[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill-nvidia-tao-train-optical-inspection":3,"mdc-wmgnx1-key":34,"related-repo-nvidia-tao-train-optical-inspection":2171,"related-org-nvidia-tao-train-optical-inspection":2275},{"slug":4,"name":4,"fn":5,"description":6,"org":7,"tags":11,"stars":23,"repoUrl":24,"updatedAt":25,"license":26,"forks":27,"topics":28,"repo":29,"sourceUrl":32,"mdContent":33},"tao-train-optical-inspection","detect manufacturing defects with Siamese networks","Optical Inspection for defect detection using Siamese networks. Compares image pairs to detect manufacturing defects, anomalies, or quality issues. Use when training, evaluating, exporting, or running inference for a TAO Optical Inspection model on AOI \u002F quality-control data. Trigger phrases include \"train optical inspection\", \"AOI defect detection\", \"Siamese defect classifier\", \"PCB \u002F manufacturing inspection\".",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},"nvidia","NVIDIA","https:\u002F\u002Fpexgzepcugksgbtrxkhf.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Forg-logos\u002Fnvidia.png",[12,16,19,20],{"name":13,"slug":14,"type":15},"Deep Learning","deep-learning","tag",{"name":17,"slug":18,"type":15},"Machine Learning","machine-learning",{"name":9,"slug":8,"type":15},{"name":21,"slug":22,"type":15},"Computer Vision","computer-vision",2473,"https:\u002F\u002Fgithub.com\u002FNVIDIA\u002Fskills","2026-07-14T05:26:40.320013","Apache-2.0",281,[],{"repoUrl":24,"stars":23,"forks":27,"topics":30,"description":31},[],"AI agent skills published by NVIDIA","https:\u002F\u002Fgithub.com\u002FNVIDIA\u002Fskills\u002Ftree\u002FHEAD\u002Fskills\u002Ftao-train-optical-inspection","---\nname: tao-train-optical-inspection\ndescription: Optical Inspection for defect detection using Siamese networks. Compares image pairs to detect manufacturing\n  defects, anomalies, or quality issues. Use when training, evaluating, exporting, or running inference for a TAO Optical\n  Inspection model on AOI \u002F quality-control data. Trigger phrases include \"train optical inspection\", \"AOI defect\n  detection\", \"Siamese defect classifier\", \"PCB \u002F manufacturing inspection\".\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- defect\n- detection\n---\n\n# Optical Inspection\n\nOptical inspection for defect detection using Siamese networks. Compares image pairs to detect manufacturing defects, anomalies, or quality issues.\n\nSet train.pretrained_model_path for pretrained Siamese weights.\n\nFor TAO Deploy TensorRT actions (`gen_trt_engine`, TensorRT `evaluate`, and TensorRT `inference`), read `references\u002Ftao-deploy-optical-inspection.md` first. The parent PyT container does not expose `optical_inspection gen_trt_engine`; TensorRT engine generation is deploy-only. Deploy spec templates live in this skill's `references\u002F` folder with the `spec_template_deploy_*.yaml` prefix.\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:** optical_inspection\n- **Formats:** default\n- **Monitoring metric:** val_acc\n\n### Per-Action Dataset Requirements\n\n| Action | Spec Key | Source | Files | List? |\n|---|---|---|---|---|\n| evaluate | dataset.test_dataset.images_dir | eval_dataset | images.tar.gz | No |\n| evaluate | dataset.test_dataset.csv_path | eval_dataset | dataset.csv | No |\n| inference | dataset.infer_dataset.images_dir | inference_dataset | images.tar.gz | No |\n| inference | dataset.infer_dataset.csv_path | inference_dataset | dataset.csv | No |\n| train | dataset.train_dataset.images_dir | train_datasets | images.tar.gz | No |\n| train | dataset.train_dataset.csv_path | train_datasets | dataset.csv | No |\n| train | dataset.validation_dataset.images_dir | eval_dataset | images.tar.gz | No |\n| train | dataset.validation_dataset.csv_path | eval_dataset | dataset.csv | No |\n| train | dataset.test_dataset.images_dir | eval_dataset | images.tar.gz | No |\n| train | dataset.test_dataset.csv_path | eval_dataset | dataset.csv | No |\n\n### Typical Spec Overrides\n\nData 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`.\n\n```python\nS3_TRAIN = \"s3:\u002F\u002Fbucket\u002Fdata\u002Ftrain\"\nS3_EVAL = \"s3:\u002F\u002Fbucket\u002Fdata\u002Feval\"\nS3_INFERENCE = \"s3:\u002F\u002Fbucket\u002Fdata\u002Finference\"\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    \"dataset.batch_size\": 8,\n    \"dataset.train_dataset.images_dir\": f\"{S3_TRAIN}\u002Fimages.tar.gz\",\n    \"dataset.train_dataset.csv_path\": f\"{S3_TRAIN}\u002Fdataset.csv\",\n    \"dataset.validation_dataset.images_dir\": f\"{S3_EVAL}\u002Fimages.tar.gz\",\n    \"dataset.validation_dataset.csv_path\": f\"{S3_EVAL}\u002Fdataset.csv\",\n    \"dataset.test_dataset.images_dir\": f\"{S3_EVAL}\u002Fimages.tar.gz\",\n    \"dataset.test_dataset.csv_path\": f\"{S3_EVAL}\u002Fdataset.csv\",\n}\n```\n\n**evaluate (mandatory data sources):**\n```python\n{\n    \"evaluate.checkpoint\": \"\u003Cselected train\u002FAutoML checkpoint>\",\n    \"dataset.test_dataset.images_dir\": f\"{S3_EVAL}\u002Fimages.tar.gz\",\n    \"dataset.test_dataset.csv_path\": f\"{S3_EVAL}\u002Fdataset.csv\",\n}\n```\n\nUse the workflow's checkpoint resolver for downstream actions instead of guessing a filename. For Optical Inspection smoke runs, AutoML may produce `model_epoch_000_step_00006.pth`; resume can then produce `model_epoch_001_step_00012.pth`. Best-checkpoint actions should use the AutoML best child job's selected checkpoint, epoch-specific actions should pass the exact epoch\u002Fstep checkpoint requested, and only explicit \"latest\" requests should resolve to the latest checkpoint.\n\n**export:**\n```python\n{\n    \"export.checkpoint\": \"\u003Cselected train\u002FAutoML checkpoint>\",\n    \"export.onnx_file\": \"\u002Fresults\u002Foptical_inspection.onnx\",\n    \"export.input_width\": 128,\n    \"export.input_height\": 512,\n    \"export.batch_size\": 1,\n}\n```\n\n**inference (mandatory data sources):**\n```python\n{\n    \"inference.checkpoint\": \"\u003Cselected train\u002FAutoML checkpoint>\",\n    \"dataset.infer_dataset.images_dir\": f\"{S3_INFERENCE}\u002Fimages.tar.gz\",\n    \"dataset.infer_dataset.csv_path\": f\"{S3_INFERENCE}\u002Fdataset.csv\",\n}\n```\n\n## Dataset Convert\n\nDataset conversion is optional for Optical Inspection. If the dataset is already in TAO-ready Optical Inspection format, start directly from the `images.tar.gz` plus `dataset.csv` splits and run `train`, `evaluate`, `inference`, and downstream checkpoint\u002Fexport\u002Fdeploy actions on that converted data.\n\nThe PyT container exposes `optical_inspection dataset_convert`, but this model skill does not package a `dataset_convert` action\u002Ftemplate. The converter expects the raw Factory PCB layout (`root_dataset_dir`, train\u002Fval\u002Fall PCB directories, `golden_csv_dir`, `project_name`, and `bot_top`). The S3 validation bucket currently contains preconverted Optical Inspection `images.tar.gz` plus `dataset.csv` splits, not the raw PCB\u002Fgolden CSV source. Do not synthesize a fake PCB dataset. In model validation reports, mark dataset conversion as `not run: preconverted dataset provided` rather than failed or blocked when only converted data is available.\n\nWhen using the preconverted S3 validation tarballs locally, verify the extracted directory before writing specs. The tarballs may unpack an `images\u002F` wrapper directory; point `dataset.*.images_dir` at the inner directory that contains `golden\u002F` and the board\u002Fimage folders referenced by `dataset.csv`, for example `...\u002F\u003Csplit>\u002Fimages\u002Fimages`, not the outer wrapper.\n\n## Eval Dataset\n\nOptional. Eval dataset uses same format (images + CSV).\n\n## Important Parameters\n\n- **model.model_type**: Siamese variant. Options include Siamese, Siamese_3.\n- **model.model_backbone**: Default custom.\n- **model.embedding_vectors**: Number of embedding dimensions. Default 5.\n- **train.optim.lr**: Learning rate. Default 5e-4.\n- **dataset.batch_size**: Training batch size. Must be greater than 1; use `2` or higher for minimal smoke runs.\n- **dataset.num_input**: Number of input images per comparison.\n- **dataset.input_map**: Mapping of input channels \u002F image pairs.\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 Siamese network, single GPU typically sufficient\n\n## Hardware\n\nMinimum 1 GPU(s), recommended 1 GPU(s). 8GB+ VRAM per GPU. Siamese networks for inspection are lightweight. Single GPU sufficient.\n\n## Error Patterns\n\n**CSV format error**: Ensure dataset.csv has the correct column format for image pair paths and labels.\n\n**Extracted image root mismatch**: If train, evaluate, or inference cannot find paths from `dataset.csv`, inspect the extracted `images.tar.gz` tree. The TAO-ready root must contain `golden\u002F` plus the board folders referenced in the CSV. For validation S3 tarballs this can be one level below the extraction target, such as `images\u002Fimages`.\n\n**Training batch size assertion**: The Optical Inspection dataloader rejects\n`dataset.batch_size: 1` for train. Keep the template default of 8 for normal\nruns, or set `dataset.batch_size: 2` for minimal AutoML smoke validation.\n\n**PyTorch checkpoint load failure on downstream actions**: For checkpoints\nproduced by the same trusted TAO train\u002FAutoML workflow, set\n`TORCH_FORCE_NO_WEIGHTS_ONLY_LOAD=1` for evaluate, inference, export, and\nresume jobs if the current PyTorch default blocks loading the full checkpoint.\nDo not use this env var for untrusted checkpoints.\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 `optical_inspection.config.json`:\n\n| Action | Spec Field | Inference Function | Meaning |\n|---|---|---|---|\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.trt_engine` | `parent_model` | model file inferred from the parent job results folder |\n| inference | `results_dir` | `output_dir` | current job results directory |\n| train | `encryption_key` | `key` | encryption key |\n| train | `results_dir` | `output_dir` | current job results directory |\n| train | `train.pretrained_model_path` | `ptm_if_no_resume_model` | PTM when no resume checkpoint exists |\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\n## Deployment\n\n- [tao-deploy-optical-inspection](references\u002Ftao-deploy-optical-inspection.md)\n",{"data":35,"body":44},{"name":4,"description":6,"license":26,"compatibility":36,"metadata":37,"allowed-tools":40,"tags":41},"Requires docker + nvidia-container-toolkit.",{"version":38,"author":39},"0.1.0","NVIDIA Corporation","Read Bash",[42,43],"defect","detection",{"type":45,"children":46},"root",[47,56,62,67,129,136,229,235,344,376,382,418,425,713,719,739,779,787,908,916,959,980,988,1049,1057,1102,1108,1146,1219,1263,1269,1274,1280,1361,1367,1384,1457,1499,1505,1510,1516,1526,1564,1590,1608,1614,1643,1656,2113,2147,2153,2165],{"type":48,"tag":49,"props":50,"children":52},"element","h1",{"id":51},"optical-inspection",[53],{"type":54,"value":55},"text","Optical Inspection",{"type":48,"tag":57,"props":58,"children":59},"p",{},[60],{"type":54,"value":61},"Optical inspection for defect detection using Siamese networks. Compares image pairs to detect manufacturing defects, anomalies, or quality issues.",{"type":48,"tag":57,"props":63,"children":64},{},[65],{"type":54,"value":66},"Set train.pretrained_model_path for pretrained Siamese weights.",{"type":48,"tag":57,"props":68,"children":69},{},[70,72,79,81,87,89,95,97,103,105,111,113,119,121,127],{"type":54,"value":71},"For TAO Deploy TensorRT actions (",{"type":48,"tag":73,"props":74,"children":76},"code",{"className":75},[],[77],{"type":54,"value":78},"gen_trt_engine",{"type":54,"value":80},", TensorRT ",{"type":48,"tag":73,"props":82,"children":84},{"className":83},[],[85],{"type":54,"value":86},"evaluate",{"type":54,"value":88},", and TensorRT ",{"type":48,"tag":73,"props":90,"children":92},{"className":91},[],[93],{"type":54,"value":94},"inference",{"type":54,"value":96},"), read ",{"type":48,"tag":73,"props":98,"children":100},{"className":99},[],[101],{"type":54,"value":102},"references\u002Ftao-deploy-optical-inspection.md",{"type":54,"value":104}," first. The parent PyT container does not expose ",{"type":48,"tag":73,"props":106,"children":108},{"className":107},[],[109],{"type":54,"value":110},"optical_inspection gen_trt_engine",{"type":54,"value":112},"; TensorRT engine generation is deploy-only. Deploy spec templates live in this skill's ",{"type":48,"tag":73,"props":114,"children":116},{"className":115},[],[117],{"type":54,"value":118},"references\u002F",{"type":54,"value":120}," folder with the ",{"type":48,"tag":73,"props":122,"children":124},{"className":123},[],[125],{"type":54,"value":126},"spec_template_deploy_*.yaml",{"type":54,"value":128}," prefix.",{"type":48,"tag":130,"props":131,"children":133},"h2",{"id":132},"dataclass-schemas",[134],{"type":54,"value":135},"Dataclass Schemas",{"type":48,"tag":57,"props":137,"children":138},{},[139,141,147,149,155,157,163,165,171,173,179,181,187,189,195,197,203,205,211,213,219,221,227],{"type":54,"value":140},"Generated TAO Core schemas are packaged in ",{"type":48,"tag":73,"props":142,"children":144},{"className":143},[],[145],{"type":54,"value":146},"schemas\u002F\u003Caction>.schema.json",{"type":54,"value":148},", with ",{"type":48,"tag":73,"props":150,"children":152},{"className":151},[],[153],{"type":54,"value":154},"schemas\u002Fmanifest.json",{"type":54,"value":156}," listing available actions. Each generated schema also emits ",{"type":48,"tag":73,"props":158,"children":160},{"className":159},[],[161],{"type":54,"value":162},"references\u002Fspec_template_\u003Caction>.yaml",{"type":54,"value":164}," from the schema top-level ",{"type":48,"tag":73,"props":166,"children":168},{"className":167},[],[169],{"type":54,"value":170},"default",{"type":54,"value":172}," field. AutoML enablement is declared at the model layer in ",{"type":48,"tag":73,"props":174,"children":176},{"className":175},[],[177],{"type":54,"value":178},"references\u002Fskill_info.yaml",{"type":54,"value":180}," via ",{"type":48,"tag":73,"props":182,"children":184},{"className":183},[],[185],{"type":54,"value":186},"automl_enabled",{"type":54,"value":188},". Runnable AutoML still requires ",{"type":48,"tag":73,"props":190,"children":192},{"className":191},[],[193],{"type":54,"value":194},"schemas\u002Ftrain.schema.json",{"type":54,"value":196}," and ",{"type":48,"tag":73,"props":198,"children":200},{"className":199},[],[201],{"type":54,"value":202},"references\u002Fspec_template_train.yaml",{"type":54,"value":204}," to exist and parse. Use the packaged train schema for ",{"type":48,"tag":73,"props":206,"children":208},{"className":207},[],[209],{"type":54,"value":210},"automl_default_parameters",{"type":54,"value":212},", ",{"type":48,"tag":73,"props":214,"children":216},{"className":215},[],[217],{"type":54,"value":218},"automl_disabled_parameters",{"type":54,"value":220},", defaults, min\u002Fmax bounds, enums, option weights, math conditions, dependencies, and popular parameters. Do not expect ",{"type":48,"tag":73,"props":222,"children":224},{"className":223},[],[225],{"type":54,"value":226},"~\u002Ftao-core",{"type":54,"value":228}," at runtime; maintainers regenerate schemas\u002Ftemplates before packaging the skill bank.",{"type":48,"tag":130,"props":230,"children":232},{"id":231},"train-action-policy",[233],{"type":54,"value":234},"Train Action Policy",{"type":48,"tag":57,"props":236,"children":237},{},[238,240,245,247,253,255,261,263,269,271,277,279,285,287,292,293,299,301,306,307,312,314,320,322,328,330,335,337,342],{"type":54,"value":239},"This model is AutoML-enabled at the model layer. Before handling any train-stage request, read ",{"type":48,"tag":73,"props":241,"children":243},{"className":242},[],[244],{"type":54,"value":178},{"type":54,"value":246}," and resolve the run override from either an explicit ",{"type":48,"tag":73,"props":248,"children":250},{"className":249},[],[251],{"type":54,"value":252},"automl_policy",{"type":54,"value":254}," value or the user's workflow request. Use ",{"type":48,"tag":73,"props":256,"children":258},{"className":257},[],[259],{"type":54,"value":260},"automl_policy: on",{"type":54,"value":262}," by default and only expose ",{"type":48,"tag":73,"props":264,"children":266},{"className":265},[],[267],{"type":54,"value":268},"on",{"type":54,"value":270}," \u002F ",{"type":48,"tag":73,"props":272,"children":274},{"className":273},[],[275],{"type":54,"value":276},"off",{"type":54,"value":278}," in new launch prompts. Treat phrases like \"turn off AutoML\", \"disable AutoML\", \"no HPO\", or \"plain training\" as ",{"type":48,"tag":73,"props":280,"children":282},{"className":281},[],[283],{"type":54,"value":284},"automl_policy: off",{"type":54,"value":286}," for this run only. When ",{"type":48,"tag":73,"props":288,"children":290},{"className":289},[],[291],{"type":54,"value":260},{"type":54,"value":212},{"type":48,"tag":73,"props":294,"children":296},{"className":295},[],[297],{"type":54,"value":298},"automl_enabled: true",{"type":54,"value":300},", and both ",{"type":48,"tag":73,"props":302,"children":304},{"className":303},[],[305],{"type":54,"value":194},{"type":54,"value":196},{"type":48,"tag":73,"props":308,"children":310},{"className":309},[],[311],{"type":54,"value":202},{"type":54,"value":313}," are packaged, route the train action through ",{"type":48,"tag":73,"props":315,"children":317},{"className":316},[],[318],{"type":54,"value":319},"tao-skill-bank:tao-run-automl",{"type":54,"value":321}," by default with this model's ",{"type":48,"tag":73,"props":323,"children":325},{"className":324},[],[326],{"type":54,"value":327},"skill_dir",{"type":54,"value":329},". Preserve workflow\u002Fapplication overrides for datasets, specs, output directories, GPU\u002Fplatform settings, parent checkpoints, and ",{"type":48,"tag":73,"props":331,"children":333},{"className":332},[],[334],{"type":54,"value":252},{"type":54,"value":336},". Use direct model training only when ",{"type":48,"tag":73,"props":338,"children":340},{"className":339},[],[341],{"type":54,"value":284},{"type":54,"value":343}," 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":48,"tag":57,"props":345,"children":346},{},[347,349,354,355,360,361,367,369,374],{"type":54,"value":348},"Non-train actions such as ",{"type":48,"tag":73,"props":350,"children":352},{"className":351},[],[353],{"type":54,"value":86},{"type":54,"value":212},{"type":48,"tag":73,"props":356,"children":358},{"className":357},[],[359],{"type":54,"value":94},{"type":54,"value":212},{"type":48,"tag":73,"props":362,"children":364},{"className":363},[],[365],{"type":54,"value":366},"export",{"type":54,"value":368},", and deploy flows stay in this model skill. The per-run ",{"type":48,"tag":73,"props":370,"children":372},{"className":371},[],[373],{"type":54,"value":252},{"type":54,"value":375}," override does not change model metadata.",{"type":48,"tag":130,"props":377,"children":379},{"id":378},"training-requirements",[380],{"type":54,"value":381},"Training Requirements",{"type":48,"tag":383,"props":384,"children":385},"ul",{},[386,398,408],{"type":48,"tag":387,"props":388,"children":389},"li",{},[390,396],{"type":48,"tag":391,"props":392,"children":393},"strong",{},[394],{"type":54,"value":395},"Dataset type:",{"type":54,"value":397}," optical_inspection",{"type":48,"tag":387,"props":399,"children":400},{},[401,406],{"type":48,"tag":391,"props":402,"children":403},{},[404],{"type":54,"value":405},"Formats:",{"type":54,"value":407}," default",{"type":48,"tag":387,"props":409,"children":410},{},[411,416],{"type":48,"tag":391,"props":412,"children":413},{},[414],{"type":54,"value":415},"Monitoring metric:",{"type":54,"value":417}," val_acc",{"type":48,"tag":419,"props":420,"children":422},"h3",{"id":421},"per-action-dataset-requirements",[423],{"type":54,"value":424},"Per-Action Dataset Requirements",{"type":48,"tag":426,"props":427,"children":428},"table",{},[429,463],{"type":48,"tag":430,"props":431,"children":432},"thead",{},[433],{"type":48,"tag":434,"props":435,"children":436},"tr",{},[437,443,448,453,458],{"type":48,"tag":438,"props":439,"children":440},"th",{},[441],{"type":54,"value":442},"Action",{"type":48,"tag":438,"props":444,"children":445},{},[446],{"type":54,"value":447},"Spec Key",{"type":48,"tag":438,"props":449,"children":450},{},[451],{"type":54,"value":452},"Source",{"type":48,"tag":438,"props":454,"children":455},{},[456],{"type":54,"value":457},"Files",{"type":48,"tag":438,"props":459,"children":460},{},[461],{"type":54,"value":462},"List?",{"type":48,"tag":464,"props":465,"children":466},"tbody",{},[467,495,520,545,569,595,619,643,667,690],{"type":48,"tag":434,"props":468,"children":469},{},[470,475,480,485,490],{"type":48,"tag":471,"props":472,"children":473},"td",{},[474],{"type":54,"value":86},{"type":48,"tag":471,"props":476,"children":477},{},[478],{"type":54,"value":479},"dataset.test_dataset.images_dir",{"type":48,"tag":471,"props":481,"children":482},{},[483],{"type":54,"value":484},"eval_dataset",{"type":48,"tag":471,"props":486,"children":487},{},[488],{"type":54,"value":489},"images.tar.gz",{"type":48,"tag":471,"props":491,"children":492},{},[493],{"type":54,"value":494},"No",{"type":48,"tag":434,"props":496,"children":497},{},[498,502,507,511,516],{"type":48,"tag":471,"props":499,"children":500},{},[501],{"type":54,"value":86},{"type":48,"tag":471,"props":503,"children":504},{},[505],{"type":54,"value":506},"dataset.test_dataset.csv_path",{"type":48,"tag":471,"props":508,"children":509},{},[510],{"type":54,"value":484},{"type":48,"tag":471,"props":512,"children":513},{},[514],{"type":54,"value":515},"dataset.csv",{"type":48,"tag":471,"props":517,"children":518},{},[519],{"type":54,"value":494},{"type":48,"tag":434,"props":521,"children":522},{},[523,527,532,537,541],{"type":48,"tag":471,"props":524,"children":525},{},[526],{"type":54,"value":94},{"type":48,"tag":471,"props":528,"children":529},{},[530],{"type":54,"value":531},"dataset.infer_dataset.images_dir",{"type":48,"tag":471,"props":533,"children":534},{},[535],{"type":54,"value":536},"inference_dataset",{"type":48,"tag":471,"props":538,"children":539},{},[540],{"type":54,"value":489},{"type":48,"tag":471,"props":542,"children":543},{},[544],{"type":54,"value":494},{"type":48,"tag":434,"props":546,"children":547},{},[548,552,557,561,565],{"type":48,"tag":471,"props":549,"children":550},{},[551],{"type":54,"value":94},{"type":48,"tag":471,"props":553,"children":554},{},[555],{"type":54,"value":556},"dataset.infer_dataset.csv_path",{"type":48,"tag":471,"props":558,"children":559},{},[560],{"type":54,"value":536},{"type":48,"tag":471,"props":562,"children":563},{},[564],{"type":54,"value":515},{"type":48,"tag":471,"props":566,"children":567},{},[568],{"type":54,"value":494},{"type":48,"tag":434,"props":570,"children":571},{},[572,577,582,587,591],{"type":48,"tag":471,"props":573,"children":574},{},[575],{"type":54,"value":576},"train",{"type":48,"tag":471,"props":578,"children":579},{},[580],{"type":54,"value":581},"dataset.train_dataset.images_dir",{"type":48,"tag":471,"props":583,"children":584},{},[585],{"type":54,"value":586},"train_datasets",{"type":48,"tag":471,"props":588,"children":589},{},[590],{"type":54,"value":489},{"type":48,"tag":471,"props":592,"children":593},{},[594],{"type":54,"value":494},{"type":48,"tag":434,"props":596,"children":597},{},[598,602,607,611,615],{"type":48,"tag":471,"props":599,"children":600},{},[601],{"type":54,"value":576},{"type":48,"tag":471,"props":603,"children":604},{},[605],{"type":54,"value":606},"dataset.train_dataset.csv_path",{"type":48,"tag":471,"props":608,"children":609},{},[610],{"type":54,"value":586},{"type":48,"tag":471,"props":612,"children":613},{},[614],{"type":54,"value":515},{"type":48,"tag":471,"props":616,"children":617},{},[618],{"type":54,"value":494},{"type":48,"tag":434,"props":620,"children":621},{},[622,626,631,635,639],{"type":48,"tag":471,"props":623,"children":624},{},[625],{"type":54,"value":576},{"type":48,"tag":471,"props":627,"children":628},{},[629],{"type":54,"value":630},"dataset.validation_dataset.images_dir",{"type":48,"tag":471,"props":632,"children":633},{},[634],{"type":54,"value":484},{"type":48,"tag":471,"props":636,"children":637},{},[638],{"type":54,"value":489},{"type":48,"tag":471,"props":640,"children":641},{},[642],{"type":54,"value":494},{"type":48,"tag":434,"props":644,"children":645},{},[646,650,655,659,663],{"type":48,"tag":471,"props":647,"children":648},{},[649],{"type":54,"value":576},{"type":48,"tag":471,"props":651,"children":652},{},[653],{"type":54,"value":654},"dataset.validation_dataset.csv_path",{"type":48,"tag":471,"props":656,"children":657},{},[658],{"type":54,"value":484},{"type":48,"tag":471,"props":660,"children":661},{},[662],{"type":54,"value":515},{"type":48,"tag":471,"props":664,"children":665},{},[666],{"type":54,"value":494},{"type":48,"tag":434,"props":668,"children":669},{},[670,674,678,682,686],{"type":48,"tag":471,"props":671,"children":672},{},[673],{"type":54,"value":576},{"type":48,"tag":471,"props":675,"children":676},{},[677],{"type":54,"value":479},{"type":48,"tag":471,"props":679,"children":680},{},[681],{"type":54,"value":484},{"type":48,"tag":471,"props":683,"children":684},{},[685],{"type":54,"value":489},{"type":48,"tag":471,"props":687,"children":688},{},[689],{"type":54,"value":494},{"type":48,"tag":434,"props":691,"children":692},{},[693,697,701,705,709],{"type":48,"tag":471,"props":694,"children":695},{},[696],{"type":54,"value":576},{"type":48,"tag":471,"props":698,"children":699},{},[700],{"type":54,"value":506},{"type":48,"tag":471,"props":702,"children":703},{},[704],{"type":54,"value":484},{"type":48,"tag":471,"props":706,"children":707},{},[708],{"type":54,"value":515},{"type":48,"tag":471,"props":710,"children":711},{},[712],{"type":54,"value":494},{"type":48,"tag":419,"props":714,"children":716},{"id":715},"typical-spec-overrides",[717],{"type":54,"value":718},"Typical Spec Overrides",{"type":48,"tag":57,"props":720,"children":721},{},[722,724,729,731,737],{"type":54,"value":723},"Data source overrides are ",{"type":48,"tag":391,"props":725,"children":726},{},[727],{"type":54,"value":728},"mandatory for every action",{"type":54,"value":730}," — the agent MUST construct data source paths from the Per-Action Dataset Requirements table above and include them in ",{"type":48,"tag":73,"props":732,"children":734},{"className":733},[],[735],{"type":54,"value":736},"spec_overrides",{"type":54,"value":738},".",{"type":48,"tag":740,"props":741,"children":746},"pre",{"className":742,"code":743,"language":744,"meta":745,"style":745},"language-python shiki shiki-themes material-theme-lighter material-theme material-theme-palenight","S3_TRAIN = \"s3:\u002F\u002Fbucket\u002Fdata\u002Ftrain\"\nS3_EVAL = \"s3:\u002F\u002Fbucket\u002Fdata\u002Feval\"\nS3_INFERENCE = \"s3:\u002F\u002Fbucket\u002Fdata\u002Finference\"\n","python","",[747],{"type":48,"tag":73,"props":748,"children":749},{"__ignoreMap":745},[750,761,770],{"type":48,"tag":751,"props":752,"children":755},"span",{"class":753,"line":754},"line",1,[756],{"type":48,"tag":751,"props":757,"children":758},{},[759],{"type":54,"value":760},"S3_TRAIN = \"s3:\u002F\u002Fbucket\u002Fdata\u002Ftrain\"\n",{"type":48,"tag":751,"props":762,"children":764},{"class":753,"line":763},2,[765],{"type":48,"tag":751,"props":766,"children":767},{},[768],{"type":54,"value":769},"S3_EVAL = \"s3:\u002F\u002Fbucket\u002Fdata\u002Feval\"\n",{"type":48,"tag":751,"props":771,"children":773},{"class":753,"line":772},3,[774],{"type":48,"tag":751,"props":775,"children":776},{},[777],{"type":54,"value":778},"S3_INFERENCE = \"s3:\u002F\u002Fbucket\u002Fdata\u002Finference\"\n",{"type":48,"tag":57,"props":780,"children":781},{},[782],{"type":48,"tag":391,"props":783,"children":784},{},[785],{"type":54,"value":786},"train (mandatory data sources):",{"type":48,"tag":740,"props":788,"children":790},{"className":742,"code":789,"language":744,"meta":745,"style":745},"{\n    \"train.num_epochs\": 30,\n    \"train.checkpoint_interval\": 10,\n    \"train.validation_interval\": 10,\n    \"train.num_gpus\": 1,\n    \"dataset.batch_size\": 8,\n    \"dataset.train_dataset.images_dir\": f\"{S3_TRAIN}\u002Fimages.tar.gz\",\n    \"dataset.train_dataset.csv_path\": f\"{S3_TRAIN}\u002Fdataset.csv\",\n    \"dataset.validation_dataset.images_dir\": f\"{S3_EVAL}\u002Fimages.tar.gz\",\n    \"dataset.validation_dataset.csv_path\": f\"{S3_EVAL}\u002Fdataset.csv\",\n    \"dataset.test_dataset.images_dir\": f\"{S3_EVAL}\u002Fimages.tar.gz\",\n    \"dataset.test_dataset.csv_path\": f\"{S3_EVAL}\u002Fdataset.csv\",\n}\n",[791],{"type":48,"tag":73,"props":792,"children":793},{"__ignoreMap":745},[794,802,810,818,827,836,845,854,863,872,881,890,899],{"type":48,"tag":751,"props":795,"children":796},{"class":753,"line":754},[797],{"type":48,"tag":751,"props":798,"children":799},{},[800],{"type":54,"value":801},"{\n",{"type":48,"tag":751,"props":803,"children":804},{"class":753,"line":763},[805],{"type":48,"tag":751,"props":806,"children":807},{},[808],{"type":54,"value":809},"    \"train.num_epochs\": 30,\n",{"type":48,"tag":751,"props":811,"children":812},{"class":753,"line":772},[813],{"type":48,"tag":751,"props":814,"children":815},{},[816],{"type":54,"value":817},"    \"train.checkpoint_interval\": 10,\n",{"type":48,"tag":751,"props":819,"children":821},{"class":753,"line":820},4,[822],{"type":48,"tag":751,"props":823,"children":824},{},[825],{"type":54,"value":826},"    \"train.validation_interval\": 10,\n",{"type":48,"tag":751,"props":828,"children":830},{"class":753,"line":829},5,[831],{"type":48,"tag":751,"props":832,"children":833},{},[834],{"type":54,"value":835},"    \"train.num_gpus\": 1,\n",{"type":48,"tag":751,"props":837,"children":839},{"class":753,"line":838},6,[840],{"type":48,"tag":751,"props":841,"children":842},{},[843],{"type":54,"value":844},"    \"dataset.batch_size\": 8,\n",{"type":48,"tag":751,"props":846,"children":848},{"class":753,"line":847},7,[849],{"type":48,"tag":751,"props":850,"children":851},{},[852],{"type":54,"value":853},"    \"dataset.train_dataset.images_dir\": f\"{S3_TRAIN}\u002Fimages.tar.gz\",\n",{"type":48,"tag":751,"props":855,"children":857},{"class":753,"line":856},8,[858],{"type":48,"tag":751,"props":859,"children":860},{},[861],{"type":54,"value":862},"    \"dataset.train_dataset.csv_path\": f\"{S3_TRAIN}\u002Fdataset.csv\",\n",{"type":48,"tag":751,"props":864,"children":866},{"class":753,"line":865},9,[867],{"type":48,"tag":751,"props":868,"children":869},{},[870],{"type":54,"value":871},"    \"dataset.validation_dataset.images_dir\": f\"{S3_EVAL}\u002Fimages.tar.gz\",\n",{"type":48,"tag":751,"props":873,"children":875},{"class":753,"line":874},10,[876],{"type":48,"tag":751,"props":877,"children":878},{},[879],{"type":54,"value":880},"    \"dataset.validation_dataset.csv_path\": f\"{S3_EVAL}\u002Fdataset.csv\",\n",{"type":48,"tag":751,"props":882,"children":884},{"class":753,"line":883},11,[885],{"type":48,"tag":751,"props":886,"children":887},{},[888],{"type":54,"value":889},"    \"dataset.test_dataset.images_dir\": f\"{S3_EVAL}\u002Fimages.tar.gz\",\n",{"type":48,"tag":751,"props":891,"children":893},{"class":753,"line":892},12,[894],{"type":48,"tag":751,"props":895,"children":896},{},[897],{"type":54,"value":898},"    \"dataset.test_dataset.csv_path\": f\"{S3_EVAL}\u002Fdataset.csv\",\n",{"type":48,"tag":751,"props":900,"children":902},{"class":753,"line":901},13,[903],{"type":48,"tag":751,"props":904,"children":905},{},[906],{"type":54,"value":907},"}\n",{"type":48,"tag":57,"props":909,"children":910},{},[911],{"type":48,"tag":391,"props":912,"children":913},{},[914],{"type":54,"value":915},"evaluate (mandatory data sources):",{"type":48,"tag":740,"props":917,"children":919},{"className":742,"code":918,"language":744,"meta":745,"style":745},"{\n    \"evaluate.checkpoint\": \"\u003Cselected train\u002FAutoML checkpoint>\",\n    \"dataset.test_dataset.images_dir\": f\"{S3_EVAL}\u002Fimages.tar.gz\",\n    \"dataset.test_dataset.csv_path\": f\"{S3_EVAL}\u002Fdataset.csv\",\n}\n",[920],{"type":48,"tag":73,"props":921,"children":922},{"__ignoreMap":745},[923,930,938,945,952],{"type":48,"tag":751,"props":924,"children":925},{"class":753,"line":754},[926],{"type":48,"tag":751,"props":927,"children":928},{},[929],{"type":54,"value":801},{"type":48,"tag":751,"props":931,"children":932},{"class":753,"line":763},[933],{"type":48,"tag":751,"props":934,"children":935},{},[936],{"type":54,"value":937},"    \"evaluate.checkpoint\": \"\u003Cselected train\u002FAutoML checkpoint>\",\n",{"type":48,"tag":751,"props":939,"children":940},{"class":753,"line":772},[941],{"type":48,"tag":751,"props":942,"children":943},{},[944],{"type":54,"value":889},{"type":48,"tag":751,"props":946,"children":947},{"class":753,"line":820},[948],{"type":48,"tag":751,"props":949,"children":950},{},[951],{"type":54,"value":898},{"type":48,"tag":751,"props":953,"children":954},{"class":753,"line":829},[955],{"type":48,"tag":751,"props":956,"children":957},{},[958],{"type":54,"value":907},{"type":48,"tag":57,"props":960,"children":961},{},[962,964,970,972,978],{"type":54,"value":963},"Use the workflow's checkpoint resolver for downstream actions instead of guessing a filename. For Optical Inspection smoke runs, AutoML may produce ",{"type":48,"tag":73,"props":965,"children":967},{"className":966},[],[968],{"type":54,"value":969},"model_epoch_000_step_00006.pth",{"type":54,"value":971},"; resume can then produce ",{"type":48,"tag":73,"props":973,"children":975},{"className":974},[],[976],{"type":54,"value":977},"model_epoch_001_step_00012.pth",{"type":54,"value":979},". Best-checkpoint actions should use the AutoML best child job's selected checkpoint, epoch-specific actions should pass the exact epoch\u002Fstep checkpoint requested, and only explicit \"latest\" requests should resolve to the latest checkpoint.",{"type":48,"tag":57,"props":981,"children":982},{},[983],{"type":48,"tag":391,"props":984,"children":985},{},[986],{"type":54,"value":987},"export:",{"type":48,"tag":740,"props":989,"children":991},{"className":742,"code":990,"language":744,"meta":745,"style":745},"{\n    \"export.checkpoint\": \"\u003Cselected train\u002FAutoML checkpoint>\",\n    \"export.onnx_file\": \"\u002Fresults\u002Foptical_inspection.onnx\",\n    \"export.input_width\": 128,\n    \"export.input_height\": 512,\n    \"export.batch_size\": 1,\n}\n",[992],{"type":48,"tag":73,"props":993,"children":994},{"__ignoreMap":745},[995,1002,1010,1018,1026,1034,1042],{"type":48,"tag":751,"props":996,"children":997},{"class":753,"line":754},[998],{"type":48,"tag":751,"props":999,"children":1000},{},[1001],{"type":54,"value":801},{"type":48,"tag":751,"props":1003,"children":1004},{"class":753,"line":763},[1005],{"type":48,"tag":751,"props":1006,"children":1007},{},[1008],{"type":54,"value":1009},"    \"export.checkpoint\": \"\u003Cselected train\u002FAutoML checkpoint>\",\n",{"type":48,"tag":751,"props":1011,"children":1012},{"class":753,"line":772},[1013],{"type":48,"tag":751,"props":1014,"children":1015},{},[1016],{"type":54,"value":1017},"    \"export.onnx_file\": \"\u002Fresults\u002Foptical_inspection.onnx\",\n",{"type":48,"tag":751,"props":1019,"children":1020},{"class":753,"line":820},[1021],{"type":48,"tag":751,"props":1022,"children":1023},{},[1024],{"type":54,"value":1025},"    \"export.input_width\": 128,\n",{"type":48,"tag":751,"props":1027,"children":1028},{"class":753,"line":829},[1029],{"type":48,"tag":751,"props":1030,"children":1031},{},[1032],{"type":54,"value":1033},"    \"export.input_height\": 512,\n",{"type":48,"tag":751,"props":1035,"children":1036},{"class":753,"line":838},[1037],{"type":48,"tag":751,"props":1038,"children":1039},{},[1040],{"type":54,"value":1041},"    \"export.batch_size\": 1,\n",{"type":48,"tag":751,"props":1043,"children":1044},{"class":753,"line":847},[1045],{"type":48,"tag":751,"props":1046,"children":1047},{},[1048],{"type":54,"value":907},{"type":48,"tag":57,"props":1050,"children":1051},{},[1052],{"type":48,"tag":391,"props":1053,"children":1054},{},[1055],{"type":54,"value":1056},"inference (mandatory data sources):",{"type":48,"tag":740,"props":1058,"children":1060},{"className":742,"code":1059,"language":744,"meta":745,"style":745},"{\n    \"inference.checkpoint\": \"\u003Cselected train\u002FAutoML checkpoint>\",\n    \"dataset.infer_dataset.images_dir\": f\"{S3_INFERENCE}\u002Fimages.tar.gz\",\n    \"dataset.infer_dataset.csv_path\": f\"{S3_INFERENCE}\u002Fdataset.csv\",\n}\n",[1061],{"type":48,"tag":73,"props":1062,"children":1063},{"__ignoreMap":745},[1064,1071,1079,1087,1095],{"type":48,"tag":751,"props":1065,"children":1066},{"class":753,"line":754},[1067],{"type":48,"tag":751,"props":1068,"children":1069},{},[1070],{"type":54,"value":801},{"type":48,"tag":751,"props":1072,"children":1073},{"class":753,"line":763},[1074],{"type":48,"tag":751,"props":1075,"children":1076},{},[1077],{"type":54,"value":1078},"    \"inference.checkpoint\": \"\u003Cselected train\u002FAutoML checkpoint>\",\n",{"type":48,"tag":751,"props":1080,"children":1081},{"class":753,"line":772},[1082],{"type":48,"tag":751,"props":1083,"children":1084},{},[1085],{"type":54,"value":1086},"    \"dataset.infer_dataset.images_dir\": f\"{S3_INFERENCE}\u002Fimages.tar.gz\",\n",{"type":48,"tag":751,"props":1088,"children":1089},{"class":753,"line":820},[1090],{"type":48,"tag":751,"props":1091,"children":1092},{},[1093],{"type":54,"value":1094},"    \"dataset.infer_dataset.csv_path\": f\"{S3_INFERENCE}\u002Fdataset.csv\",\n",{"type":48,"tag":751,"props":1096,"children":1097},{"class":753,"line":829},[1098],{"type":48,"tag":751,"props":1099,"children":1100},{},[1101],{"type":54,"value":907},{"type":48,"tag":130,"props":1103,"children":1105},{"id":1104},"dataset-convert",[1106],{"type":54,"value":1107},"Dataset Convert",{"type":48,"tag":57,"props":1109,"children":1110},{},[1111,1113,1118,1120,1125,1127,1132,1133,1138,1139,1144],{"type":54,"value":1112},"Dataset conversion is optional for Optical Inspection. If the dataset is already in TAO-ready Optical Inspection format, start directly from the ",{"type":48,"tag":73,"props":1114,"children":1116},{"className":1115},[],[1117],{"type":54,"value":489},{"type":54,"value":1119}," plus ",{"type":48,"tag":73,"props":1121,"children":1123},{"className":1122},[],[1124],{"type":54,"value":515},{"type":54,"value":1126}," splits and run ",{"type":48,"tag":73,"props":1128,"children":1130},{"className":1129},[],[1131],{"type":54,"value":576},{"type":54,"value":212},{"type":48,"tag":73,"props":1134,"children":1136},{"className":1135},[],[1137],{"type":54,"value":86},{"type":54,"value":212},{"type":48,"tag":73,"props":1140,"children":1142},{"className":1141},[],[1143],{"type":54,"value":94},{"type":54,"value":1145},", and downstream checkpoint\u002Fexport\u002Fdeploy actions on that converted data.",{"type":48,"tag":57,"props":1147,"children":1148},{},[1149,1151,1157,1159,1165,1167,1173,1175,1181,1182,1188,1190,1196,1198,1203,1204,1209,1211,1217],{"type":54,"value":1150},"The PyT container exposes ",{"type":48,"tag":73,"props":1152,"children":1154},{"className":1153},[],[1155],{"type":54,"value":1156},"optical_inspection dataset_convert",{"type":54,"value":1158},", but this model skill does not package a ",{"type":48,"tag":73,"props":1160,"children":1162},{"className":1161},[],[1163],{"type":54,"value":1164},"dataset_convert",{"type":54,"value":1166}," action\u002Ftemplate. The converter expects the raw Factory PCB layout (",{"type":48,"tag":73,"props":1168,"children":1170},{"className":1169},[],[1171],{"type":54,"value":1172},"root_dataset_dir",{"type":54,"value":1174},", train\u002Fval\u002Fall PCB directories, ",{"type":48,"tag":73,"props":1176,"children":1178},{"className":1177},[],[1179],{"type":54,"value":1180},"golden_csv_dir",{"type":54,"value":212},{"type":48,"tag":73,"props":1183,"children":1185},{"className":1184},[],[1186],{"type":54,"value":1187},"project_name",{"type":54,"value":1189},", and ",{"type":48,"tag":73,"props":1191,"children":1193},{"className":1192},[],[1194],{"type":54,"value":1195},"bot_top",{"type":54,"value":1197},"). The S3 validation bucket currently contains preconverted Optical Inspection ",{"type":48,"tag":73,"props":1199,"children":1201},{"className":1200},[],[1202],{"type":54,"value":489},{"type":54,"value":1119},{"type":48,"tag":73,"props":1205,"children":1207},{"className":1206},[],[1208],{"type":54,"value":515},{"type":54,"value":1210}," splits, not the raw PCB\u002Fgolden CSV source. Do not synthesize a fake PCB dataset. In model validation reports, mark dataset conversion as ",{"type":48,"tag":73,"props":1212,"children":1214},{"className":1213},[],[1215],{"type":54,"value":1216},"not run: preconverted dataset provided",{"type":54,"value":1218}," rather than failed or blocked when only converted data is available.",{"type":48,"tag":57,"props":1220,"children":1221},{},[1222,1224,1230,1232,1238,1240,1246,1248,1253,1255,1261],{"type":54,"value":1223},"When using the preconverted S3 validation tarballs locally, verify the extracted directory before writing specs. The tarballs may unpack an ",{"type":48,"tag":73,"props":1225,"children":1227},{"className":1226},[],[1228],{"type":54,"value":1229},"images\u002F",{"type":54,"value":1231}," wrapper directory; point ",{"type":48,"tag":73,"props":1233,"children":1235},{"className":1234},[],[1236],{"type":54,"value":1237},"dataset.*.images_dir",{"type":54,"value":1239}," at the inner directory that contains ",{"type":48,"tag":73,"props":1241,"children":1243},{"className":1242},[],[1244],{"type":54,"value":1245},"golden\u002F",{"type":54,"value":1247}," and the board\u002Fimage folders referenced by ",{"type":48,"tag":73,"props":1249,"children":1251},{"className":1250},[],[1252],{"type":54,"value":515},{"type":54,"value":1254},", for example ",{"type":48,"tag":73,"props":1256,"children":1258},{"className":1257},[],[1259],{"type":54,"value":1260},"...\u002F\u003Csplit>\u002Fimages\u002Fimages",{"type":54,"value":1262},", not the outer wrapper.",{"type":48,"tag":130,"props":1264,"children":1266},{"id":1265},"eval-dataset",[1267],{"type":54,"value":1268},"Eval Dataset",{"type":48,"tag":57,"props":1270,"children":1271},{},[1272],{"type":54,"value":1273},"Optional. Eval dataset uses same format (images + CSV).",{"type":48,"tag":130,"props":1275,"children":1277},{"id":1276},"important-parameters",[1278],{"type":54,"value":1279},"Important Parameters",{"type":48,"tag":383,"props":1281,"children":1282},{},[1283,1293,1303,1313,1323,1341,1351],{"type":48,"tag":387,"props":1284,"children":1285},{},[1286,1291],{"type":48,"tag":391,"props":1287,"children":1288},{},[1289],{"type":54,"value":1290},"model.model_type",{"type":54,"value":1292},": Siamese variant. Options include Siamese, Siamese_3.",{"type":48,"tag":387,"props":1294,"children":1295},{},[1296,1301],{"type":48,"tag":391,"props":1297,"children":1298},{},[1299],{"type":54,"value":1300},"model.model_backbone",{"type":54,"value":1302},": Default custom.",{"type":48,"tag":387,"props":1304,"children":1305},{},[1306,1311],{"type":48,"tag":391,"props":1307,"children":1308},{},[1309],{"type":54,"value":1310},"model.embedding_vectors",{"type":54,"value":1312},": Number of embedding dimensions. Default 5.",{"type":48,"tag":387,"props":1314,"children":1315},{},[1316,1321],{"type":48,"tag":391,"props":1317,"children":1318},{},[1319],{"type":54,"value":1320},"train.optim.lr",{"type":54,"value":1322},": Learning rate. Default 5e-4.",{"type":48,"tag":387,"props":1324,"children":1325},{},[1326,1331,1333,1339],{"type":48,"tag":391,"props":1327,"children":1328},{},[1329],{"type":54,"value":1330},"dataset.batch_size",{"type":54,"value":1332},": Training batch size. Must be greater than 1; use ",{"type":48,"tag":73,"props":1334,"children":1336},{"className":1335},[],[1337],{"type":54,"value":1338},"2",{"type":54,"value":1340}," or higher for minimal smoke runs.",{"type":48,"tag":387,"props":1342,"children":1343},{},[1344,1349],{"type":48,"tag":391,"props":1345,"children":1346},{},[1347],{"type":54,"value":1348},"dataset.num_input",{"type":54,"value":1350},": Number of input images per comparison.",{"type":48,"tag":387,"props":1352,"children":1353},{},[1354,1359],{"type":48,"tag":391,"props":1355,"children":1356},{},[1357],{"type":54,"value":1358},"dataset.input_map",{"type":54,"value":1360},": Mapping of input channels \u002F image pairs.",{"type":48,"tag":130,"props":1362,"children":1364},{"id":1363},"multi-gpu-multi-node",[1365],{"type":54,"value":1366},"Multi-GPU \u002F Multi-Node",{"type":48,"tag":57,"props":1368,"children":1369},{},[1370,1375,1377,1382],{"type":48,"tag":391,"props":1371,"children":1372},{},[1373],{"type":54,"value":1374},"Launch method:",{"type":54,"value":1376}," Lightning-managed (single ",{"type":48,"tag":73,"props":1378,"children":1380},{"className":1379},[],[1381],{"type":54,"value":744},{"type":54,"value":1383}," process, Lightning spawns workers).",{"type":48,"tag":426,"props":1385,"children":1386},{},[1387,1407],{"type":48,"tag":430,"props":1388,"children":1389},{},[1390],{"type":48,"tag":434,"props":1391,"children":1392},{},[1393,1397,1402],{"type":48,"tag":438,"props":1394,"children":1395},{},[1396],{"type":54,"value":447},{"type":48,"tag":438,"props":1398,"children":1399},{},[1400],{"type":54,"value":1401},"Description",{"type":48,"tag":438,"props":1403,"children":1404},{},[1405],{"type":54,"value":1406},"Default",{"type":48,"tag":464,"props":1408,"children":1409},{},[1410,1432],{"type":48,"tag":434,"props":1411,"children":1412},{},[1413,1422,1427],{"type":48,"tag":471,"props":1414,"children":1415},{},[1416],{"type":48,"tag":73,"props":1417,"children":1419},{"className":1418},[],[1420],{"type":54,"value":1421},"train.num_gpus",{"type":48,"tag":471,"props":1423,"children":1424},{},[1425],{"type":54,"value":1426},"Number of GPUs",{"type":48,"tag":471,"props":1428,"children":1429},{},[1430],{"type":54,"value":1431},"1",{"type":48,"tag":434,"props":1433,"children":1434},{},[1435,1444,1449],{"type":48,"tag":471,"props":1436,"children":1437},{},[1438],{"type":48,"tag":73,"props":1439,"children":1441},{"className":1440},[],[1442],{"type":54,"value":1443},"train.gpu_ids",{"type":48,"tag":471,"props":1445,"children":1446},{},[1447],{"type":54,"value":1448},"GPU device indices",{"type":48,"tag":471,"props":1450,"children":1451},{},[1452],{"type":48,"tag":751,"props":1453,"children":1454},{},[1455],{"type":54,"value":1456},"0",{"type":48,"tag":383,"props":1458,"children":1459},{},[1460,1473,1494],{"type":48,"tag":387,"props":1461,"children":1462},{},[1463,1465,1471],{"type":54,"value":1464},"Strategy: ",{"type":48,"tag":73,"props":1466,"children":1468},{"className":1467},[],[1469],{"type":54,"value":1470},"auto",{"type":54,"value":1472}," (Lightning picks best strategy automatically)",{"type":48,"tag":387,"props":1474,"children":1475},{},[1476,1478,1484,1486,1492],{"type":54,"value":1477},"No explicit ",{"type":48,"tag":73,"props":1479,"children":1481},{"className":1480},[],[1482],{"type":54,"value":1483},"num_nodes",{"type":54,"value":1485}," or ",{"type":48,"tag":73,"props":1487,"children":1489},{"className":1488},[],[1490],{"type":54,"value":1491},"distributed_strategy",{"type":54,"value":1493}," config — single-node only",{"type":48,"tag":387,"props":1495,"children":1496},{},[1497],{"type":54,"value":1498},"Lightweight Siamese network, single GPU typically sufficient",{"type":48,"tag":130,"props":1500,"children":1502},{"id":1501},"hardware",[1503],{"type":54,"value":1504},"Hardware",{"type":48,"tag":57,"props":1506,"children":1507},{},[1508],{"type":54,"value":1509},"Minimum 1 GPU(s), recommended 1 GPU(s). 8GB+ VRAM per GPU. Siamese networks for inspection are lightweight. Single GPU sufficient.",{"type":48,"tag":130,"props":1511,"children":1513},{"id":1512},"error-patterns",[1514],{"type":54,"value":1515},"Error Patterns",{"type":48,"tag":57,"props":1517,"children":1518},{},[1519,1524],{"type":48,"tag":391,"props":1520,"children":1521},{},[1522],{"type":54,"value":1523},"CSV format error",{"type":54,"value":1525},": Ensure dataset.csv has the correct column format for image pair paths and labels.",{"type":48,"tag":57,"props":1527,"children":1528},{},[1529,1534,1536,1541,1543,1548,1550,1555,1557,1563],{"type":48,"tag":391,"props":1530,"children":1531},{},[1532],{"type":54,"value":1533},"Extracted image root mismatch",{"type":54,"value":1535},": If train, evaluate, or inference cannot find paths from ",{"type":48,"tag":73,"props":1537,"children":1539},{"className":1538},[],[1540],{"type":54,"value":515},{"type":54,"value":1542},", inspect the extracted ",{"type":48,"tag":73,"props":1544,"children":1546},{"className":1545},[],[1547],{"type":54,"value":489},{"type":54,"value":1549}," tree. The TAO-ready root must contain ",{"type":48,"tag":73,"props":1551,"children":1553},{"className":1552},[],[1554],{"type":54,"value":1245},{"type":54,"value":1556}," plus the board folders referenced in the CSV. For validation S3 tarballs this can be one level below the extraction target, such as ",{"type":48,"tag":73,"props":1558,"children":1560},{"className":1559},[],[1561],{"type":54,"value":1562},"images\u002Fimages",{"type":54,"value":738},{"type":48,"tag":57,"props":1565,"children":1566},{},[1567,1572,1574,1580,1582,1588],{"type":48,"tag":391,"props":1568,"children":1569},{},[1570],{"type":54,"value":1571},"Training batch size assertion",{"type":54,"value":1573},": The Optical Inspection dataloader rejects\n",{"type":48,"tag":73,"props":1575,"children":1577},{"className":1576},[],[1578],{"type":54,"value":1579},"dataset.batch_size: 1",{"type":54,"value":1581}," for train. Keep the template default of 8 for normal\nruns, or set ",{"type":48,"tag":73,"props":1583,"children":1585},{"className":1584},[],[1586],{"type":54,"value":1587},"dataset.batch_size: 2",{"type":54,"value":1589}," for minimal AutoML smoke validation.",{"type":48,"tag":57,"props":1591,"children":1592},{},[1593,1598,1600,1606],{"type":48,"tag":391,"props":1594,"children":1595},{},[1596],{"type":54,"value":1597},"PyTorch checkpoint load failure on downstream actions",{"type":54,"value":1599},": For checkpoints\nproduced by the same trusted TAO train\u002FAutoML workflow, set\n",{"type":48,"tag":73,"props":1601,"children":1603},{"className":1602},[],[1604],{"type":54,"value":1605},"TORCH_FORCE_NO_WEIGHTS_ONLY_LOAD=1",{"type":54,"value":1607}," for evaluate, inference, export, and\nresume jobs if the current PyTorch default blocks loading the full checkpoint.\nDo not use this env var for untrusted checkpoints.",{"type":48,"tag":130,"props":1609,"children":1611},{"id":1610},"spec-param-parent-model-inference",[1612],{"type":54,"value":1613},"Spec Param \u002F Parent Model Inference",{"type":48,"tag":57,"props":1615,"children":1616},{},[1617,1619,1625,1627,1633,1635,1641],{"type":54,"value":1618},"Model-specific inference mappings belong in this MD file, not in ",{"type":48,"tag":73,"props":1620,"children":1622},{"className":1621},[],[1623],{"type":54,"value":1624},"config.json",{"type":54,"value":1626},". Generated runners should read this section and apply the mappings with SDK helpers before ",{"type":48,"tag":73,"props":1628,"children":1630},{"className":1629},[],[1631],{"type":54,"value":1632},"create_job()",{"type":54,"value":1634},". This mirrors the old microservices ",{"type":48,"tag":73,"props":1636,"children":1638},{"className":1637},[],[1639],{"type":54,"value":1640},"infer_params.py",{"type":54,"value":1642}," flow.",{"type":48,"tag":57,"props":1644,"children":1645},{},[1646,1648,1654],{"type":54,"value":1647},"Inference mappings from TAO Core ",{"type":48,"tag":73,"props":1649,"children":1651},{"className":1650},[],[1652],{"type":54,"value":1653},"optical_inspection.config.json",{"type":54,"value":1655},":",{"type":48,"tag":426,"props":1657,"children":1658},{},[1659,1684],{"type":48,"tag":430,"props":1660,"children":1661},{},[1662],{"type":48,"tag":434,"props":1663,"children":1664},{},[1665,1669,1674,1679],{"type":48,"tag":438,"props":1666,"children":1667},{},[1668],{"type":54,"value":442},{"type":48,"tag":438,"props":1670,"children":1671},{},[1672],{"type":54,"value":1673},"Spec Field",{"type":48,"tag":438,"props":1675,"children":1676},{},[1677],{"type":54,"value":1678},"Inference Function",{"type":48,"tag":438,"props":1680,"children":1681},{},[1682],{"type":54,"value":1683},"Meaning",{"type":48,"tag":464,"props":1685,"children":1686},{},[1687,1717,1747,1777,1804,1832,1862,1889,1916,1944,1972,1999,2026,2053,2083],{"type":48,"tag":434,"props":1688,"children":1689},{},[1690,1694,1703,1712],{"type":48,"tag":471,"props":1691,"children":1692},{},[1693],{"type":54,"value":86},{"type":48,"tag":471,"props":1695,"children":1696},{},[1697],{"type":48,"tag":73,"props":1698,"children":1700},{"className":1699},[],[1701],{"type":54,"value":1702},"encryption_key",{"type":48,"tag":471,"props":1704,"children":1705},{},[1706],{"type":48,"tag":73,"props":1707,"children":1709},{"className":1708},[],[1710],{"type":54,"value":1711},"key",{"type":48,"tag":471,"props":1713,"children":1714},{},[1715],{"type":54,"value":1716},"encryption key",{"type":48,"tag":434,"props":1718,"children":1719},{},[1720,1724,1733,1742],{"type":48,"tag":471,"props":1721,"children":1722},{},[1723],{"type":54,"value":86},{"type":48,"tag":471,"props":1725,"children":1726},{},[1727],{"type":48,"tag":73,"props":1728,"children":1730},{"className":1729},[],[1731],{"type":54,"value":1732},"evaluate.checkpoint",{"type":48,"tag":471,"props":1734,"children":1735},{},[1736],{"type":48,"tag":73,"props":1737,"children":1739},{"className":1738},[],[1740],{"type":54,"value":1741},"parent_model",{"type":48,"tag":471,"props":1743,"children":1744},{},[1745],{"type":54,"value":1746},"model file inferred from the parent job results folder",{"type":48,"tag":434,"props":1748,"children":1749},{},[1750,1754,1763,1772],{"type":48,"tag":471,"props":1751,"children":1752},{},[1753],{"type":54,"value":86},{"type":48,"tag":471,"props":1755,"children":1756},{},[1757],{"type":48,"tag":73,"props":1758,"children":1760},{"className":1759},[],[1761],{"type":54,"value":1762},"results_dir",{"type":48,"tag":471,"props":1764,"children":1765},{},[1766],{"type":48,"tag":73,"props":1767,"children":1769},{"className":1768},[],[1770],{"type":54,"value":1771},"output_dir",{"type":48,"tag":471,"props":1773,"children":1774},{},[1775],{"type":54,"value":1776},"current job results directory",{"type":48,"tag":434,"props":1778,"children":1779},{},[1780,1784,1792,1800],{"type":48,"tag":471,"props":1781,"children":1782},{},[1783],{"type":54,"value":366},{"type":48,"tag":471,"props":1785,"children":1786},{},[1787],{"type":48,"tag":73,"props":1788,"children":1790},{"className":1789},[],[1791],{"type":54,"value":1702},{"type":48,"tag":471,"props":1793,"children":1794},{},[1795],{"type":48,"tag":73,"props":1796,"children":1798},{"className":1797},[],[1799],{"type":54,"value":1711},{"type":48,"tag":471,"props":1801,"children":1802},{},[1803],{"type":54,"value":1716},{"type":48,"tag":434,"props":1805,"children":1806},{},[1807,1811,1820,1828],{"type":48,"tag":471,"props":1808,"children":1809},{},[1810],{"type":54,"value":366},{"type":48,"tag":471,"props":1812,"children":1813},{},[1814],{"type":48,"tag":73,"props":1815,"children":1817},{"className":1816},[],[1818],{"type":54,"value":1819},"export.checkpoint",{"type":48,"tag":471,"props":1821,"children":1822},{},[1823],{"type":48,"tag":73,"props":1824,"children":1826},{"className":1825},[],[1827],{"type":54,"value":1741},{"type":48,"tag":471,"props":1829,"children":1830},{},[1831],{"type":54,"value":1746},{"type":48,"tag":434,"props":1833,"children":1834},{},[1835,1839,1848,1857],{"type":48,"tag":471,"props":1836,"children":1837},{},[1838],{"type":54,"value":366},{"type":48,"tag":471,"props":1840,"children":1841},{},[1842],{"type":48,"tag":73,"props":1843,"children":1845},{"className":1844},[],[1846],{"type":54,"value":1847},"export.onnx_file",{"type":48,"tag":471,"props":1849,"children":1850},{},[1851],{"type":48,"tag":73,"props":1852,"children":1854},{"className":1853},[],[1855],{"type":54,"value":1856},"create_onnx_file",{"type":48,"tag":471,"props":1858,"children":1859},{},[1860],{"type":54,"value":1861},"output ONNX path",{"type":48,"tag":434,"props":1863,"children":1864},{},[1865,1869,1877,1885],{"type":48,"tag":471,"props":1866,"children":1867},{},[1868],{"type":54,"value":366},{"type":48,"tag":471,"props":1870,"children":1871},{},[1872],{"type":48,"tag":73,"props":1873,"children":1875},{"className":1874},[],[1876],{"type":54,"value":1762},{"type":48,"tag":471,"props":1878,"children":1879},{},[1880],{"type":48,"tag":73,"props":1881,"children":1883},{"className":1882},[],[1884],{"type":54,"value":1771},{"type":48,"tag":471,"props":1886,"children":1887},{},[1888],{"type":54,"value":1776},{"type":48,"tag":434,"props":1890,"children":1891},{},[1892,1896,1904,1912],{"type":48,"tag":471,"props":1893,"children":1894},{},[1895],{"type":54,"value":94},{"type":48,"tag":471,"props":1897,"children":1898},{},[1899],{"type":48,"tag":73,"props":1900,"children":1902},{"className":1901},[],[1903],{"type":54,"value":1702},{"type":48,"tag":471,"props":1905,"children":1906},{},[1907],{"type":48,"tag":73,"props":1908,"children":1910},{"className":1909},[],[1911],{"type":54,"value":1711},{"type":48,"tag":471,"props":1913,"children":1914},{},[1915],{"type":54,"value":1716},{"type":48,"tag":434,"props":1917,"children":1918},{},[1919,1923,1932,1940],{"type":48,"tag":471,"props":1920,"children":1921},{},[1922],{"type":54,"value":94},{"type":48,"tag":471,"props":1924,"children":1925},{},[1926],{"type":48,"tag":73,"props":1927,"children":1929},{"className":1928},[],[1930],{"type":54,"value":1931},"inference.checkpoint",{"type":48,"tag":471,"props":1933,"children":1934},{},[1935],{"type":48,"tag":73,"props":1936,"children":1938},{"className":1937},[],[1939],{"type":54,"value":1741},{"type":48,"tag":471,"props":1941,"children":1942},{},[1943],{"type":54,"value":1746},{"type":48,"tag":434,"props":1945,"children":1946},{},[1947,1951,1960,1968],{"type":48,"tag":471,"props":1948,"children":1949},{},[1950],{"type":54,"value":94},{"type":48,"tag":471,"props":1952,"children":1953},{},[1954],{"type":48,"tag":73,"props":1955,"children":1957},{"className":1956},[],[1958],{"type":54,"value":1959},"inference.trt_engine",{"type":48,"tag":471,"props":1961,"children":1962},{},[1963],{"type":48,"tag":73,"props":1964,"children":1966},{"className":1965},[],[1967],{"type":54,"value":1741},{"type":48,"tag":471,"props":1969,"children":1970},{},[1971],{"type":54,"value":1746},{"type":48,"tag":434,"props":1973,"children":1974},{},[1975,1979,1987,1995],{"type":48,"tag":471,"props":1976,"children":1977},{},[1978],{"type":54,"value":94},{"type":48,"tag":471,"props":1980,"children":1981},{},[1982],{"type":48,"tag":73,"props":1983,"children":1985},{"className":1984},[],[1986],{"type":54,"value":1762},{"type":48,"tag":471,"props":1988,"children":1989},{},[1990],{"type":48,"tag":73,"props":1991,"children":1993},{"className":1992},[],[1994],{"type":54,"value":1771},{"type":48,"tag":471,"props":1996,"children":1997},{},[1998],{"type":54,"value":1776},{"type":48,"tag":434,"props":2000,"children":2001},{},[2002,2006,2014,2022],{"type":48,"tag":471,"props":2003,"children":2004},{},[2005],{"type":54,"value":576},{"type":48,"tag":471,"props":2007,"children":2008},{},[2009],{"type":48,"tag":73,"props":2010,"children":2012},{"className":2011},[],[2013],{"type":54,"value":1702},{"type":48,"tag":471,"props":2015,"children":2016},{},[2017],{"type":48,"tag":73,"props":2018,"children":2020},{"className":2019},[],[2021],{"type":54,"value":1711},{"type":48,"tag":471,"props":2023,"children":2024},{},[2025],{"type":54,"value":1716},{"type":48,"tag":434,"props":2027,"children":2028},{},[2029,2033,2041,2049],{"type":48,"tag":471,"props":2030,"children":2031},{},[2032],{"type":54,"value":576},{"type":48,"tag":471,"props":2034,"children":2035},{},[2036],{"type":48,"tag":73,"props":2037,"children":2039},{"className":2038},[],[2040],{"type":54,"value":1762},{"type":48,"tag":471,"props":2042,"children":2043},{},[2044],{"type":48,"tag":73,"props":2045,"children":2047},{"className":2046},[],[2048],{"type":54,"value":1771},{"type":48,"tag":471,"props":2050,"children":2051},{},[2052],{"type":54,"value":1776},{"type":48,"tag":434,"props":2054,"children":2055},{},[2056,2060,2069,2078],{"type":48,"tag":471,"props":2057,"children":2058},{},[2059],{"type":54,"value":576},{"type":48,"tag":471,"props":2061,"children":2062},{},[2063],{"type":48,"tag":73,"props":2064,"children":2066},{"className":2065},[],[2067],{"type":54,"value":2068},"train.pretrained_model_path",{"type":48,"tag":471,"props":2070,"children":2071},{},[2072],{"type":48,"tag":73,"props":2073,"children":2075},{"className":2074},[],[2076],{"type":54,"value":2077},"ptm_if_no_resume_model",{"type":48,"tag":471,"props":2079,"children":2080},{},[2081],{"type":54,"value":2082},"PTM when no resume checkpoint exists",{"type":48,"tag":434,"props":2084,"children":2085},{},[2086,2090,2099,2108],{"type":48,"tag":471,"props":2087,"children":2088},{},[2089],{"type":54,"value":576},{"type":48,"tag":471,"props":2091,"children":2092},{},[2093],{"type":48,"tag":73,"props":2094,"children":2096},{"className":2095},[],[2097],{"type":54,"value":2098},"train.resume_training_checkpoint_path",{"type":48,"tag":471,"props":2100,"children":2101},{},[2102],{"type":48,"tag":73,"props":2103,"children":2105},{"className":2104},[],[2106],{"type":54,"value":2107},"resume_model",{"type":48,"tag":471,"props":2109,"children":2110},{},[2111],{"type":54,"value":2112},"model file inferred from the current job results folder",{"type":48,"tag":57,"props":2114,"children":2115},{},[2116,2118,2123,2124,2130,2132,2138,2140,2145],{"type":54,"value":2117},"For ",{"type":48,"tag":73,"props":2119,"children":2121},{"className":2120},[],[2122],{"type":54,"value":1741},{"type":54,"value":1485},{"type":48,"tag":73,"props":2125,"children":2127},{"className":2126},[],[2128],{"type":54,"value":2129},"parent_model_folder",{"type":54,"value":2131},", pass the upstream train\u002Fexport\u002FAutoML child job id as ",{"type":48,"tag":73,"props":2133,"children":2135},{"className":2134},[],[2136],{"type":54,"value":2137},"parent_job_id",{"type":54,"value":2139},". 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":48,"tag":73,"props":2141,"children":2143},{"className":2142},[],[2144],{"type":54,"value":1624},{"type":54,"value":2146}," and do not patch generated runner scripts to guess checkpoint paths.",{"type":48,"tag":130,"props":2148,"children":2150},{"id":2149},"deployment",[2151],{"type":54,"value":2152},"Deployment",{"type":48,"tag":383,"props":2154,"children":2155},{},[2156],{"type":48,"tag":387,"props":2157,"children":2158},{},[2159],{"type":48,"tag":2160,"props":2161,"children":2162},"a",{"href":102},[2163],{"type":54,"value":2164},"tao-deploy-optical-inspection",{"type":48,"tag":2166,"props":2167,"children":2168},"style",{},[2169],{"type":54,"value":2170},"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":2172,"total":2274},[2173,2190,2202,2216,2228,2245,2260],{"slug":2174,"name":2174,"fn":2175,"description":2176,"org":2177,"tags":2178,"stars":23,"repoUrl":24,"updatedAt":2189},"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},[2179,2182,2185,2186],{"name":2180,"slug":2181,"type":15},"Data Analysis","data-analysis",{"name":2183,"slug":2184,"type":15},"Data Engineering","data-engineering",{"name":9,"slug":8,"type":15},{"name":2187,"slug":2188,"type":15},"Performance","performance","2026-07-14T05:28:43.176466",{"slug":2191,"name":2191,"fn":2192,"description":2193,"org":2194,"tags":2195,"stars":23,"repoUrl":24,"updatedAt":2201},"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},[2196,2197,2200],{"name":2152,"slug":2149,"type":15},{"name":2198,"slug":2199,"type":15},"Infrastructure","infrastructure",{"name":9,"slug":8,"type":15},"2026-07-14T05:29:06.667109",{"slug":2203,"name":2203,"fn":2204,"description":2205,"org":2206,"tags":2207,"stars":23,"repoUrl":24,"updatedAt":2215},"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},[2208,2211,2212],{"name":2209,"slug":2210,"type":15},"Agents","agents",{"name":9,"slug":8,"type":15},{"name":2213,"slug":2214,"type":15},"Research","research","2026-07-14T05:28:06.816956",{"slug":2217,"name":2217,"fn":2218,"description":2219,"org":2220,"tags":2221,"stars":23,"repoUrl":24,"updatedAt":2227},"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},[2222,2223,2224],{"name":2180,"slug":2181,"type":15},{"name":9,"slug":8,"type":15},{"name":2225,"slug":2226,"type":15},"Testing","testing","2026-07-17T05:29:03.913266",{"slug":2229,"name":2229,"fn":2230,"description":2231,"org":2232,"tags":2233,"stars":23,"repoUrl":24,"updatedAt":2244},"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},[2234,2237,2240,2241],{"name":2235,"slug":2236,"type":15},"Automation","automation",{"name":2238,"slug":2239,"type":15},"Imaging","imaging",{"name":9,"slug":8,"type":15},{"name":2242,"slug":2243,"type":15},"Video","video","2026-07-17T05:28:53.905004",{"slug":2246,"name":2246,"fn":2247,"description":2248,"org":2249,"tags":2250,"stars":23,"repoUrl":24,"updatedAt":2259},"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},[2251,2252,2255,2256],{"name":2152,"slug":2149,"type":15},{"name":2253,"slug":2254,"type":15},"Docker","docker",{"name":9,"slug":8,"type":15},{"name":2257,"slug":2258,"type":15},"Operations","operations","2026-07-17T05:28:56.913999",{"slug":2261,"name":2261,"fn":2262,"description":2263,"org":2264,"tags":2265,"stars":23,"repoUrl":24,"updatedAt":2273},"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},[2266,2267,2270],{"name":9,"slug":8,"type":15},{"name":2268,"slug":2269,"type":15},"Quantum Computing","quantum-computing",{"name":2271,"slug":2272,"type":15},"Simulation","simulation","2026-07-14T05:26:58.898253",305,{"items":2276,"total":2424},[2277,2295,2310,2321,2333,2347,2360,2372,2383,2392,2406,2415],{"slug":2278,"name":2278,"fn":2279,"description":2280,"org":2281,"tags":2282,"stars":2292,"repoUrl":2293,"updatedAt":2294},"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},[2283,2286,2289],{"name":2284,"slug":2285,"type":15},"Documentation","documentation",{"name":2287,"slug":2288,"type":15},"MCP","mcp",{"name":2290,"slug":2291,"type":15},"Search","search",21777,"https:\u002F\u002Fgithub.com\u002FNVIDIA\u002FNemoClaw","2026-07-20T06:00:01.461044",{"slug":2296,"name":2296,"fn":2297,"description":2298,"org":2299,"tags":2300,"stars":2307,"repoUrl":2308,"updatedAt":2309},"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},[2301,2304,2305],{"name":2302,"slug":2303,"type":15},"Containers","containers",{"name":2152,"slug":2149,"type":15},{"name":2306,"slug":744,"type":15},"Python",17049,"https:\u002F\u002Fgithub.com\u002FNVIDIA\u002FMegatron-LM","2026-07-27T06:06:11.249662",{"slug":2311,"name":2311,"fn":2312,"description":2313,"org":2314,"tags":2315,"stars":2307,"repoUrl":2308,"updatedAt":2320},"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},[2316,2319],{"name":2317,"slug":2318,"type":15},"CI\u002FCD","ci-cd",{"name":2152,"slug":2149,"type":15},"2026-07-14T05:25:59.97109",{"slug":2322,"name":2322,"fn":2323,"description":2324,"org":2325,"tags":2326,"stars":2307,"repoUrl":2308,"updatedAt":2332},"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},[2327,2328,2329],{"name":2317,"slug":2318,"type":15},{"name":2152,"slug":2149,"type":15},{"name":2330,"slug":2331,"type":15},"GitHub","github","2026-07-27T06:06:12.278222",{"slug":2334,"name":2334,"fn":2335,"description":2336,"org":2337,"tags":2338,"stars":2307,"repoUrl":2308,"updatedAt":2346},"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},[2339,2342,2343],{"name":2340,"slug":2341,"type":15},"Debugging","debugging",{"name":2330,"slug":2331,"type":15},{"name":2344,"slug":2345,"type":15},"Triage","triage","2026-07-14T05:25:57.442089",{"slug":2348,"name":2348,"fn":2349,"description":2350,"org":2351,"tags":2352,"stars":2307,"repoUrl":2308,"updatedAt":2359},"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},[2353,2356],{"name":2354,"slug":2355,"type":15},"Best Practices","best-practices",{"name":2357,"slug":2358,"type":15},"Code Analysis","code-analysis","2026-07-14T05:25:56.18433",{"slug":2361,"name":2361,"fn":2362,"description":2363,"org":2364,"tags":2365,"stars":2307,"repoUrl":2308,"updatedAt":2371},"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},[2366,2367,2370],{"name":17,"slug":18,"type":15},{"name":2368,"slug":2369,"type":15},"Migration","migration",{"name":9,"slug":8,"type":15},"2026-07-17T06:07:11.777011",{"slug":2373,"name":2373,"fn":2374,"description":2375,"org":2376,"tags":2377,"stars":2307,"repoUrl":2308,"updatedAt":2382},"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},[2378,2381],{"name":2379,"slug":2380,"type":15},"QA","qa",{"name":2225,"slug":2226,"type":15},"2026-07-14T05:25:53.673039",{"slug":2384,"name":2384,"fn":2385,"description":2386,"org":2387,"tags":2388,"stars":2307,"repoUrl":2308,"updatedAt":2391},"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},[2389,2390],{"name":2152,"slug":2149,"type":15},{"name":2198,"slug":2199,"type":15},"2026-07-14T05:25:49.362534",{"slug":2393,"name":2393,"fn":2394,"description":2395,"org":2396,"tags":2397,"stars":2307,"repoUrl":2308,"updatedAt":2405},"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},[2398,2401,2402],{"name":2399,"slug":2400,"type":15},"Code Review","code-review",{"name":2330,"slug":2331,"type":15},{"name":2403,"slug":2404,"type":15},"Pull Requests","pull-requests","2026-07-14T05:26:01.226578",{"slug":2407,"name":2407,"fn":2408,"description":2409,"org":2410,"tags":2411,"stars":2307,"repoUrl":2308,"updatedAt":2414},"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},[2412,2413],{"name":2379,"slug":2380,"type":15},{"name":2225,"slug":2226,"type":15},"2026-07-14T05:25:54.928983",{"slug":2416,"name":2416,"fn":2417,"description":2418,"org":2419,"tags":2420,"stars":2307,"repoUrl":2308,"updatedAt":2423},"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},[2421,2422],{"name":2235,"slug":2236,"type":15},{"name":2317,"slug":2318,"type":15},"2026-07-30T05:29:03.275638",496]