[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill-nvidia-tao-train-pose-classification":3,"mdc--7qcud-key":31,"related-repo-nvidia-tao-train-pose-classification":2396,"related-org-nvidia-tao-train-pose-classification":2502},{"slug":4,"name":4,"fn":5,"description":6,"org":7,"tags":11,"stars":20,"repoUrl":21,"updatedAt":22,"license":23,"forks":24,"topics":25,"repo":26,"sourceUrl":29,"mdContent":30},"tao-train-pose-classification","train ST-GCN models for pose classification","Pose classification using ST-GCN (Spatial Temporal Graph Convolutional Network). Classifies skeleton sequences into action categories from pose-keypoint data. Use when training, evaluating, exporting, or running inference for a TAO pose-classification model. Trigger phrases include \"train pose classification\", \"skeleton action recognition\", \"ST-GCN\", \"keypoint sequence classifier\".",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},"nvidia","NVIDIA","https:\u002F\u002Fpexgzepcugksgbtrxkhf.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Forg-logos\u002Fnvidia.png",[12,16,17],{"name":13,"slug":14,"type":15},"Machine Learning","machine-learning","tag",{"name":9,"slug":8,"type":15},{"name":18,"slug":19,"type":15},"Computer Vision","computer-vision",2473,"https:\u002F\u002Fgithub.com\u002FNVIDIA\u002Fskills","2026-07-14T05:30:29.18188","Apache-2.0",281,[],{"repoUrl":21,"stars":20,"forks":24,"topics":27,"description":28},[],"AI agent skills published by NVIDIA","https:\u002F\u002Fgithub.com\u002FNVIDIA\u002Fskills\u002Ftree\u002FHEAD\u002Fskills\u002Ftao-train-pose-classification","---\nname: tao-train-pose-classification\ndescription: Pose classification using ST-GCN (Spatial Temporal Graph Convolutional Network). Classifies skeleton sequences\n  into action categories from pose-keypoint data. Use when training, evaluating, exporting, or running inference for a TAO\n  pose-classification model. Trigger phrases include \"train pose classification\", \"skeleton action recognition\", \"ST-GCN\",\n  \"keypoint sequence classifier\".\nlicense: Apache-2.0\ncompatibility: Requires docker + nvidia-container-toolkit.\nmetadata:\n  version: \"0.1.0\"\n  author: NVIDIA Corporation\nallowed-tools: Read Bash\ntags:\n- pose\n- classification\n---\n\n# Pose Classification\n\nPose classification using ST-GCN (Spatial Temporal Graph Convolutional Network). Classifies skeleton sequences into action categories from pose keypoint data.\n\nTypically trained from scratch on skeleton data.\n\nThe packaged PyTorch Pose Classification CLI supports `dataset_convert`, `train`, `evaluate`, `export`, and `inference`. `dataset_convert` is conditional: run it only when the input is raw DeepStream BodyPose JSON. If the dataset is already converted to TAO-ready `.npy` \u002F `.pkl` files, start directly with `train` on those files and mark dataset conversion as `not run: preconverted dataset provided` in validation reports. This model does not expose deploy, prune, quantize, or standalone retrain actions. Resume\u002Fretrain behavior uses `pose_classification train -e ...` with `train.resume_training_checkpoint_path` populated.\n\n## Dataclass Schemas\n\nGenerated TAO Core schemas are packaged in `schemas\u002F\u003Caction>.schema.json`, with `schemas\u002Fmanifest.json` listing available actions. Each generated schema also emits `references\u002Fspec_template_\u003Caction>.yaml` from the schema top-level `default` field. AutoML enablement is declared at the model layer in `references\u002Fskill_info.yaml` via `automl_enabled`. Runnable AutoML still requires `schemas\u002Ftrain.schema.json` and `references\u002Fspec_template_train.yaml` to exist and parse. Use the packaged train schema for `automl_default_parameters`, `automl_disabled_parameters`, defaults, min\u002Fmax bounds, enums, option weights, math conditions, dependencies, and popular parameters. Do not expect `~\u002Ftao-core` at runtime; maintainers regenerate schemas\u002Ftemplates before packaging the skill bank.\n\n## Train Action Policy\n\nThis model is AutoML-enabled at the model layer. Before handling any train-stage request, read `references\u002Fskill_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` \u002F `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\u002Ftrain.schema.json` and `references\u002Fspec_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\u002Fapplication overrides for datasets, specs, output directories, GPU\u002Fplatform settings, parent checkpoints, and `automl_policy`. Use direct model training only when `automl_policy: off` or the packaged train schema\u002Ftemplate is missing; in the missing-schema case, report that AutoML is enabled but not runnable for this model until schemas are generated.\n\nNon-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.\n\n## Training Requirements\n\n- **Dataset type:** pose_classification\n- **Formats:** default\n- **Monitoring metric:** val_loss\n\n### Per-Action Dataset Requirements\n\n| Action | Spec Key | Source | Files | List? |\n|---|---|---|---|---|\n| dataset_convert (optional) | dataset_convert.data | id | DeepStream BodyPose JSON | No |\n| evaluate | evaluate.test_dataset.data_path | train_datasets | val_data.npy | No |\n| evaluate | evaluate.test_dataset.label_path | train_datasets | val_label.pkl | No |\n| inference | inference.test_dataset.data_path | train_datasets | test_data.npy | No |\n| train | dataset.train_dataset.data_path | train_datasets | train_data.npy | No |\n| train | dataset.train_dataset.label_path | train_datasets | train_label.pkl | No |\n| train | dataset.val_dataset.data_path | train_datasets | val_data.npy | No |\n| train | dataset.val_dataset.label_path | train_datasets | val_label.pkl | No |\n\n### Typical Spec Overrides\n\nData source overrides are **mandatory for every action being run** — the agent MUST construct data source paths from the Per-Action Dataset Requirements table above and include them in `spec_overrides`. Do not run `dataset_convert` when the supplied dataset is already converted to `.npy` \u002F `.pkl` files.\n\n```python\nS3_TRAIN = \"s3:\u002F\u002Fbucket\u002Fdata\u002Fpurpose_built_models_pose_classification_train\u002Fnvidia\"\nCHECKPOINT = \"\u002Fresults\u002F{train_job_id}\u002Fresults_dir\u002Fmodel_epoch_000_step_00007.pth\"\n```\n\n**dataset_convert (optional; raw DeepStream BodyPose JSON only):**\n```python\n{\n    \"dataset_convert.data\": \"s3:\u002F\u002Fbucket\u002Fdata\u002F\u003Cdeepstream-bodypose-output>.json\",\n}\n```\n\n**train (mandatory data sources):**\n```python\n{\n    \"train.num_epochs\": 30,\n    \"train.checkpoint_interval\": 10,\n    \"train.validation_interval\": 10,\n    \"train.num_gpus\": 1,\n    \"wandb.enable\": False,\n    \"dataset.num_classes\": 6,\n    \"dataset.label_map\": {\n        \"class_0\": 0,\n        \"class_1\": 1,\n        \"class_2\": 2,\n        \"class_3\": 3,\n        \"class_4\": 4,\n        \"class_5\": 5,\n    },\n    \"model.graph_layout\": \"nvidia\",\n    \"dataset.train_dataset.data_path\": f\"{S3_TRAIN}\u002Ftrain_data.npy\",\n    \"dataset.train_dataset.label_path\": f\"{S3_TRAIN}\u002Ftrain_label.pkl\",\n    \"dataset.val_dataset.data_path\": f\"{S3_TRAIN}\u002Fval_data.npy\",\n    \"dataset.val_dataset.label_path\": f\"{S3_TRAIN}\u002Fval_label.pkl\",\n}\n```\n\n**resume train (mandatory checkpoint):**\n```python\n{\n    \"train.num_epochs\": 31,\n    \"train.resume_training_checkpoint_path\": CHECKPOINT,\n    \"dataset.train_dataset.data_path\": f\"{S3_TRAIN}\u002Ftrain_data.npy\",\n    \"dataset.train_dataset.label_path\": f\"{S3_TRAIN}\u002Ftrain_label.pkl\",\n    \"dataset.val_dataset.data_path\": f\"{S3_TRAIN}\u002Fval_data.npy\",\n    \"dataset.val_dataset.label_path\": f\"{S3_TRAIN}\u002Fval_label.pkl\",\n}\n```\n\n**evaluate (mandatory data sources):**\n```python\n{\n    \"evaluate.test_dataset.data_path\": f\"{S3_TRAIN}\u002Fval_data.npy\",\n    \"evaluate.test_dataset.label_path\": f\"{S3_TRAIN}\u002Fval_label.pkl\",\n    \"evaluate.checkpoint\": CHECKPOINT,\n}\n```\n\n**export (mandatory checkpoint and output):**\n```python\n{\n    \"export.checkpoint\": CHECKPOINT,\n    \"export.onnx_file\": \"\u002Fresults\u002F{export_job_id}\u002Fresults_dir\u002Fpose_classification.onnx\",\n}\n```\n\n**inference (mandatory data sources):**\n```python\n{\n    \"inference.test_dataset.data_path\": f\"{S3_TRAIN}\u002Ftest_data.npy\",\n    \"inference.test_dataset.label_path\": f\"{S3_TRAIN}\u002Ftest_label.pkl\",\n    \"inference.checkpoint\": CHECKPOINT,\n    \"inference.output_file\": \"\u002Fresults\u002Fpose_classification_inference.txt\",\n}\n```\n## Dataset Convert\n\nDataset conversion is optional for Pose Classification. Run `pose_classification dataset_convert` only when the user supplies raw DeepStream BodyPose JSON. For the common S3 validation dataset, the data is already converted to `train_data.npy`, `train_label.pkl`, `val_data.npy`, `val_label.pkl`, `test_data.npy`, and `test_label.pkl`; use those files directly for train\u002Fevaluate\u002Finference\u002Fexport flows and do not synthesize fake BodyPose JSON.\n\n## Eval Dataset\n\nOptional. Validation data is provided alongside training as val_data.npy \u002F val_label.pkl. TAO training emits `val_loss` as the TensorBoard validation scalar for this model; use `val_loss` with minimize direction for AutoML selection unless a custom evaluation hook supplies a different metric.\n\n## Important Parameters\n\n- **dataset.num_classes**: Number of pose action classes. Default 6.\n- **model.graph_layout**: Skeleton graph layout. Options: nvidia, openpose. Determines joint connectivity.\n- **model.graph_strategy**: Graph partitioning strategy for GCN.\n- **train.optim.lr**: Learning rate. Default 0.1 (SGD). Higher than vision models due to graph convolution properties.\n- **model.dropout**: Dropout rate for regularization.\n\n## Multi-GPU \u002F Multi-Node\n\n**Launch method:** Lightning-managed (single `python` process, Lightning spawns workers).\n\n| Spec Key | Description | Default |\n|----------|-------------|---------|\n| `train.num_gpus` | Number of GPUs | 1 |\n| `train.gpu_ids` | GPU device indices | [0] |\n\n- Strategy: `auto` (Lightning picks best strategy automatically)\n- No explicit `num_nodes` or `distributed_strategy` config — single-node only\n- Lightweight model, single GPU typically sufficient\n\n## Hardware\n\nMinimum 1 GPU(s), recommended 1 GPU(s). 8GB+ VRAM per GPU. Pose classification is very lightweight — skeleton data is small. Single GPU is sufficient.\n\n## Error Patterns\n\n**Graph layout mismatch**: Ensure model.graph_layout matches the skeleton format in your .npy data files.\n\n**Label shape mismatch**: train_label.pkl class indices must be in range [0, num_classes).\n\n**Missing label map**: The training dataloader expects `dataset.label_map` to be a dictionary. If the dataset only supplies numeric class IDs, set a synthetic contiguous map such as `class_0: 0` through `class_5: 5` for the six-class NVIDIA sample data.\n\n**Checkpoint handoff**: After AutoML\u002Ftrain, use the checkpoint resolver to select the intended saved `.pth` checkpoint under the parent result folder, such as `model_epoch_000_step_00007.pth`, and pass that exact file as `evaluate.checkpoint`, `export.checkpoint`, `inference.checkpoint`, or `train.resume_training_checkpoint_path`. `pc_model_latest.pth` is a latest-checkpoint symlink; use it only when the user explicitly asks for latest rather than a specific\u002Fbest checkpoint. Keep the same `dataset.num_classes`, `dataset.label_map`, and `model.graph_layout` overrides for downstream actions.\n\n**Dataset conversion source**: `dataset_convert` expects the raw JSON output from the DeepStream BodyPose app. The common NVIDIA sample S3 folder is already converted to `train_data.npy`, `train_label.pkl`, `val_data.npy`, `val_label.pkl`, `test_data.npy`, and `test_label.pkl`; skip conversion and start from the converted files when those are present.\n\n**Action-specific dataset paths**: The evaluate and inference templates also contain the training `dataset.train_dataset` and `dataset.val_dataset` blocks. For evaluate, populate `evaluate.test_dataset.data_path` and `evaluate.test_dataset.label_path`. For inference, populate `inference.test_dataset.data_path` and set `inference.output_file`; do not stop after replacing the first `data_path` or `label_path` in the file.\n\n**Output files**: Export needs an explicit `export.onnx_file` path. Inference must set `inference.output_file` to a writable file path; the packaged template default is an empty string, and the current PyTorch inference code opens that value directly.\n\n## Spec Param \u002F Parent Model Inference\n\nModel-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.\n\nInference mappings from TAO Core `pose_classification.config.json`:\n\n| Action | Spec Field | Inference Function | Meaning |\n|---|---|---|---|\n| dataset_convert | `dataset_convert.results_dir` | `output_dir` | current job results directory |\n| evaluate | `encryption_key` | `key` | encryption key |\n| evaluate | `evaluate.checkpoint` | `parent_model` | model file inferred from the parent job results folder |\n| evaluate | `results_dir` | `output_dir` | current job results directory |\n| export | `encryption_key` | `key` | encryption key |\n| export | `export.checkpoint` | `parent_model` | model file inferred from the parent job results folder |\n| export | `export.onnx_file` | `create_onnx_file` | output ONNX path |\n| export | `results_dir` | `output_dir` | current job results directory |\n| inference | `encryption_key` | `key` | encryption key |\n| inference | `inference.checkpoint` | `parent_model` | model file inferred from the parent job results folder |\n| inference | `inference.output_file` | `create_inference_result_file_pose` | pose inference result file |\n| inference | `results_dir` | `output_dir` | current job results directory |\n| train | `encryption_key` | `key` | encryption key |\n| train | `model.pretrained_model_path` | `ptm_if_no_resume_model` | PTM when no resume checkpoint exists |\n| train | `results_dir` | `output_dir` | current job results directory |\n| train | `train.resume_training_checkpoint_path` | `resume_model` | model file inferred from the current job results folder |\n\nFor `parent_model` or `parent_model_folder`, pass the upstream train\u002Fexport\u002FAutoML 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.\n",{"data":32,"body":41},{"name":4,"description":6,"license":23,"compatibility":33,"metadata":34,"allowed-tools":37,"tags":38},"Requires docker + nvidia-container-toolkit.",{"version":35,"author":36},"0.1.0","NVIDIA Corporation","Read Bash",[39,40],"pose","classification",{"type":42,"children":43},"root",[44,53,59,64,162,169,261,267,375,406,412,448,455,700,706,746,777,785,817,825,1016,1024,1089,1097,1142,1150,1187,1195,1248,1254,1305,1311,1331,1337,1390,1396,1413,1486,1528,1534,1539,1545,1555,1565,1599,1680,1734,1802,1827,1833,1862,1875,2356,2390],{"type":45,"tag":46,"props":47,"children":49},"element","h1",{"id":48},"pose-classification",[50],{"type":51,"value":52},"text","Pose Classification",{"type":45,"tag":54,"props":55,"children":56},"p",{},[57],{"type":51,"value":58},"Pose classification using ST-GCN (Spatial Temporal Graph Convolutional Network). Classifies skeleton sequences into action categories from pose keypoint data.",{"type":45,"tag":54,"props":60,"children":61},{},[62],{"type":51,"value":63},"Typically trained from scratch on skeleton data.",{"type":45,"tag":54,"props":65,"children":66},{},[67,69,76,78,84,85,91,92,98,100,106,108,113,115,121,123,129,131,136,138,144,146,152,154,160],{"type":51,"value":68},"The packaged PyTorch Pose Classification CLI supports ",{"type":45,"tag":70,"props":71,"children":73},"code",{"className":72},[],[74],{"type":51,"value":75},"dataset_convert",{"type":51,"value":77},", ",{"type":45,"tag":70,"props":79,"children":81},{"className":80},[],[82],{"type":51,"value":83},"train",{"type":51,"value":77},{"type":45,"tag":70,"props":86,"children":88},{"className":87},[],[89],{"type":51,"value":90},"evaluate",{"type":51,"value":77},{"type":45,"tag":70,"props":93,"children":95},{"className":94},[],[96],{"type":51,"value":97},"export",{"type":51,"value":99},", and ",{"type":45,"tag":70,"props":101,"children":103},{"className":102},[],[104],{"type":51,"value":105},"inference",{"type":51,"value":107},". ",{"type":45,"tag":70,"props":109,"children":111},{"className":110},[],[112],{"type":51,"value":75},{"type":51,"value":114}," is conditional: run it only when the input is raw DeepStream BodyPose JSON. If the dataset is already converted to TAO-ready ",{"type":45,"tag":70,"props":116,"children":118},{"className":117},[],[119],{"type":51,"value":120},".npy",{"type":51,"value":122}," \u002F ",{"type":45,"tag":70,"props":124,"children":126},{"className":125},[],[127],{"type":51,"value":128},".pkl",{"type":51,"value":130}," files, start directly with ",{"type":45,"tag":70,"props":132,"children":134},{"className":133},[],[135],{"type":51,"value":83},{"type":51,"value":137}," on those files and mark dataset conversion as ",{"type":45,"tag":70,"props":139,"children":141},{"className":140},[],[142],{"type":51,"value":143},"not run: preconverted dataset provided",{"type":51,"value":145}," in validation reports. This model does not expose deploy, prune, quantize, or standalone retrain actions. Resume\u002Fretrain behavior uses ",{"type":45,"tag":70,"props":147,"children":149},{"className":148},[],[150],{"type":51,"value":151},"pose_classification train -e ...",{"type":51,"value":153}," with ",{"type":45,"tag":70,"props":155,"children":157},{"className":156},[],[158],{"type":51,"value":159},"train.resume_training_checkpoint_path",{"type":51,"value":161}," populated.",{"type":45,"tag":163,"props":164,"children":166},"h2",{"id":165},"dataclass-schemas",[167],{"type":51,"value":168},"Dataclass Schemas",{"type":45,"tag":54,"props":170,"children":171},{},[172,174,180,182,188,190,196,198,204,206,212,214,220,222,228,230,236,238,244,245,251,253,259],{"type":51,"value":173},"Generated TAO Core schemas are packaged in ",{"type":45,"tag":70,"props":175,"children":177},{"className":176},[],[178],{"type":51,"value":179},"schemas\u002F\u003Caction>.schema.json",{"type":51,"value":181},", with ",{"type":45,"tag":70,"props":183,"children":185},{"className":184},[],[186],{"type":51,"value":187},"schemas\u002Fmanifest.json",{"type":51,"value":189}," listing available actions. Each generated schema also emits ",{"type":45,"tag":70,"props":191,"children":193},{"className":192},[],[194],{"type":51,"value":195},"references\u002Fspec_template_\u003Caction>.yaml",{"type":51,"value":197}," from the schema top-level ",{"type":45,"tag":70,"props":199,"children":201},{"className":200},[],[202],{"type":51,"value":203},"default",{"type":51,"value":205}," field. AutoML enablement is declared at the model layer in ",{"type":45,"tag":70,"props":207,"children":209},{"className":208},[],[210],{"type":51,"value":211},"references\u002Fskill_info.yaml",{"type":51,"value":213}," via ",{"type":45,"tag":70,"props":215,"children":217},{"className":216},[],[218],{"type":51,"value":219},"automl_enabled",{"type":51,"value":221},". Runnable AutoML still requires ",{"type":45,"tag":70,"props":223,"children":225},{"className":224},[],[226],{"type":51,"value":227},"schemas\u002Ftrain.schema.json",{"type":51,"value":229}," and ",{"type":45,"tag":70,"props":231,"children":233},{"className":232},[],[234],{"type":51,"value":235},"references\u002Fspec_template_train.yaml",{"type":51,"value":237}," to exist and parse. Use the packaged train schema for ",{"type":45,"tag":70,"props":239,"children":241},{"className":240},[],[242],{"type":51,"value":243},"automl_default_parameters",{"type":51,"value":77},{"type":45,"tag":70,"props":246,"children":248},{"className":247},[],[249],{"type":51,"value":250},"automl_disabled_parameters",{"type":51,"value":252},", defaults, min\u002Fmax bounds, enums, option weights, math conditions, dependencies, and popular parameters. Do not expect ",{"type":45,"tag":70,"props":254,"children":256},{"className":255},[],[257],{"type":51,"value":258},"~\u002Ftao-core",{"type":51,"value":260}," at runtime; maintainers regenerate schemas\u002Ftemplates before packaging the skill bank.",{"type":45,"tag":163,"props":262,"children":264},{"id":263},"train-action-policy",[265],{"type":51,"value":266},"Train Action Policy",{"type":45,"tag":54,"props":268,"children":269},{},[270,272,277,279,285,287,293,295,301,302,308,310,316,318,323,324,330,332,337,338,343,345,351,353,359,361,366,368,373],{"type":51,"value":271},"This model is AutoML-enabled at the model layer. Before handling any train-stage request, read ",{"type":45,"tag":70,"props":273,"children":275},{"className":274},[],[276],{"type":51,"value":211},{"type":51,"value":278}," and resolve the run override from either an explicit ",{"type":45,"tag":70,"props":280,"children":282},{"className":281},[],[283],{"type":51,"value":284},"automl_policy",{"type":51,"value":286}," value or the user's workflow request. Use ",{"type":45,"tag":70,"props":288,"children":290},{"className":289},[],[291],{"type":51,"value":292},"automl_policy: on",{"type":51,"value":294}," by default and only expose ",{"type":45,"tag":70,"props":296,"children":298},{"className":297},[],[299],{"type":51,"value":300},"on",{"type":51,"value":122},{"type":45,"tag":70,"props":303,"children":305},{"className":304},[],[306],{"type":51,"value":307},"off",{"type":51,"value":309}," in new launch prompts. Treat phrases like \"turn off AutoML\", \"disable AutoML\", \"no HPO\", or \"plain training\" as ",{"type":45,"tag":70,"props":311,"children":313},{"className":312},[],[314],{"type":51,"value":315},"automl_policy: off",{"type":51,"value":317}," for this run only. When ",{"type":45,"tag":70,"props":319,"children":321},{"className":320},[],[322],{"type":51,"value":292},{"type":51,"value":77},{"type":45,"tag":70,"props":325,"children":327},{"className":326},[],[328],{"type":51,"value":329},"automl_enabled: true",{"type":51,"value":331},", and both ",{"type":45,"tag":70,"props":333,"children":335},{"className":334},[],[336],{"type":51,"value":227},{"type":51,"value":229},{"type":45,"tag":70,"props":339,"children":341},{"className":340},[],[342],{"type":51,"value":235},{"type":51,"value":344}," are packaged, route the train action through ",{"type":45,"tag":70,"props":346,"children":348},{"className":347},[],[349],{"type":51,"value":350},"tao-skill-bank:tao-run-automl",{"type":51,"value":352}," by default with this model's ",{"type":45,"tag":70,"props":354,"children":356},{"className":355},[],[357],{"type":51,"value":358},"skill_dir",{"type":51,"value":360},". Preserve workflow\u002Fapplication overrides for datasets, specs, output directories, GPU\u002Fplatform settings, parent checkpoints, and ",{"type":45,"tag":70,"props":362,"children":364},{"className":363},[],[365],{"type":51,"value":284},{"type":51,"value":367},". Use direct model training only when ",{"type":45,"tag":70,"props":369,"children":371},{"className":370},[],[372],{"type":51,"value":315},{"type":51,"value":374}," or the packaged train schema\u002Ftemplate is missing; in the missing-schema case, report that AutoML is enabled but not runnable for this model until schemas are generated.",{"type":45,"tag":54,"props":376,"children":377},{},[378,380,385,386,391,392,397,399,404],{"type":51,"value":379},"Non-train actions such as ",{"type":45,"tag":70,"props":381,"children":383},{"className":382},[],[384],{"type":51,"value":90},{"type":51,"value":77},{"type":45,"tag":70,"props":387,"children":389},{"className":388},[],[390],{"type":51,"value":105},{"type":51,"value":77},{"type":45,"tag":70,"props":393,"children":395},{"className":394},[],[396],{"type":51,"value":97},{"type":51,"value":398},", and deploy flows stay in this model skill. The per-run ",{"type":45,"tag":70,"props":400,"children":402},{"className":401},[],[403],{"type":51,"value":284},{"type":51,"value":405}," override does not change model metadata.",{"type":45,"tag":163,"props":407,"children":409},{"id":408},"training-requirements",[410],{"type":51,"value":411},"Training Requirements",{"type":45,"tag":413,"props":414,"children":415},"ul",{},[416,428,438],{"type":45,"tag":417,"props":418,"children":419},"li",{},[420,426],{"type":45,"tag":421,"props":422,"children":423},"strong",{},[424],{"type":51,"value":425},"Dataset type:",{"type":51,"value":427}," pose_classification",{"type":45,"tag":417,"props":429,"children":430},{},[431,436],{"type":45,"tag":421,"props":432,"children":433},{},[434],{"type":51,"value":435},"Formats:",{"type":51,"value":437}," default",{"type":45,"tag":417,"props":439,"children":440},{},[441,446],{"type":45,"tag":421,"props":442,"children":443},{},[444],{"type":51,"value":445},"Monitoring metric:",{"type":51,"value":447}," val_loss",{"type":45,"tag":449,"props":450,"children":452},"h3",{"id":451},"per-action-dataset-requirements",[453],{"type":51,"value":454},"Per-Action Dataset Requirements",{"type":45,"tag":456,"props":457,"children":458},"table",{},[459,493],{"type":45,"tag":460,"props":461,"children":462},"thead",{},[463],{"type":45,"tag":464,"props":465,"children":466},"tr",{},[467,473,478,483,488],{"type":45,"tag":468,"props":469,"children":470},"th",{},[471],{"type":51,"value":472},"Action",{"type":45,"tag":468,"props":474,"children":475},{},[476],{"type":51,"value":477},"Spec Key",{"type":45,"tag":468,"props":479,"children":480},{},[481],{"type":51,"value":482},"Source",{"type":45,"tag":468,"props":484,"children":485},{},[486],{"type":51,"value":487},"Files",{"type":45,"tag":468,"props":489,"children":490},{},[491],{"type":51,"value":492},"List?",{"type":45,"tag":494,"props":495,"children":496},"tbody",{},[497,526,552,577,602,627,652,676],{"type":45,"tag":464,"props":498,"children":499},{},[500,506,511,516,521],{"type":45,"tag":501,"props":502,"children":503},"td",{},[504],{"type":51,"value":505},"dataset_convert (optional)",{"type":45,"tag":501,"props":507,"children":508},{},[509],{"type":51,"value":510},"dataset_convert.data",{"type":45,"tag":501,"props":512,"children":513},{},[514],{"type":51,"value":515},"id",{"type":45,"tag":501,"props":517,"children":518},{},[519],{"type":51,"value":520},"DeepStream BodyPose JSON",{"type":45,"tag":501,"props":522,"children":523},{},[524],{"type":51,"value":525},"No",{"type":45,"tag":464,"props":527,"children":528},{},[529,533,538,543,548],{"type":45,"tag":501,"props":530,"children":531},{},[532],{"type":51,"value":90},{"type":45,"tag":501,"props":534,"children":535},{},[536],{"type":51,"value":537},"evaluate.test_dataset.data_path",{"type":45,"tag":501,"props":539,"children":540},{},[541],{"type":51,"value":542},"train_datasets",{"type":45,"tag":501,"props":544,"children":545},{},[546],{"type":51,"value":547},"val_data.npy",{"type":45,"tag":501,"props":549,"children":550},{},[551],{"type":51,"value":525},{"type":45,"tag":464,"props":553,"children":554},{},[555,559,564,568,573],{"type":45,"tag":501,"props":556,"children":557},{},[558],{"type":51,"value":90},{"type":45,"tag":501,"props":560,"children":561},{},[562],{"type":51,"value":563},"evaluate.test_dataset.label_path",{"type":45,"tag":501,"props":565,"children":566},{},[567],{"type":51,"value":542},{"type":45,"tag":501,"props":569,"children":570},{},[571],{"type":51,"value":572},"val_label.pkl",{"type":45,"tag":501,"props":574,"children":575},{},[576],{"type":51,"value":525},{"type":45,"tag":464,"props":578,"children":579},{},[580,584,589,593,598],{"type":45,"tag":501,"props":581,"children":582},{},[583],{"type":51,"value":105},{"type":45,"tag":501,"props":585,"children":586},{},[587],{"type":51,"value":588},"inference.test_dataset.data_path",{"type":45,"tag":501,"props":590,"children":591},{},[592],{"type":51,"value":542},{"type":45,"tag":501,"props":594,"children":595},{},[596],{"type":51,"value":597},"test_data.npy",{"type":45,"tag":501,"props":599,"children":600},{},[601],{"type":51,"value":525},{"type":45,"tag":464,"props":603,"children":604},{},[605,609,614,618,623],{"type":45,"tag":501,"props":606,"children":607},{},[608],{"type":51,"value":83},{"type":45,"tag":501,"props":610,"children":611},{},[612],{"type":51,"value":613},"dataset.train_dataset.data_path",{"type":45,"tag":501,"props":615,"children":616},{},[617],{"type":51,"value":542},{"type":45,"tag":501,"props":619,"children":620},{},[621],{"type":51,"value":622},"train_data.npy",{"type":45,"tag":501,"props":624,"children":625},{},[626],{"type":51,"value":525},{"type":45,"tag":464,"props":628,"children":629},{},[630,634,639,643,648],{"type":45,"tag":501,"props":631,"children":632},{},[633],{"type":51,"value":83},{"type":45,"tag":501,"props":635,"children":636},{},[637],{"type":51,"value":638},"dataset.train_dataset.label_path",{"type":45,"tag":501,"props":640,"children":641},{},[642],{"type":51,"value":542},{"type":45,"tag":501,"props":644,"children":645},{},[646],{"type":51,"value":647},"train_label.pkl",{"type":45,"tag":501,"props":649,"children":650},{},[651],{"type":51,"value":525},{"type":45,"tag":464,"props":653,"children":654},{},[655,659,664,668,672],{"type":45,"tag":501,"props":656,"children":657},{},[658],{"type":51,"value":83},{"type":45,"tag":501,"props":660,"children":661},{},[662],{"type":51,"value":663},"dataset.val_dataset.data_path",{"type":45,"tag":501,"props":665,"children":666},{},[667],{"type":51,"value":542},{"type":45,"tag":501,"props":669,"children":670},{},[671],{"type":51,"value":547},{"type":45,"tag":501,"props":673,"children":674},{},[675],{"type":51,"value":525},{"type":45,"tag":464,"props":677,"children":678},{},[679,683,688,692,696],{"type":45,"tag":501,"props":680,"children":681},{},[682],{"type":51,"value":83},{"type":45,"tag":501,"props":684,"children":685},{},[686],{"type":51,"value":687},"dataset.val_dataset.label_path",{"type":45,"tag":501,"props":689,"children":690},{},[691],{"type":51,"value":542},{"type":45,"tag":501,"props":693,"children":694},{},[695],{"type":51,"value":572},{"type":45,"tag":501,"props":697,"children":698},{},[699],{"type":51,"value":525},{"type":45,"tag":449,"props":701,"children":703},{"id":702},"typical-spec-overrides",[704],{"type":51,"value":705},"Typical Spec Overrides",{"type":45,"tag":54,"props":707,"children":708},{},[709,711,716,718,724,726,731,733,738,739,744],{"type":51,"value":710},"Data source overrides are ",{"type":45,"tag":421,"props":712,"children":713},{},[714],{"type":51,"value":715},"mandatory for every action being run",{"type":51,"value":717}," — the agent MUST construct data source paths from the Per-Action Dataset Requirements table above and include them in ",{"type":45,"tag":70,"props":719,"children":721},{"className":720},[],[722],{"type":51,"value":723},"spec_overrides",{"type":51,"value":725},". Do not run ",{"type":45,"tag":70,"props":727,"children":729},{"className":728},[],[730],{"type":51,"value":75},{"type":51,"value":732}," when the supplied dataset is already converted to ",{"type":45,"tag":70,"props":734,"children":736},{"className":735},[],[737],{"type":51,"value":120},{"type":51,"value":122},{"type":45,"tag":70,"props":740,"children":742},{"className":741},[],[743],{"type":51,"value":128},{"type":51,"value":745}," files.",{"type":45,"tag":747,"props":748,"children":753},"pre",{"className":749,"code":750,"language":751,"meta":752,"style":752},"language-python shiki shiki-themes material-theme-lighter material-theme material-theme-palenight","S3_TRAIN = \"s3:\u002F\u002Fbucket\u002Fdata\u002Fpurpose_built_models_pose_classification_train\u002Fnvidia\"\nCHECKPOINT = \"\u002Fresults\u002F{train_job_id}\u002Fresults_dir\u002Fmodel_epoch_000_step_00007.pth\"\n","python","",[754],{"type":45,"tag":70,"props":755,"children":756},{"__ignoreMap":752},[757,768],{"type":45,"tag":758,"props":759,"children":762},"span",{"class":760,"line":761},"line",1,[763],{"type":45,"tag":758,"props":764,"children":765},{},[766],{"type":51,"value":767},"S3_TRAIN = \"s3:\u002F\u002Fbucket\u002Fdata\u002Fpurpose_built_models_pose_classification_train\u002Fnvidia\"\n",{"type":45,"tag":758,"props":769,"children":771},{"class":760,"line":770},2,[772],{"type":45,"tag":758,"props":773,"children":774},{},[775],{"type":51,"value":776},"CHECKPOINT = \"\u002Fresults\u002F{train_job_id}\u002Fresults_dir\u002Fmodel_epoch_000_step_00007.pth\"\n",{"type":45,"tag":54,"props":778,"children":779},{},[780],{"type":45,"tag":421,"props":781,"children":782},{},[783],{"type":51,"value":784},"dataset_convert (optional; raw DeepStream BodyPose JSON only):",{"type":45,"tag":747,"props":786,"children":788},{"className":749,"code":787,"language":751,"meta":752,"style":752},"{\n    \"dataset_convert.data\": \"s3:\u002F\u002Fbucket\u002Fdata\u002F\u003Cdeepstream-bodypose-output>.json\",\n}\n",[789],{"type":45,"tag":70,"props":790,"children":791},{"__ignoreMap":752},[792,800,808],{"type":45,"tag":758,"props":793,"children":794},{"class":760,"line":761},[795],{"type":45,"tag":758,"props":796,"children":797},{},[798],{"type":51,"value":799},"{\n",{"type":45,"tag":758,"props":801,"children":802},{"class":760,"line":770},[803],{"type":45,"tag":758,"props":804,"children":805},{},[806],{"type":51,"value":807},"    \"dataset_convert.data\": \"s3:\u002F\u002Fbucket\u002Fdata\u002F\u003Cdeepstream-bodypose-output>.json\",\n",{"type":45,"tag":758,"props":809,"children":811},{"class":760,"line":810},3,[812],{"type":45,"tag":758,"props":813,"children":814},{},[815],{"type":51,"value":816},"}\n",{"type":45,"tag":54,"props":818,"children":819},{},[820],{"type":45,"tag":421,"props":821,"children":822},{},[823],{"type":51,"value":824},"train (mandatory data sources):",{"type":45,"tag":747,"props":826,"children":828},{"className":749,"code":827,"language":751,"meta":752,"style":752},"{\n    \"train.num_epochs\": 30,\n    \"train.checkpoint_interval\": 10,\n    \"train.validation_interval\": 10,\n    \"train.num_gpus\": 1,\n    \"wandb.enable\": False,\n    \"dataset.num_classes\": 6,\n    \"dataset.label_map\": {\n        \"class_0\": 0,\n        \"class_1\": 1,\n        \"class_2\": 2,\n        \"class_3\": 3,\n        \"class_4\": 4,\n        \"class_5\": 5,\n    },\n    \"model.graph_layout\": \"nvidia\",\n    \"dataset.train_dataset.data_path\": f\"{S3_TRAIN}\u002Ftrain_data.npy\",\n    \"dataset.train_dataset.label_path\": f\"{S3_TRAIN}\u002Ftrain_label.pkl\",\n    \"dataset.val_dataset.data_path\": f\"{S3_TRAIN}\u002Fval_data.npy\",\n    \"dataset.val_dataset.label_path\": f\"{S3_TRAIN}\u002Fval_label.pkl\",\n}\n",[829],{"type":45,"tag":70,"props":830,"children":831},{"__ignoreMap":752},[832,839,847,855,864,873,882,891,900,909,918,927,936,945,954,963,972,981,990,999,1008],{"type":45,"tag":758,"props":833,"children":834},{"class":760,"line":761},[835],{"type":45,"tag":758,"props":836,"children":837},{},[838],{"type":51,"value":799},{"type":45,"tag":758,"props":840,"children":841},{"class":760,"line":770},[842],{"type":45,"tag":758,"props":843,"children":844},{},[845],{"type":51,"value":846},"    \"train.num_epochs\": 30,\n",{"type":45,"tag":758,"props":848,"children":849},{"class":760,"line":810},[850],{"type":45,"tag":758,"props":851,"children":852},{},[853],{"type":51,"value":854},"    \"train.checkpoint_interval\": 10,\n",{"type":45,"tag":758,"props":856,"children":858},{"class":760,"line":857},4,[859],{"type":45,"tag":758,"props":860,"children":861},{},[862],{"type":51,"value":863},"    \"train.validation_interval\": 10,\n",{"type":45,"tag":758,"props":865,"children":867},{"class":760,"line":866},5,[868],{"type":45,"tag":758,"props":869,"children":870},{},[871],{"type":51,"value":872},"    \"train.num_gpus\": 1,\n",{"type":45,"tag":758,"props":874,"children":876},{"class":760,"line":875},6,[877],{"type":45,"tag":758,"props":878,"children":879},{},[880],{"type":51,"value":881},"    \"wandb.enable\": False,\n",{"type":45,"tag":758,"props":883,"children":885},{"class":760,"line":884},7,[886],{"type":45,"tag":758,"props":887,"children":888},{},[889],{"type":51,"value":890},"    \"dataset.num_classes\": 6,\n",{"type":45,"tag":758,"props":892,"children":894},{"class":760,"line":893},8,[895],{"type":45,"tag":758,"props":896,"children":897},{},[898],{"type":51,"value":899},"    \"dataset.label_map\": {\n",{"type":45,"tag":758,"props":901,"children":903},{"class":760,"line":902},9,[904],{"type":45,"tag":758,"props":905,"children":906},{},[907],{"type":51,"value":908},"        \"class_0\": 0,\n",{"type":45,"tag":758,"props":910,"children":912},{"class":760,"line":911},10,[913],{"type":45,"tag":758,"props":914,"children":915},{},[916],{"type":51,"value":917},"        \"class_1\": 1,\n",{"type":45,"tag":758,"props":919,"children":921},{"class":760,"line":920},11,[922],{"type":45,"tag":758,"props":923,"children":924},{},[925],{"type":51,"value":926},"        \"class_2\": 2,\n",{"type":45,"tag":758,"props":928,"children":930},{"class":760,"line":929},12,[931],{"type":45,"tag":758,"props":932,"children":933},{},[934],{"type":51,"value":935},"        \"class_3\": 3,\n",{"type":45,"tag":758,"props":937,"children":939},{"class":760,"line":938},13,[940],{"type":45,"tag":758,"props":941,"children":942},{},[943],{"type":51,"value":944},"        \"class_4\": 4,\n",{"type":45,"tag":758,"props":946,"children":948},{"class":760,"line":947},14,[949],{"type":45,"tag":758,"props":950,"children":951},{},[952],{"type":51,"value":953},"        \"class_5\": 5,\n",{"type":45,"tag":758,"props":955,"children":957},{"class":760,"line":956},15,[958],{"type":45,"tag":758,"props":959,"children":960},{},[961],{"type":51,"value":962},"    },\n",{"type":45,"tag":758,"props":964,"children":966},{"class":760,"line":965},16,[967],{"type":45,"tag":758,"props":968,"children":969},{},[970],{"type":51,"value":971},"    \"model.graph_layout\": \"nvidia\",\n",{"type":45,"tag":758,"props":973,"children":975},{"class":760,"line":974},17,[976],{"type":45,"tag":758,"props":977,"children":978},{},[979],{"type":51,"value":980},"    \"dataset.train_dataset.data_path\": f\"{S3_TRAIN}\u002Ftrain_data.npy\",\n",{"type":45,"tag":758,"props":982,"children":984},{"class":760,"line":983},18,[985],{"type":45,"tag":758,"props":986,"children":987},{},[988],{"type":51,"value":989},"    \"dataset.train_dataset.label_path\": f\"{S3_TRAIN}\u002Ftrain_label.pkl\",\n",{"type":45,"tag":758,"props":991,"children":993},{"class":760,"line":992},19,[994],{"type":45,"tag":758,"props":995,"children":996},{},[997],{"type":51,"value":998},"    \"dataset.val_dataset.data_path\": f\"{S3_TRAIN}\u002Fval_data.npy\",\n",{"type":45,"tag":758,"props":1000,"children":1002},{"class":760,"line":1001},20,[1003],{"type":45,"tag":758,"props":1004,"children":1005},{},[1006],{"type":51,"value":1007},"    \"dataset.val_dataset.label_path\": f\"{S3_TRAIN}\u002Fval_label.pkl\",\n",{"type":45,"tag":758,"props":1009,"children":1011},{"class":760,"line":1010},21,[1012],{"type":45,"tag":758,"props":1013,"children":1014},{},[1015],{"type":51,"value":816},{"type":45,"tag":54,"props":1017,"children":1018},{},[1019],{"type":45,"tag":421,"props":1020,"children":1021},{},[1022],{"type":51,"value":1023},"resume train (mandatory checkpoint):",{"type":45,"tag":747,"props":1025,"children":1027},{"className":749,"code":1026,"language":751,"meta":752,"style":752},"{\n    \"train.num_epochs\": 31,\n    \"train.resume_training_checkpoint_path\": CHECKPOINT,\n    \"dataset.train_dataset.data_path\": f\"{S3_TRAIN}\u002Ftrain_data.npy\",\n    \"dataset.train_dataset.label_path\": f\"{S3_TRAIN}\u002Ftrain_label.pkl\",\n    \"dataset.val_dataset.data_path\": f\"{S3_TRAIN}\u002Fval_data.npy\",\n    \"dataset.val_dataset.label_path\": f\"{S3_TRAIN}\u002Fval_label.pkl\",\n}\n",[1028],{"type":45,"tag":70,"props":1029,"children":1030},{"__ignoreMap":752},[1031,1038,1046,1054,1061,1068,1075,1082],{"type":45,"tag":758,"props":1032,"children":1033},{"class":760,"line":761},[1034],{"type":45,"tag":758,"props":1035,"children":1036},{},[1037],{"type":51,"value":799},{"type":45,"tag":758,"props":1039,"children":1040},{"class":760,"line":770},[1041],{"type":45,"tag":758,"props":1042,"children":1043},{},[1044],{"type":51,"value":1045},"    \"train.num_epochs\": 31,\n",{"type":45,"tag":758,"props":1047,"children":1048},{"class":760,"line":810},[1049],{"type":45,"tag":758,"props":1050,"children":1051},{},[1052],{"type":51,"value":1053},"    \"train.resume_training_checkpoint_path\": CHECKPOINT,\n",{"type":45,"tag":758,"props":1055,"children":1056},{"class":760,"line":857},[1057],{"type":45,"tag":758,"props":1058,"children":1059},{},[1060],{"type":51,"value":980},{"type":45,"tag":758,"props":1062,"children":1063},{"class":760,"line":866},[1064],{"type":45,"tag":758,"props":1065,"children":1066},{},[1067],{"type":51,"value":989},{"type":45,"tag":758,"props":1069,"children":1070},{"class":760,"line":875},[1071],{"type":45,"tag":758,"props":1072,"children":1073},{},[1074],{"type":51,"value":998},{"type":45,"tag":758,"props":1076,"children":1077},{"class":760,"line":884},[1078],{"type":45,"tag":758,"props":1079,"children":1080},{},[1081],{"type":51,"value":1007},{"type":45,"tag":758,"props":1083,"children":1084},{"class":760,"line":893},[1085],{"type":45,"tag":758,"props":1086,"children":1087},{},[1088],{"type":51,"value":816},{"type":45,"tag":54,"props":1090,"children":1091},{},[1092],{"type":45,"tag":421,"props":1093,"children":1094},{},[1095],{"type":51,"value":1096},"evaluate (mandatory data sources):",{"type":45,"tag":747,"props":1098,"children":1100},{"className":749,"code":1099,"language":751,"meta":752,"style":752},"{\n    \"evaluate.test_dataset.data_path\": f\"{S3_TRAIN}\u002Fval_data.npy\",\n    \"evaluate.test_dataset.label_path\": f\"{S3_TRAIN}\u002Fval_label.pkl\",\n    \"evaluate.checkpoint\": CHECKPOINT,\n}\n",[1101],{"type":45,"tag":70,"props":1102,"children":1103},{"__ignoreMap":752},[1104,1111,1119,1127,1135],{"type":45,"tag":758,"props":1105,"children":1106},{"class":760,"line":761},[1107],{"type":45,"tag":758,"props":1108,"children":1109},{},[1110],{"type":51,"value":799},{"type":45,"tag":758,"props":1112,"children":1113},{"class":760,"line":770},[1114],{"type":45,"tag":758,"props":1115,"children":1116},{},[1117],{"type":51,"value":1118},"    \"evaluate.test_dataset.data_path\": f\"{S3_TRAIN}\u002Fval_data.npy\",\n",{"type":45,"tag":758,"props":1120,"children":1121},{"class":760,"line":810},[1122],{"type":45,"tag":758,"props":1123,"children":1124},{},[1125],{"type":51,"value":1126},"    \"evaluate.test_dataset.label_path\": f\"{S3_TRAIN}\u002Fval_label.pkl\",\n",{"type":45,"tag":758,"props":1128,"children":1129},{"class":760,"line":857},[1130],{"type":45,"tag":758,"props":1131,"children":1132},{},[1133],{"type":51,"value":1134},"    \"evaluate.checkpoint\": CHECKPOINT,\n",{"type":45,"tag":758,"props":1136,"children":1137},{"class":760,"line":866},[1138],{"type":45,"tag":758,"props":1139,"children":1140},{},[1141],{"type":51,"value":816},{"type":45,"tag":54,"props":1143,"children":1144},{},[1145],{"type":45,"tag":421,"props":1146,"children":1147},{},[1148],{"type":51,"value":1149},"export (mandatory checkpoint and output):",{"type":45,"tag":747,"props":1151,"children":1153},{"className":749,"code":1152,"language":751,"meta":752,"style":752},"{\n    \"export.checkpoint\": CHECKPOINT,\n    \"export.onnx_file\": \"\u002Fresults\u002F{export_job_id}\u002Fresults_dir\u002Fpose_classification.onnx\",\n}\n",[1154],{"type":45,"tag":70,"props":1155,"children":1156},{"__ignoreMap":752},[1157,1164,1172,1180],{"type":45,"tag":758,"props":1158,"children":1159},{"class":760,"line":761},[1160],{"type":45,"tag":758,"props":1161,"children":1162},{},[1163],{"type":51,"value":799},{"type":45,"tag":758,"props":1165,"children":1166},{"class":760,"line":770},[1167],{"type":45,"tag":758,"props":1168,"children":1169},{},[1170],{"type":51,"value":1171},"    \"export.checkpoint\": CHECKPOINT,\n",{"type":45,"tag":758,"props":1173,"children":1174},{"class":760,"line":810},[1175],{"type":45,"tag":758,"props":1176,"children":1177},{},[1178],{"type":51,"value":1179},"    \"export.onnx_file\": \"\u002Fresults\u002F{export_job_id}\u002Fresults_dir\u002Fpose_classification.onnx\",\n",{"type":45,"tag":758,"props":1181,"children":1182},{"class":760,"line":857},[1183],{"type":45,"tag":758,"props":1184,"children":1185},{},[1186],{"type":51,"value":816},{"type":45,"tag":54,"props":1188,"children":1189},{},[1190],{"type":45,"tag":421,"props":1191,"children":1192},{},[1193],{"type":51,"value":1194},"inference (mandatory data sources):",{"type":45,"tag":747,"props":1196,"children":1198},{"className":749,"code":1197,"language":751,"meta":752,"style":752},"{\n    \"inference.test_dataset.data_path\": f\"{S3_TRAIN}\u002Ftest_data.npy\",\n    \"inference.test_dataset.label_path\": f\"{S3_TRAIN}\u002Ftest_label.pkl\",\n    \"inference.checkpoint\": CHECKPOINT,\n    \"inference.output_file\": \"\u002Fresults\u002Fpose_classification_inference.txt\",\n}\n",[1199],{"type":45,"tag":70,"props":1200,"children":1201},{"__ignoreMap":752},[1202,1209,1217,1225,1233,1241],{"type":45,"tag":758,"props":1203,"children":1204},{"class":760,"line":761},[1205],{"type":45,"tag":758,"props":1206,"children":1207},{},[1208],{"type":51,"value":799},{"type":45,"tag":758,"props":1210,"children":1211},{"class":760,"line":770},[1212],{"type":45,"tag":758,"props":1213,"children":1214},{},[1215],{"type":51,"value":1216},"    \"inference.test_dataset.data_path\": f\"{S3_TRAIN}\u002Ftest_data.npy\",\n",{"type":45,"tag":758,"props":1218,"children":1219},{"class":760,"line":810},[1220],{"type":45,"tag":758,"props":1221,"children":1222},{},[1223],{"type":51,"value":1224},"    \"inference.test_dataset.label_path\": f\"{S3_TRAIN}\u002Ftest_label.pkl\",\n",{"type":45,"tag":758,"props":1226,"children":1227},{"class":760,"line":857},[1228],{"type":45,"tag":758,"props":1229,"children":1230},{},[1231],{"type":51,"value":1232},"    \"inference.checkpoint\": CHECKPOINT,\n",{"type":45,"tag":758,"props":1234,"children":1235},{"class":760,"line":866},[1236],{"type":45,"tag":758,"props":1237,"children":1238},{},[1239],{"type":51,"value":1240},"    \"inference.output_file\": \"\u002Fresults\u002Fpose_classification_inference.txt\",\n",{"type":45,"tag":758,"props":1242,"children":1243},{"class":760,"line":875},[1244],{"type":45,"tag":758,"props":1245,"children":1246},{},[1247],{"type":51,"value":816},{"type":45,"tag":163,"props":1249,"children":1251},{"id":1250},"dataset-convert",[1252],{"type":51,"value":1253},"Dataset Convert",{"type":45,"tag":54,"props":1255,"children":1256},{},[1257,1259,1265,1267,1272,1273,1278,1279,1284,1285,1290,1291,1296,1297,1303],{"type":51,"value":1258},"Dataset conversion is optional for Pose Classification. Run ",{"type":45,"tag":70,"props":1260,"children":1262},{"className":1261},[],[1263],{"type":51,"value":1264},"pose_classification dataset_convert",{"type":51,"value":1266}," only when the user supplies raw DeepStream BodyPose JSON. For the common S3 validation dataset, the data is already converted to ",{"type":45,"tag":70,"props":1268,"children":1270},{"className":1269},[],[1271],{"type":51,"value":622},{"type":51,"value":77},{"type":45,"tag":70,"props":1274,"children":1276},{"className":1275},[],[1277],{"type":51,"value":647},{"type":51,"value":77},{"type":45,"tag":70,"props":1280,"children":1282},{"className":1281},[],[1283],{"type":51,"value":547},{"type":51,"value":77},{"type":45,"tag":70,"props":1286,"children":1288},{"className":1287},[],[1289],{"type":51,"value":572},{"type":51,"value":77},{"type":45,"tag":70,"props":1292,"children":1294},{"className":1293},[],[1295],{"type":51,"value":597},{"type":51,"value":99},{"type":45,"tag":70,"props":1298,"children":1300},{"className":1299},[],[1301],{"type":51,"value":1302},"test_label.pkl",{"type":51,"value":1304},"; use those files directly for train\u002Fevaluate\u002Finference\u002Fexport flows and do not synthesize fake BodyPose JSON.",{"type":45,"tag":163,"props":1306,"children":1308},{"id":1307},"eval-dataset",[1309],{"type":51,"value":1310},"Eval Dataset",{"type":45,"tag":54,"props":1312,"children":1313},{},[1314,1316,1322,1324,1329],{"type":51,"value":1315},"Optional. Validation data is provided alongside training as val_data.npy \u002F val_label.pkl. TAO training emits ",{"type":45,"tag":70,"props":1317,"children":1319},{"className":1318},[],[1320],{"type":51,"value":1321},"val_loss",{"type":51,"value":1323}," as the TensorBoard validation scalar for this model; use ",{"type":45,"tag":70,"props":1325,"children":1327},{"className":1326},[],[1328],{"type":51,"value":1321},{"type":51,"value":1330}," with minimize direction for AutoML selection unless a custom evaluation hook supplies a different metric.",{"type":45,"tag":163,"props":1332,"children":1334},{"id":1333},"important-parameters",[1335],{"type":51,"value":1336},"Important Parameters",{"type":45,"tag":413,"props":1338,"children":1339},{},[1340,1350,1360,1370,1380],{"type":45,"tag":417,"props":1341,"children":1342},{},[1343,1348],{"type":45,"tag":421,"props":1344,"children":1345},{},[1346],{"type":51,"value":1347},"dataset.num_classes",{"type":51,"value":1349},": Number of pose action classes. Default 6.",{"type":45,"tag":417,"props":1351,"children":1352},{},[1353,1358],{"type":45,"tag":421,"props":1354,"children":1355},{},[1356],{"type":51,"value":1357},"model.graph_layout",{"type":51,"value":1359},": Skeleton graph layout. Options: nvidia, openpose. Determines joint connectivity.",{"type":45,"tag":417,"props":1361,"children":1362},{},[1363,1368],{"type":45,"tag":421,"props":1364,"children":1365},{},[1366],{"type":51,"value":1367},"model.graph_strategy",{"type":51,"value":1369},": Graph partitioning strategy for GCN.",{"type":45,"tag":417,"props":1371,"children":1372},{},[1373,1378],{"type":45,"tag":421,"props":1374,"children":1375},{},[1376],{"type":51,"value":1377},"train.optim.lr",{"type":51,"value":1379},": Learning rate. Default 0.1 (SGD). Higher than vision models due to graph convolution properties.",{"type":45,"tag":417,"props":1381,"children":1382},{},[1383,1388],{"type":45,"tag":421,"props":1384,"children":1385},{},[1386],{"type":51,"value":1387},"model.dropout",{"type":51,"value":1389},": Dropout rate for regularization.",{"type":45,"tag":163,"props":1391,"children":1393},{"id":1392},"multi-gpu-multi-node",[1394],{"type":51,"value":1395},"Multi-GPU \u002F Multi-Node",{"type":45,"tag":54,"props":1397,"children":1398},{},[1399,1404,1406,1411],{"type":45,"tag":421,"props":1400,"children":1401},{},[1402],{"type":51,"value":1403},"Launch method:",{"type":51,"value":1405}," Lightning-managed (single ",{"type":45,"tag":70,"props":1407,"children":1409},{"className":1408},[],[1410],{"type":51,"value":751},{"type":51,"value":1412}," process, Lightning spawns workers).",{"type":45,"tag":456,"props":1414,"children":1415},{},[1416,1436],{"type":45,"tag":460,"props":1417,"children":1418},{},[1419],{"type":45,"tag":464,"props":1420,"children":1421},{},[1422,1426,1431],{"type":45,"tag":468,"props":1423,"children":1424},{},[1425],{"type":51,"value":477},{"type":45,"tag":468,"props":1427,"children":1428},{},[1429],{"type":51,"value":1430},"Description",{"type":45,"tag":468,"props":1432,"children":1433},{},[1434],{"type":51,"value":1435},"Default",{"type":45,"tag":494,"props":1437,"children":1438},{},[1439,1461],{"type":45,"tag":464,"props":1440,"children":1441},{},[1442,1451,1456],{"type":45,"tag":501,"props":1443,"children":1444},{},[1445],{"type":45,"tag":70,"props":1446,"children":1448},{"className":1447},[],[1449],{"type":51,"value":1450},"train.num_gpus",{"type":45,"tag":501,"props":1452,"children":1453},{},[1454],{"type":51,"value":1455},"Number of GPUs",{"type":45,"tag":501,"props":1457,"children":1458},{},[1459],{"type":51,"value":1460},"1",{"type":45,"tag":464,"props":1462,"children":1463},{},[1464,1473,1478],{"type":45,"tag":501,"props":1465,"children":1466},{},[1467],{"type":45,"tag":70,"props":1468,"children":1470},{"className":1469},[],[1471],{"type":51,"value":1472},"train.gpu_ids",{"type":45,"tag":501,"props":1474,"children":1475},{},[1476],{"type":51,"value":1477},"GPU device indices",{"type":45,"tag":501,"props":1479,"children":1480},{},[1481],{"type":45,"tag":758,"props":1482,"children":1483},{},[1484],{"type":51,"value":1485},"0",{"type":45,"tag":413,"props":1487,"children":1488},{},[1489,1502,1523],{"type":45,"tag":417,"props":1490,"children":1491},{},[1492,1494,1500],{"type":51,"value":1493},"Strategy: ",{"type":45,"tag":70,"props":1495,"children":1497},{"className":1496},[],[1498],{"type":51,"value":1499},"auto",{"type":51,"value":1501}," (Lightning picks best strategy automatically)",{"type":45,"tag":417,"props":1503,"children":1504},{},[1505,1507,1513,1515,1521],{"type":51,"value":1506},"No explicit ",{"type":45,"tag":70,"props":1508,"children":1510},{"className":1509},[],[1511],{"type":51,"value":1512},"num_nodes",{"type":51,"value":1514}," or ",{"type":45,"tag":70,"props":1516,"children":1518},{"className":1517},[],[1519],{"type":51,"value":1520},"distributed_strategy",{"type":51,"value":1522}," config — single-node only",{"type":45,"tag":417,"props":1524,"children":1525},{},[1526],{"type":51,"value":1527},"Lightweight model, single GPU typically sufficient",{"type":45,"tag":163,"props":1529,"children":1531},{"id":1530},"hardware",[1532],{"type":51,"value":1533},"Hardware",{"type":45,"tag":54,"props":1535,"children":1536},{},[1537],{"type":51,"value":1538},"Minimum 1 GPU(s), recommended 1 GPU(s). 8GB+ VRAM per GPU. Pose classification is very lightweight — skeleton data is small. Single GPU is sufficient.",{"type":45,"tag":163,"props":1540,"children":1542},{"id":1541},"error-patterns",[1543],{"type":51,"value":1544},"Error Patterns",{"type":45,"tag":54,"props":1546,"children":1547},{},[1548,1553],{"type":45,"tag":421,"props":1549,"children":1550},{},[1551],{"type":51,"value":1552},"Graph layout mismatch",{"type":51,"value":1554},": Ensure model.graph_layout matches the skeleton format in your .npy data files.",{"type":45,"tag":54,"props":1556,"children":1557},{},[1558,1563],{"type":45,"tag":421,"props":1559,"children":1560},{},[1561],{"type":51,"value":1562},"Label shape mismatch",{"type":51,"value":1564},": train_label.pkl class indices must be in range [0, num_classes).",{"type":45,"tag":54,"props":1566,"children":1567},{},[1568,1573,1575,1581,1583,1589,1591,1597],{"type":45,"tag":421,"props":1569,"children":1570},{},[1571],{"type":51,"value":1572},"Missing label map",{"type":51,"value":1574},": The training dataloader expects ",{"type":45,"tag":70,"props":1576,"children":1578},{"className":1577},[],[1579],{"type":51,"value":1580},"dataset.label_map",{"type":51,"value":1582}," to be a dictionary. If the dataset only supplies numeric class IDs, set a synthetic contiguous map such as ",{"type":45,"tag":70,"props":1584,"children":1586},{"className":1585},[],[1587],{"type":51,"value":1588},"class_0: 0",{"type":51,"value":1590}," through ",{"type":45,"tag":70,"props":1592,"children":1594},{"className":1593},[],[1595],{"type":51,"value":1596},"class_5: 5",{"type":51,"value":1598}," for the six-class NVIDIA sample data.",{"type":45,"tag":54,"props":1600,"children":1601},{},[1602,1607,1609,1615,1617,1623,1625,1631,1632,1638,1639,1645,1647,1652,1653,1659,1661,1666,1667,1672,1673,1678],{"type":45,"tag":421,"props":1603,"children":1604},{},[1605],{"type":51,"value":1606},"Checkpoint handoff",{"type":51,"value":1608},": After AutoML\u002Ftrain, use the checkpoint resolver to select the intended saved ",{"type":45,"tag":70,"props":1610,"children":1612},{"className":1611},[],[1613],{"type":51,"value":1614},".pth",{"type":51,"value":1616}," checkpoint under the parent result folder, such as ",{"type":45,"tag":70,"props":1618,"children":1620},{"className":1619},[],[1621],{"type":51,"value":1622},"model_epoch_000_step_00007.pth",{"type":51,"value":1624},", and pass that exact file as ",{"type":45,"tag":70,"props":1626,"children":1628},{"className":1627},[],[1629],{"type":51,"value":1630},"evaluate.checkpoint",{"type":51,"value":77},{"type":45,"tag":70,"props":1633,"children":1635},{"className":1634},[],[1636],{"type":51,"value":1637},"export.checkpoint",{"type":51,"value":77},{"type":45,"tag":70,"props":1640,"children":1642},{"className":1641},[],[1643],{"type":51,"value":1644},"inference.checkpoint",{"type":51,"value":1646},", or ",{"type":45,"tag":70,"props":1648,"children":1650},{"className":1649},[],[1651],{"type":51,"value":159},{"type":51,"value":107},{"type":45,"tag":70,"props":1654,"children":1656},{"className":1655},[],[1657],{"type":51,"value":1658},"pc_model_latest.pth",{"type":51,"value":1660}," is a latest-checkpoint symlink; use it only when the user explicitly asks for latest rather than a specific\u002Fbest checkpoint. Keep the same ",{"type":45,"tag":70,"props":1662,"children":1664},{"className":1663},[],[1665],{"type":51,"value":1347},{"type":51,"value":77},{"type":45,"tag":70,"props":1668,"children":1670},{"className":1669},[],[1671],{"type":51,"value":1580},{"type":51,"value":99},{"type":45,"tag":70,"props":1674,"children":1676},{"className":1675},[],[1677],{"type":51,"value":1357},{"type":51,"value":1679}," overrides for downstream actions.",{"type":45,"tag":54,"props":1681,"children":1682},{},[1683,1688,1690,1695,1697,1702,1703,1708,1709,1714,1715,1720,1721,1726,1727,1732],{"type":45,"tag":421,"props":1684,"children":1685},{},[1686],{"type":51,"value":1687},"Dataset conversion source",{"type":51,"value":1689},": ",{"type":45,"tag":70,"props":1691,"children":1693},{"className":1692},[],[1694],{"type":51,"value":75},{"type":51,"value":1696}," expects the raw JSON output from the DeepStream BodyPose app. The common NVIDIA sample S3 folder is already converted to ",{"type":45,"tag":70,"props":1698,"children":1700},{"className":1699},[],[1701],{"type":51,"value":622},{"type":51,"value":77},{"type":45,"tag":70,"props":1704,"children":1706},{"className":1705},[],[1707],{"type":51,"value":647},{"type":51,"value":77},{"type":45,"tag":70,"props":1710,"children":1712},{"className":1711},[],[1713],{"type":51,"value":547},{"type":51,"value":77},{"type":45,"tag":70,"props":1716,"children":1718},{"className":1717},[],[1719],{"type":51,"value":572},{"type":51,"value":77},{"type":45,"tag":70,"props":1722,"children":1724},{"className":1723},[],[1725],{"type":51,"value":597},{"type":51,"value":99},{"type":45,"tag":70,"props":1728,"children":1730},{"className":1729},[],[1731],{"type":51,"value":1302},{"type":51,"value":1733},"; skip conversion and start from the converted files when those are present.",{"type":45,"tag":54,"props":1735,"children":1736},{},[1737,1742,1744,1750,1751,1757,1759,1764,1765,1770,1772,1777,1779,1785,1787,1793,1794,1800],{"type":45,"tag":421,"props":1738,"children":1739},{},[1740],{"type":51,"value":1741},"Action-specific dataset paths",{"type":51,"value":1743},": The evaluate and inference templates also contain the training ",{"type":45,"tag":70,"props":1745,"children":1747},{"className":1746},[],[1748],{"type":51,"value":1749},"dataset.train_dataset",{"type":51,"value":229},{"type":45,"tag":70,"props":1752,"children":1754},{"className":1753},[],[1755],{"type":51,"value":1756},"dataset.val_dataset",{"type":51,"value":1758}," blocks. For evaluate, populate ",{"type":45,"tag":70,"props":1760,"children":1762},{"className":1761},[],[1763],{"type":51,"value":537},{"type":51,"value":229},{"type":45,"tag":70,"props":1766,"children":1768},{"className":1767},[],[1769],{"type":51,"value":563},{"type":51,"value":1771},". For inference, populate ",{"type":45,"tag":70,"props":1773,"children":1775},{"className":1774},[],[1776],{"type":51,"value":588},{"type":51,"value":1778}," and set ",{"type":45,"tag":70,"props":1780,"children":1782},{"className":1781},[],[1783],{"type":51,"value":1784},"inference.output_file",{"type":51,"value":1786},"; do not stop after replacing the first ",{"type":45,"tag":70,"props":1788,"children":1790},{"className":1789},[],[1791],{"type":51,"value":1792},"data_path",{"type":51,"value":1514},{"type":45,"tag":70,"props":1795,"children":1797},{"className":1796},[],[1798],{"type":51,"value":1799},"label_path",{"type":51,"value":1801}," in the file.",{"type":45,"tag":54,"props":1803,"children":1804},{},[1805,1810,1812,1818,1820,1825],{"type":45,"tag":421,"props":1806,"children":1807},{},[1808],{"type":51,"value":1809},"Output files",{"type":51,"value":1811},": Export needs an explicit ",{"type":45,"tag":70,"props":1813,"children":1815},{"className":1814},[],[1816],{"type":51,"value":1817},"export.onnx_file",{"type":51,"value":1819}," path. Inference must set ",{"type":45,"tag":70,"props":1821,"children":1823},{"className":1822},[],[1824],{"type":51,"value":1784},{"type":51,"value":1826}," to a writable file path; the packaged template default is an empty string, and the current PyTorch inference code opens that value directly.",{"type":45,"tag":163,"props":1828,"children":1830},{"id":1829},"spec-param-parent-model-inference",[1831],{"type":51,"value":1832},"Spec Param \u002F Parent Model Inference",{"type":45,"tag":54,"props":1834,"children":1835},{},[1836,1838,1844,1846,1852,1854,1860],{"type":51,"value":1837},"Model-specific inference mappings belong in this MD file, not in ",{"type":45,"tag":70,"props":1839,"children":1841},{"className":1840},[],[1842],{"type":51,"value":1843},"config.json",{"type":51,"value":1845},". Generated runners should read this section and apply the mappings with SDK helpers before ",{"type":45,"tag":70,"props":1847,"children":1849},{"className":1848},[],[1850],{"type":51,"value":1851},"create_job()",{"type":51,"value":1853},". This mirrors the old microservices ",{"type":45,"tag":70,"props":1855,"children":1857},{"className":1856},[],[1858],{"type":51,"value":1859},"infer_params.py",{"type":51,"value":1861}," flow.",{"type":45,"tag":54,"props":1863,"children":1864},{},[1865,1867,1873],{"type":51,"value":1866},"Inference mappings from TAO Core ",{"type":45,"tag":70,"props":1868,"children":1870},{"className":1869},[],[1871],{"type":51,"value":1872},"pose_classification.config.json",{"type":51,"value":1874},":",{"type":45,"tag":456,"props":1876,"children":1877},{},[1878,1903],{"type":45,"tag":460,"props":1879,"children":1880},{},[1881],{"type":45,"tag":464,"props":1882,"children":1883},{},[1884,1888,1893,1898],{"type":45,"tag":468,"props":1885,"children":1886},{},[1887],{"type":51,"value":472},{"type":45,"tag":468,"props":1889,"children":1890},{},[1891],{"type":51,"value":1892},"Spec Field",{"type":45,"tag":468,"props":1894,"children":1895},{},[1896],{"type":51,"value":1897},"Inference Function",{"type":45,"tag":468,"props":1899,"children":1900},{},[1901],{"type":51,"value":1902},"Meaning",{"type":45,"tag":494,"props":1904,"children":1905},{},[1906,1936,1966,1995,2023,2050,2077,2106,2133,2160,2187,2216,2243,2270,2300,2327],{"type":45,"tag":464,"props":1907,"children":1908},{},[1909,1913,1922,1931],{"type":45,"tag":501,"props":1910,"children":1911},{},[1912],{"type":51,"value":75},{"type":45,"tag":501,"props":1914,"children":1915},{},[1916],{"type":45,"tag":70,"props":1917,"children":1919},{"className":1918},[],[1920],{"type":51,"value":1921},"dataset_convert.results_dir",{"type":45,"tag":501,"props":1923,"children":1924},{},[1925],{"type":45,"tag":70,"props":1926,"children":1928},{"className":1927},[],[1929],{"type":51,"value":1930},"output_dir",{"type":45,"tag":501,"props":1932,"children":1933},{},[1934],{"type":51,"value":1935},"current job results directory",{"type":45,"tag":464,"props":1937,"children":1938},{},[1939,1943,1952,1961],{"type":45,"tag":501,"props":1940,"children":1941},{},[1942],{"type":51,"value":90},{"type":45,"tag":501,"props":1944,"children":1945},{},[1946],{"type":45,"tag":70,"props":1947,"children":1949},{"className":1948},[],[1950],{"type":51,"value":1951},"encryption_key",{"type":45,"tag":501,"props":1953,"children":1954},{},[1955],{"type":45,"tag":70,"props":1956,"children":1958},{"className":1957},[],[1959],{"type":51,"value":1960},"key",{"type":45,"tag":501,"props":1962,"children":1963},{},[1964],{"type":51,"value":1965},"encryption key",{"type":45,"tag":464,"props":1967,"children":1968},{},[1969,1973,1981,1990],{"type":45,"tag":501,"props":1970,"children":1971},{},[1972],{"type":51,"value":90},{"type":45,"tag":501,"props":1974,"children":1975},{},[1976],{"type":45,"tag":70,"props":1977,"children":1979},{"className":1978},[],[1980],{"type":51,"value":1630},{"type":45,"tag":501,"props":1982,"children":1983},{},[1984],{"type":45,"tag":70,"props":1985,"children":1987},{"className":1986},[],[1988],{"type":51,"value":1989},"parent_model",{"type":45,"tag":501,"props":1991,"children":1992},{},[1993],{"type":51,"value":1994},"model file inferred from the parent job results folder",{"type":45,"tag":464,"props":1996,"children":1997},{},[1998,2002,2011,2019],{"type":45,"tag":501,"props":1999,"children":2000},{},[2001],{"type":51,"value":90},{"type":45,"tag":501,"props":2003,"children":2004},{},[2005],{"type":45,"tag":70,"props":2006,"children":2008},{"className":2007},[],[2009],{"type":51,"value":2010},"results_dir",{"type":45,"tag":501,"props":2012,"children":2013},{},[2014],{"type":45,"tag":70,"props":2015,"children":2017},{"className":2016},[],[2018],{"type":51,"value":1930},{"type":45,"tag":501,"props":2020,"children":2021},{},[2022],{"type":51,"value":1935},{"type":45,"tag":464,"props":2024,"children":2025},{},[2026,2030,2038,2046],{"type":45,"tag":501,"props":2027,"children":2028},{},[2029],{"type":51,"value":97},{"type":45,"tag":501,"props":2031,"children":2032},{},[2033],{"type":45,"tag":70,"props":2034,"children":2036},{"className":2035},[],[2037],{"type":51,"value":1951},{"type":45,"tag":501,"props":2039,"children":2040},{},[2041],{"type":45,"tag":70,"props":2042,"children":2044},{"className":2043},[],[2045],{"type":51,"value":1960},{"type":45,"tag":501,"props":2047,"children":2048},{},[2049],{"type":51,"value":1965},{"type":45,"tag":464,"props":2051,"children":2052},{},[2053,2057,2065,2073],{"type":45,"tag":501,"props":2054,"children":2055},{},[2056],{"type":51,"value":97},{"type":45,"tag":501,"props":2058,"children":2059},{},[2060],{"type":45,"tag":70,"props":2061,"children":2063},{"className":2062},[],[2064],{"type":51,"value":1637},{"type":45,"tag":501,"props":2066,"children":2067},{},[2068],{"type":45,"tag":70,"props":2069,"children":2071},{"className":2070},[],[2072],{"type":51,"value":1989},{"type":45,"tag":501,"props":2074,"children":2075},{},[2076],{"type":51,"value":1994},{"type":45,"tag":464,"props":2078,"children":2079},{},[2080,2084,2092,2101],{"type":45,"tag":501,"props":2081,"children":2082},{},[2083],{"type":51,"value":97},{"type":45,"tag":501,"props":2085,"children":2086},{},[2087],{"type":45,"tag":70,"props":2088,"children":2090},{"className":2089},[],[2091],{"type":51,"value":1817},{"type":45,"tag":501,"props":2093,"children":2094},{},[2095],{"type":45,"tag":70,"props":2096,"children":2098},{"className":2097},[],[2099],{"type":51,"value":2100},"create_onnx_file",{"type":45,"tag":501,"props":2102,"children":2103},{},[2104],{"type":51,"value":2105},"output ONNX path",{"type":45,"tag":464,"props":2107,"children":2108},{},[2109,2113,2121,2129],{"type":45,"tag":501,"props":2110,"children":2111},{},[2112],{"type":51,"value":97},{"type":45,"tag":501,"props":2114,"children":2115},{},[2116],{"type":45,"tag":70,"props":2117,"children":2119},{"className":2118},[],[2120],{"type":51,"value":2010},{"type":45,"tag":501,"props":2122,"children":2123},{},[2124],{"type":45,"tag":70,"props":2125,"children":2127},{"className":2126},[],[2128],{"type":51,"value":1930},{"type":45,"tag":501,"props":2130,"children":2131},{},[2132],{"type":51,"value":1935},{"type":45,"tag":464,"props":2134,"children":2135},{},[2136,2140,2148,2156],{"type":45,"tag":501,"props":2137,"children":2138},{},[2139],{"type":51,"value":105},{"type":45,"tag":501,"props":2141,"children":2142},{},[2143],{"type":45,"tag":70,"props":2144,"children":2146},{"className":2145},[],[2147],{"type":51,"value":1951},{"type":45,"tag":501,"props":2149,"children":2150},{},[2151],{"type":45,"tag":70,"props":2152,"children":2154},{"className":2153},[],[2155],{"type":51,"value":1960},{"type":45,"tag":501,"props":2157,"children":2158},{},[2159],{"type":51,"value":1965},{"type":45,"tag":464,"props":2161,"children":2162},{},[2163,2167,2175,2183],{"type":45,"tag":501,"props":2164,"children":2165},{},[2166],{"type":51,"value":105},{"type":45,"tag":501,"props":2168,"children":2169},{},[2170],{"type":45,"tag":70,"props":2171,"children":2173},{"className":2172},[],[2174],{"type":51,"value":1644},{"type":45,"tag":501,"props":2176,"children":2177},{},[2178],{"type":45,"tag":70,"props":2179,"children":2181},{"className":2180},[],[2182],{"type":51,"value":1989},{"type":45,"tag":501,"props":2184,"children":2185},{},[2186],{"type":51,"value":1994},{"type":45,"tag":464,"props":2188,"children":2189},{},[2190,2194,2202,2211],{"type":45,"tag":501,"props":2191,"children":2192},{},[2193],{"type":51,"value":105},{"type":45,"tag":501,"props":2195,"children":2196},{},[2197],{"type":45,"tag":70,"props":2198,"children":2200},{"className":2199},[],[2201],{"type":51,"value":1784},{"type":45,"tag":501,"props":2203,"children":2204},{},[2205],{"type":45,"tag":70,"props":2206,"children":2208},{"className":2207},[],[2209],{"type":51,"value":2210},"create_inference_result_file_pose",{"type":45,"tag":501,"props":2212,"children":2213},{},[2214],{"type":51,"value":2215},"pose inference result file",{"type":45,"tag":464,"props":2217,"children":2218},{},[2219,2223,2231,2239],{"type":45,"tag":501,"props":2220,"children":2221},{},[2222],{"type":51,"value":105},{"type":45,"tag":501,"props":2224,"children":2225},{},[2226],{"type":45,"tag":70,"props":2227,"children":2229},{"className":2228},[],[2230],{"type":51,"value":2010},{"type":45,"tag":501,"props":2232,"children":2233},{},[2234],{"type":45,"tag":70,"props":2235,"children":2237},{"className":2236},[],[2238],{"type":51,"value":1930},{"type":45,"tag":501,"props":2240,"children":2241},{},[2242],{"type":51,"value":1935},{"type":45,"tag":464,"props":2244,"children":2245},{},[2246,2250,2258,2266],{"type":45,"tag":501,"props":2247,"children":2248},{},[2249],{"type":51,"value":83},{"type":45,"tag":501,"props":2251,"children":2252},{},[2253],{"type":45,"tag":70,"props":2254,"children":2256},{"className":2255},[],[2257],{"type":51,"value":1951},{"type":45,"tag":501,"props":2259,"children":2260},{},[2261],{"type":45,"tag":70,"props":2262,"children":2264},{"className":2263},[],[2265],{"type":51,"value":1960},{"type":45,"tag":501,"props":2267,"children":2268},{},[2269],{"type":51,"value":1965},{"type":45,"tag":464,"props":2271,"children":2272},{},[2273,2277,2286,2295],{"type":45,"tag":501,"props":2274,"children":2275},{},[2276],{"type":51,"value":83},{"type":45,"tag":501,"props":2278,"children":2279},{},[2280],{"type":45,"tag":70,"props":2281,"children":2283},{"className":2282},[],[2284],{"type":51,"value":2285},"model.pretrained_model_path",{"type":45,"tag":501,"props":2287,"children":2288},{},[2289],{"type":45,"tag":70,"props":2290,"children":2292},{"className":2291},[],[2293],{"type":51,"value":2294},"ptm_if_no_resume_model",{"type":45,"tag":501,"props":2296,"children":2297},{},[2298],{"type":51,"value":2299},"PTM when no resume checkpoint exists",{"type":45,"tag":464,"props":2301,"children":2302},{},[2303,2307,2315,2323],{"type":45,"tag":501,"props":2304,"children":2305},{},[2306],{"type":51,"value":83},{"type":45,"tag":501,"props":2308,"children":2309},{},[2310],{"type":45,"tag":70,"props":2311,"children":2313},{"className":2312},[],[2314],{"type":51,"value":2010},{"type":45,"tag":501,"props":2316,"children":2317},{},[2318],{"type":45,"tag":70,"props":2319,"children":2321},{"className":2320},[],[2322],{"type":51,"value":1930},{"type":45,"tag":501,"props":2324,"children":2325},{},[2326],{"type":51,"value":1935},{"type":45,"tag":464,"props":2328,"children":2329},{},[2330,2334,2342,2351],{"type":45,"tag":501,"props":2331,"children":2332},{},[2333],{"type":51,"value":83},{"type":45,"tag":501,"props":2335,"children":2336},{},[2337],{"type":45,"tag":70,"props":2338,"children":2340},{"className":2339},[],[2341],{"type":51,"value":159},{"type":45,"tag":501,"props":2343,"children":2344},{},[2345],{"type":45,"tag":70,"props":2346,"children":2348},{"className":2347},[],[2349],{"type":51,"value":2350},"resume_model",{"type":45,"tag":501,"props":2352,"children":2353},{},[2354],{"type":51,"value":2355},"model file inferred from the current job results folder",{"type":45,"tag":54,"props":2357,"children":2358},{},[2359,2361,2366,2367,2373,2375,2381,2383,2388],{"type":51,"value":2360},"For ",{"type":45,"tag":70,"props":2362,"children":2364},{"className":2363},[],[2365],{"type":51,"value":1989},{"type":51,"value":1514},{"type":45,"tag":70,"props":2368,"children":2370},{"className":2369},[],[2371],{"type":51,"value":2372},"parent_model_folder",{"type":51,"value":2374},", pass the upstream train\u002Fexport\u002FAutoML child job id as ",{"type":45,"tag":70,"props":2376,"children":2378},{"className":2377},[],[2379],{"type":51,"value":2380},"parent_job_id",{"type":51,"value":2382},". 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 ",{"type":45,"tag":70,"props":2384,"children":2386},{"className":2385},[],[2387],{"type":51,"value":1843},{"type":51,"value":2389}," and do not patch generated runner scripts to guess checkpoint paths.",{"type":45,"tag":2391,"props":2392,"children":2393},"style",{},[2394],{"type":51,"value":2395},"html .light .shiki span {color: var(--shiki-light);background: var(--shiki-light-bg);font-style: var(--shiki-light-font-style);font-weight: var(--shiki-light-font-weight);text-decoration: var(--shiki-light-text-decoration);}html.light .shiki span {color: var(--shiki-light);background: var(--shiki-light-bg);font-style: var(--shiki-light-font-style);font-weight: var(--shiki-light-font-weight);text-decoration: var(--shiki-light-text-decoration);}html .default .shiki span {color: var(--shiki-default);background: var(--shiki-default-bg);font-style: var(--shiki-default-font-style);font-weight: var(--shiki-default-font-weight);text-decoration: var(--shiki-default-text-decoration);}html .shiki span {color: var(--shiki-default);background: var(--shiki-default-bg);font-style: var(--shiki-default-font-style);font-weight: var(--shiki-default-font-weight);text-decoration: var(--shiki-default-text-decoration);}html .dark .shiki span {color: var(--shiki-dark);background: var(--shiki-dark-bg);font-style: var(--shiki-dark-font-style);font-weight: var(--shiki-dark-font-weight);text-decoration: var(--shiki-dark-text-decoration);}html.dark .shiki span {color: var(--shiki-dark);background: var(--shiki-dark-bg);font-style: var(--shiki-dark-font-style);font-weight: var(--shiki-dark-font-weight);text-decoration: var(--shiki-dark-text-decoration);}",{"items":2397,"total":2501},[2398,2415,2429,2443,2455,2472,2487],{"slug":2399,"name":2399,"fn":2400,"description":2401,"org":2402,"tags":2403,"stars":20,"repoUrl":21,"updatedAt":2414},"accelerated-computing-cudf","accelerate data processing with cuDF","Official NVIDIA-authored guidance for NVIDIA cuDF GPU DataFrames, pandas acceleration, dask-cuDF, ETL, joins, groupby, CSV\u002FParquet I\u002FO, nullable semantics, and multi-GPU DataFrame workloads.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[2404,2407,2410,2411],{"name":2405,"slug":2406,"type":15},"Data Analysis","data-analysis",{"name":2408,"slug":2409,"type":15},"Data Engineering","data-engineering",{"name":9,"slug":8,"type":15},{"name":2412,"slug":2413,"type":15},"Performance","performance","2026-07-14T05:28:43.176466",{"slug":2416,"name":2416,"fn":2417,"description":2418,"org":2419,"tags":2420,"stars":20,"repoUrl":21,"updatedAt":2428},"aiq-deploy","deploy and manage NVIDIA AI-Q infrastructure","Use when asked to install, deploy, run, validate, troubleshoot, or stop NVIDIA AI-Q Blueprint infrastructure.\n",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[2421,2424,2427],{"name":2422,"slug":2423,"type":15},"Deployment","deployment",{"name":2425,"slug":2426,"type":15},"Infrastructure","infrastructure",{"name":9,"slug":8,"type":15},"2026-07-14T05:29:06.667109",{"slug":2430,"name":2430,"fn":2431,"description":2432,"org":2433,"tags":2434,"stars":20,"repoUrl":21,"updatedAt":2442},"aiq-research","conduct deep research with AI-Q","Use when asked to run deep research or AI-Q research through a reachable NVIDIA AI-Q Blueprint backend.\n",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[2435,2438,2439],{"name":2436,"slug":2437,"type":15},"Agents","agents",{"name":9,"slug":8,"type":15},{"name":2440,"slug":2441,"type":15},"Research","research","2026-07-14T05:28:06.816956",{"slug":2444,"name":2444,"fn":2445,"description":2446,"org":2447,"tags":2448,"stars":20,"repoUrl":21,"updatedAt":2454},"amc-run-sample-calibration","run AMC sample dataset calibration","Run end-to-end calibration on the shipped sample dataset (sdg_08_2_sample_data_010926.zip) against a running AMC microservice. Use when user says 'test sample dataset', 'run sample calibration', 'verify AMC install', or 'launch and test'.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[2449,2450,2451],{"name":2405,"slug":2406,"type":15},{"name":9,"slug":8,"type":15},{"name":2452,"slug":2453,"type":15},"Testing","testing","2026-07-17T05:29:03.913266",{"slug":2456,"name":2456,"fn":2457,"description":2458,"org":2459,"tags":2460,"stars":20,"repoUrl":21,"updatedAt":2471},"amc-run-video-calibration","calibrate video datasets with AutoMagicCalib","Calibrate a new dataset from pre-recorded video files via the AutoMagicCalib REST API. Use when user has local MP4s and says 'calibrate my videos', 'run AMC on these videos', or similar. For RTSP\u002Flive streams, use amc-run-rtsp-calibration instead.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[2461,2464,2467,2468],{"name":2462,"slug":2463,"type":15},"Automation","automation",{"name":2465,"slug":2466,"type":15},"Imaging","imaging",{"name":9,"slug":8,"type":15},{"name":2469,"slug":2470,"type":15},"Video","video","2026-07-17T05:28:53.905004",{"slug":2473,"name":2473,"fn":2474,"description":2475,"org":2476,"tags":2477,"stars":20,"repoUrl":21,"updatedAt":2486},"amc-setup-calibration-stack","deploy AutoMagicCalib microservice with Docker","Launch AutoMagicCalib microservice and web UI from NGC release images via Docker Compose. Use when user says 'deploy auto calibration', 'launch auto calibration', 'launch AMC', 'start MS+UI', or 'set up auto-magic-calib'. Requires NGC API key.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[2478,2479,2482,2483],{"name":2422,"slug":2423,"type":15},{"name":2480,"slug":2481,"type":15},"Docker","docker",{"name":9,"slug":8,"type":15},{"name":2484,"slug":2485,"type":15},"Operations","operations","2026-07-17T05:28:56.913999",{"slug":2488,"name":2488,"fn":2489,"description":2490,"org":2491,"tags":2492,"stars":20,"repoUrl":21,"updatedAt":2500},"cudaq-guide","develop quantum applications with CUDA-Q","CUDA-Q onboarding guide for installation, test programs, GPU simulation, QPU hardware, and quantum applications.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[2493,2494,2497],{"name":9,"slug":8,"type":15},{"name":2495,"slug":2496,"type":15},"Quantum Computing","quantum-computing",{"name":2498,"slug":2499,"type":15},"Simulation","simulation","2026-07-14T05:26:58.898253",305,{"items":2503,"total":2651},[2504,2522,2537,2548,2560,2574,2587,2599,2610,2619,2633,2642],{"slug":2505,"name":2505,"fn":2506,"description":2507,"org":2508,"tags":2509,"stars":2519,"repoUrl":2520,"updatedAt":2521},"nemoclaw-user-guide","retrieve NemoClaw documentation and configuration","Guides human users' AI agents to the NemoClaw docs MCP server and canonical Fern documentation in Markdown form. Use when users ask how to install, configure, operate, troubleshoot, secure, or learn NemoClaw with an AI coding assistant. Trigger keywords - nemoclaw docs, use nemoclaw with ai agent, nemoclaw mcp docs, nemoclaw install help, nemoclaw quickstart, nemoclaw markdown docs, llms.txt, agent skills.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[2510,2513,2516],{"name":2511,"slug":2512,"type":15},"Documentation","documentation",{"name":2514,"slug":2515,"type":15},"MCP","mcp",{"name":2517,"slug":2518,"type":15},"Search","search",21777,"https:\u002F\u002Fgithub.com\u002FNVIDIA\u002FNemoClaw","2026-07-20T06:00:01.461044",{"slug":2523,"name":2523,"fn":2524,"description":2525,"org":2526,"tags":2527,"stars":2534,"repoUrl":2535,"updatedAt":2536},"mcore-build-and-dependency","manage Megatron-LM development environments","Container-based dev environment setup and dependency management for Megatron-LM. Covers acquiring and launching the CI container, uv package management, and updating uv.lock.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[2528,2531,2532],{"name":2529,"slug":2530,"type":15},"Containers","containers",{"name":2422,"slug":2423,"type":15},{"name":2533,"slug":751,"type":15},"Python",17049,"https:\u002F\u002Fgithub.com\u002FNVIDIA\u002FMegatron-LM","2026-07-27T06:06:11.249662",{"slug":2538,"name":2538,"fn":2539,"description":2540,"org":2541,"tags":2542,"stars":2534,"repoUrl":2535,"updatedAt":2547},"mcore-bump-base-image","update NVIDIA PyTorch base images","Bump the NVIDIA PyTorch base image (`nvcr.io\u002Fnvidia\u002Fpytorch:YY.MM-py3`) used by Megatron-LM CI. Covers the two pin sites (GitHub CI in `docker\u002F.ngc_version.dev` and GitLab CI in `.gitlab\u002Fstages\u002F01.build.yml`), the post-bump CI loop (re-run functional tests, refresh golden values, mark broken tests), and the gotchas that bit PRs",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[2543,2546],{"name":2544,"slug":2545,"type":15},"CI\u002FCD","ci-cd",{"name":2422,"slug":2423,"type":15},"2026-07-14T05:25:59.97109",{"slug":2549,"name":2549,"fn":2550,"description":2551,"org":2552,"tags":2553,"stars":2534,"repoUrl":2535,"updatedAt":2559},"mcore-cicd","manage CI\u002FCD pipelines for Megatron-LM","CI\u002FCD reference for Megatron-LM. Covers CI pipeline structure, PR scope labels, triggering internal GitLab CI (which force-pushes the current branch to a pull-request\u002FBRANCH ref — always dry-run and verify the destination first; never run against shared or protected branches), and CI failure investigation.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[2554,2555,2556],{"name":2544,"slug":2545,"type":15},{"name":2422,"slug":2423,"type":15},{"name":2557,"slug":2558,"type":15},"GitHub","github","2026-07-27T06:06:12.278222",{"slug":2561,"name":2561,"fn":2562,"description":2563,"org":2564,"tags":2565,"stars":2534,"repoUrl":2535,"updatedAt":2573},"mcore-create-issue","investigate CI failures and create issues","Investigate a failing GitHub Actions run or job and create a GitHub issue for the failure.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[2566,2569,2570],{"name":2567,"slug":2568,"type":15},"Debugging","debugging",{"name":2557,"slug":2558,"type":15},{"name":2571,"slug":2572,"type":15},"Triage","triage","2026-07-14T05:25:57.442089",{"slug":2575,"name":2575,"fn":2576,"description":2577,"org":2578,"tags":2579,"stars":2534,"repoUrl":2535,"updatedAt":2586},"mcore-linting-and-formatting","lint and format Megatron-LM code","Linting and formatting for Megatron-LM. Covers running autoformat.sh, tools (ruff, black, isort, pylint, mypy), and code style rules.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[2580,2583],{"name":2581,"slug":2582,"type":15},"Best Practices","best-practices",{"name":2584,"slug":2585,"type":15},"Code Analysis","code-analysis","2026-07-14T05:25:56.18433",{"slug":2588,"name":2588,"fn":2589,"description":2590,"org":2591,"tags":2592,"stars":2534,"repoUrl":2535,"updatedAt":2598},"mcore-migrate-gpt-to-hybrid","migrate Megatron-LM models to HybridModel","Migration guide for moving Megatron Core GPTModel checkpoints, model providers, training commands, and layer mappings to HybridModel.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[2593,2594,2597],{"name":13,"slug":14,"type":15},{"name":2595,"slug":2596,"type":15},"Migration","migration",{"name":9,"slug":8,"type":15},"2026-07-17T06:07:11.777011",{"slug":2600,"name":2600,"fn":2601,"description":2602,"org":2603,"tags":2604,"stars":2534,"repoUrl":2535,"updatedAt":2609},"mcore-onboard-gb200-1node-tests","onboard functional tests for GB200","Onboard 1-node GitHub MR functional tests for GB200 from existing mr-scoped 2-node tests.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[2605,2608],{"name":2606,"slug":2607,"type":15},"QA","qa",{"name":2452,"slug":2453,"type":15},"2026-07-14T05:25:53.673039",{"slug":2611,"name":2611,"fn":2612,"description":2613,"org":2614,"tags":2615,"stars":2534,"repoUrl":2535,"updatedAt":2618},"mcore-run-on-slurm","launch distributed training jobs on SLURM","How to launch distributed Megatron-LM training jobs on a SLURM cluster. Covers a minimal sbatch skeleton, environment-variable setup for torch.distributed.run, CUDA_DEVICE_MAX_CONNECTIONS rules across hardware and parallelism modes, container conventions, monitoring, and per-rank failure diagnosis.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[2616,2617],{"name":2422,"slug":2423,"type":15},{"name":2425,"slug":2426,"type":15},"2026-07-14T05:25:49.362534",{"slug":2620,"name":2620,"fn":2621,"description":2622,"org":2623,"tags":2624,"stars":2534,"repoUrl":2535,"updatedAt":2632},"mcore-split-pr","split pull requests to reduce review load","Split a PR into multiple PRs to reduce the number of required CODEOWNERS reviewer groups.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[2625,2628,2629],{"name":2626,"slug":2627,"type":15},"Code Review","code-review",{"name":2557,"slug":2558,"type":15},{"name":2630,"slug":2631,"type":15},"Pull Requests","pull-requests","2026-07-14T05:26:01.226578",{"slug":2634,"name":2634,"fn":2635,"description":2636,"org":2637,"tags":2638,"stars":2534,"repoUrl":2535,"updatedAt":2641},"mcore-testing","run and manage Megatron-LM tests","Test system for Megatron-LM. Covers test layout, recipe YAML structure, adding and running unit and functional tests, golden values, marker filters, and CI parity.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[2639,2640],{"name":2606,"slug":2607,"type":15},{"name":2452,"slug":2453,"type":15},"2026-07-14T05:25:54.928983",{"slug":2643,"name":2643,"fn":2644,"description":2645,"org":2646,"tags":2647,"stars":2534,"repoUrl":2535,"updatedAt":2650},"nightly-sync","manage nightly main-to-dev sync workflows","Domain knowledge for the nightly main-to-dev sync workflow. Covers merge strategy, CI architecture, failure investigation, and known issues.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[2648,2649],{"name":2462,"slug":2463,"type":15},{"name":2544,"slug":2545,"type":15},"2026-07-30T05:29:03.275638",496]