mask-auto-label when accelerating segmentation dataset creation by generating pseudo-label masks from sparse box or point prompts before human review and downstream Mask2Former or detector training. AutoML HPO searches train.lr and train.wd with conservative Bayesian ranges, and the skill recommends 24GB+ GPU VRAM for ViT-MAE at 512px crops. Inference produces MAL label-dump JSON artifacts via create_inference_result_file_mal mappings.

  • TAO mask auto-label training
  • Unlabeled image ingestion
  • Pseudo-mask generation
  • Label export formats
  • Annotation throughput boost

Tao Train Mask Auto Label by the numbers

  • 1,446 all-time installs (skills.sh)
  • +25 installs in the week ending Aug 5, 2026 (Skillselion tracking)
  • Ranked #196 of 2,064 Data Science & ML skills by installs in the Skillselion catalog
  • Data as of Aug 5, 2026 (Skillselion catalog sync)
npx skills add https://github.com/nvidia/skills --skill tao-train-mask-auto-label

Add your badge

Show developers this skill is listed on Skillselion. Paste this into your README.

Listed on Skillselion
Installs1.4k
repo stars★ 2.8k
Last updatedAugust 4, 2026
Repositorynvidia/skills ↗

How do you auto-label segmentation masks in TAO?

Train NVIDIA TAO mask auto-label models to accelerate segmentation dataset creation by generating pseudo-labels for images before human review and downstream detector training.

Who is it for?

Computer vision teams with sparse box or point annotations who need TAO-guided MAL training to bootstrap dense segmentation masks before full segmenter training.

Skip if: Projects with fully labeled panoptic COCO datasets ready for direct Mask2Former training without a pseudo-labeling preprocessing step.

When should I use this skill?

The user asks to train MAL, generate mask auto-labels, run weakly supervised segmentation, or produce pseudo-labels from minimal annotations in NVIDIA TAO.

What you get

MAL checkpoints, mIoU metrics, and inference label-dump JSON files with pseudo-label segmentation masks for human review.

  • Fine-tuned MAL checkpoints
  • Pseudo-label mask JSON dumps
  • mIoU evaluation metrics

By the numbers

  • Default dataset.crop_size 512 for ViT-MAE MAL training
  • Default train.lr 1e-6 for ViT-MAE fine-tuning
  • Skill metadata version 0.1.0 recommending 24GB+ GPU VRAM

Files

SKILL.mdMarkdownGitHub ↗

MAL

MAL (Mask Auto-Label) for weakly-supervised segmentation. Produces segmentation masks from minimal annotations (e.g., point or box annotations). Uses ViT-MAE backbone.

Set train.pretrained_model_path for ViT-MAE pretrained weights.

Dataclass Schemas

Generated TAO Core schemas are packaged in schemas/<action>.schema.json, with schemas/manifest.json listing available actions. Each generated schema also emits references/spec_template_<action>.yaml from the schema top-level default field. AutoML enablement is declared at the model layer in references/skill_info.yaml via automl_enabled. Runnable AutoML still requires schemas/train.schema.json and references/spec_template_train.yaml to exist and parse. Use the packaged train schema for automl_default_parameters, automl_disabled_parameters, defaults, min/max bounds, enums, option weights, math conditions, dependencies, and popular parameters. Do not expect ~/tao-core at runtime; maintainers regenerate schemas/templates before packaging the skill bank.

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: segmentation
  • Formats: default
  • Monitoring metric: mIoU

Per-Action Dataset Requirements

ActionSpec KeySourceFilesList?
evaluatedataset.val_img_direval_datasetimages.tar.gzNo
evaluatedataset.val_ann_patheval_datasetannotations.jsonNo
inferenceinference.img_dirinference_datasetimages.tar.gzNo
inferenceinference.ann_pathinference_datasetannotations.jsonNo
traindataset.train_img_dirtrain_datasetsimages.tar.gzNo
traindataset.train_ann_pathtrain_datasetsannotations.jsonNo
traindataset.val_img_direval_datasetimages.tar.gzNo
traindataset.val_ann_patheval_datasetannotations.jsonNo

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. MAL expects COCO-style annotation JSON plus image paths that match the JSON file_name entries after the data source is prepared. Archive-only CSV/image datasets are not compatible unless they are converted to this format first.

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

train (mandatory data sources):

{
    "train.num_gpus": 1,
    "train.gpu_ids": [
        0
    ],
    "train.num_epochs": 5,
    "train.checkpoint_interval": 5,
    "train.validation_interval": 5,
    "dataset.train_img_dir": f"{S3_TRAIN}/images.tar.gz",
    "dataset.train_ann_path": f"{S3_TRAIN}/annotations.json",
    "dataset.val_img_dir": f"{S3_EVAL}/images.tar.gz",
    "dataset.val_ann_path": f"{S3_EVAL}/annotations.json",
}

evaluate (mandatory data sources):

{
    "evaluate.checkpoint": "<selected train/AutoML checkpoint>",
    "dataset.val_img_dir": f"{S3_EVAL}/images.tar.gz",
    "dataset.val_ann_path": f"{S3_EVAL}/annotations.json",
}

inference (mandatory data sources):

{
    "inference.checkpoint": "<selected train/AutoML checkpoint>",
    "inference.img_dir": f"{S3_EVAL}/images.tar.gz",
    "inference.ann_path": f"{S3_EVAL}/annotations.json",
}

For checkpoint-dependent actions, use the model resolver declared in references/skill_info.yaml. Select the exact epoch/step checkpoint requested by the user or the best checkpoint when a best-checkpoint action is requested. The mal_model_latest.pth symlink is only appropriate when the user explicitly asks for the latest checkpoint.

Eval Dataset

Optional. Val images and annotations configured alongside train paths.

Important Parameters

  • model.arch: ViT-MAE backbone variant. Default vit-mae-base/16.

Avoid vit-deit-tiny/16; the current runtime rejects tiny ViT variants.

  • train.lr: Learning rate. Default 1e-6 (very low — fine-tuning ViT).
  • dataset.crop_size: Training crop size. Default 512. Use this key, not

model.crop_size.

  • train.warmup_epochs: Warmup epochs before full learning rate.
  • model.load_mask: Whether to load pre-computed masks.

AutoML / HPO Notes

For MAL AutoML launches, keep the default smoke search space narrow and pass automl_hyperparameters=["train.lr", "train.wd"]. Use conservative Bayesian ranges around the ViT-MAE fine-tuning defaults, for example train.lr from 1e-7 to 1e-5 and train.wd from 1e-5 to 1e-2. The packaged train schema marks these two parameters as the default AutoML parameters; pass them explicitly when using a runtime that still derives MAL search metadata from its bundled config module.

Multi-GPU / Multi-Node

Launch method: Lightning-managed (single python process, Lightning spawns workers).

Spec KeyDescriptionDefault
train.num_gpusNumber of GPUs1
train.gpu_idsGPU device indices[0]
train.num_nodesNumber of nodes1
  • Multi-GPU strategy: ddp_find_unused_parameters_true
  • No fsdp support
  • LR auto-scaling: lr = lr * num_devices * batch_size (learning rate is scaled automatically by device count and batch size)

Multi-node env vars (set by orchestrator): WORLD_SIZE, NODE_RANK, MASTER_ADDR, MASTER_PORT, NUM_GPU_PER_NODE.

Hardware

Minimum 1 GPU(s), recommended 2 GPU(s). 24GB+ (A100 recommended) VRAM per GPU. ViT-MAE backbone at crop_size=512 needs 24GB+ GPU memory.

Error Patterns

CUDA out of memory: Reduce dataset.crop_size (512 -> 384 -> 256) or use a smaller ViT-MAE variant (base vs large).

Key `crop_size` not in `MALModelConfig`: The crop-size override was placed under model.crop_size. Move it to dataset.crop_size.

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 mal.config.json:

ActionSpec FieldInference FunctionMeaning
evaluateevaluate.checkpointparent_modelmodel file inferred from the parent job results folder
evaluateresults_diroutput_dircurrent job results directory
inferenceinference.checkpointparent_modelmodel file inferred from the parent job results folder
inferenceinference.label_dump_pathcreate_inference_result_file_malMAL inference JSON path
inferenceresults_diroutput_dircurrent job results directory
traintrain.pretrained_model_pathptm_if_no_resume_modeloptional pretrained model when not resuming
traintrain.resume_training_checkpoint_pathresume_modelexact checkpoint for resume runs
trainresults_diroutput_dircurrent job results directory

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.

Related skills

microsoftMicrosoft FoundryDeploy, evaluate, and continuously improve Microsoft Foundry agents from a single agent interface.490k1.3klllllllamaAi Research ReproductionOrchestrate trustworthy, auditable reproduction of deep learning repositories directly from their READMEs.176k511lllllllamaRun TrainSafely execute selected deep learning training commands with standardized evidence capture.176k511lllllllamaExplore RunSafely run isolated exploratory experiments with clear recording and conservative selection before committing changes.176k511lllllllamaPaper Context ResolverFetch precise reproduction-critical details like dataset splits, preprocessing steps, or evaluation protocols from the original academic paper when the repo README leav141k511lllllllamaRepo Intake And PlanScan unfamiliar AI research repositories and receive a minimal, trustworthy reproduction target before investing significant time.140k511

How it compares

Use tao-train-mask-auto-label to bootstrap masks from sparse prompts; use tao-train-mask2former when full panoptic ground truth is already available for universal segmentation training.

FAQ

What annotations does tao-train-mask-auto-label accept?

tao-train-mask-auto-label expects COCO-style annotation JSON plus image paths matching file_name entries after data preparation. Archive-only CSV datasets are incompatible unless converted to this format with images.tar.gz and annotations.json train and validation splits.

What ViT-MAE settings does MAL training use?

tao-train-mask-auto-label defaults to vit-mae-base/16 architecture with train.lr 1e-6 and dataset.crop_size 512. The skill warns against vit-deit-tiny/16 because the current runtime rejects tiny ViT variants, and crop_size overrides belong under dataset not model.

Does MAL support AutoML hyperparameter tuning?

tao-train-mask-auto-label is AutoML-enabled and recommends searching train.lr from 1e-7 to 1e-5 and train.wd from 1e-5 to 1e-2. The packaged train schema marks these two parameters as default AutoML hyperparameters for ViT-MAE fine-tuning runs.

Data Science & MLpipelinesetlanalytics
Browse by category
AI & AgentsFrontend & DesignBackend & DataDev ToolsAutomationTesting & ReviewSecurityMonitoring & CloudDocs & Planning
Design & frontend
Design systemsAnimation & motionAccessibility (a11y)CSS & TailwindUI designUXDesign to codeshadcn/uiLanding pagesBest design skills
Most searched skills
find-skillscavemanfrontend-designbrainstormingsystematic-debuggingstop-slopemil-design-engui-ux-pro-maxkarpathy-guidelinesweb-design-guidelinesreact-best-practicesgrill-meimprovemicrosoft-foundryagent-browser
Browse all listings
All agent skillsAll MCP serversAll marketplacesAll plugins

This week in AI coding

Five minutes, every Monday - the tools, releases and tactics for developers.

unsubscribe anytime.