[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill-nvidia-tao-train-ocdnet":3,"mdc-vhii3y-key":31,"related-repo-nvidia-tao-train-ocdnet":2688,"related-org-nvidia-tao-train-ocdnet":2792},{"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-ocdnet","train OCDNet for scene text detection","OCDNet for scene text detection. Detects arbitrary-oriented text regions in natural images using a differentiable binarization approach. Use when training, evaluating, exporting, pruning, quantizing, retraining, or running inference for a TAO OCDNet model. Trigger phrases include \"train OCDNet\", \"scene text detection\", \"arbitrary-oriented text boxes\", \"differentiable binarization detector\".",{"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","2026-07-14T05:30:24.137271","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-ocdnet","---\nname: tao-train-ocdnet\ndescription: OCDNet for scene text detection. Detects arbitrary-oriented text regions in natural images using a\n  differentiable binarization approach. Use when training, evaluating, exporting, pruning, quantizing, retraining, or running\n  inference for a TAO OCDNet model. Trigger phrases include \"train OCDNet\", \"scene text detection\", \"arbitrary-oriented text\n  boxes\", \"differentiable binarization detector\".\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- text\n- detection\n---\n\n# OCDNet\n\nOCDNet for scene text detection. Detects arbitrary-oriented text regions in natural images using a differentiable binarization approach.\n\nSet `model.pretrained_model_path` for pretrained weights.\n\nFor TAO Deploy TensorRT actions (`gen_trt_engine`, TensorRT `evaluate`, and TensorRT `inference`), read `references\u002Ftao-deploy-ocdnet.md` first. Deploy spec templates live in this skill's `references\u002F` folder with the `spec_template_deploy_*.yaml` prefix.\n\nThe PyT OCDNet CLI supports `train`, `evaluate`, `export`, `inference`, `prune`, `quantize`, and `default_specs`. It does not expose PyT-side `retrain` or `gen_trt_engine` subcommands. The model skill exposes `retrain` by running `ocdnet train` with `model.load_pruned_graph: true` and `model.pruned_graph_path`. Resume from an epoch checkpoint uses `ocdnet train` plus `train.resume_training_checkpoint_path`. TensorRT engine generation is owned by the deploy workflow.\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\nFor AutoML train, use `train_loss_epoch` or `train_loss` as the optimization\nmetric with `direction=minimize`. The Lightning progress log emits\n`train_loss_epoch`, and TAO `status.json` records the same final value under\n`train_loss`. For one-epoch local AutoML smoke runs, set\n`train.lr_scheduler.args.warmup_epoch: 0`; leaving warmup equal to the epoch\nbudget causes the trainer to fail before a recommendation can report a metric.\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:** ocdnet\n- **Formats:** default\n- **Monitoring metric:** hmean\n\n### Per-Action Dataset Requirements\n\n| Action | Spec Key | Source | Runtime value | List? |\n|---|---|---|---|---|\n| evaluate | dataset.validate_dataset.data_path | eval_dataset | extracted validation split folder with `img\u002F` and `gt\u002F` | Yes |\n| inference | inference.input_folder | inference_dataset or eval_dataset | extracted image folder | No |\n| prune | dataset.validate_dataset.data_path | eval_dataset | extracted validation split folder with `img\u002F` and `gt\u002F` | Yes |\n| quantize | dataset.train_dataset.data_path | train_datasets | extracted train split folder with `img\u002F` and `gt\u002F` | Yes |\n| quantize | dataset.validate_dataset.data_path | eval_dataset | extracted validation split folder with `img\u002F` and `gt\u002F` | Yes |\n| quantize | dataset.quant_calibration_dataset.images_dir | train_datasets or calibration_dataset | extracted calibration image folder | No |\n| train | dataset.train_dataset.data_path | train_datasets | extracted train split folder with `img\u002F` and `gt\u002F` | Yes |\n| train | dataset.validate_dataset.data_path | eval_dataset | extracted validation split folder with `img\u002F` and `gt\u002F` | Yes |\n| retrain | dataset.train_dataset.data_path | train_datasets | extracted train split folder with `img\u002F` and `gt\u002F` | Yes |\n| retrain | dataset.validate_dataset.data_path | eval_dataset | extracted validation split folder with `img\u002F` and `gt\u002F` | Yes |\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`. OCDNet does not unpack dataset archives at runtime. If the source is `train.tar.gz`, `test.tar.gz`, or `img.tar.gz`, extract it first and pass the split folder or image folder into the spec. The split folder must contain `img\u002F` and `gt\u002F`; alternatively, pass a UTF-8 datalist text file whose lines map image paths to label paths.\n\n```python\nTRAIN_ROOT = \"\u002Fpath\u002Fto\u002Fextracted\u002Ftrain\"\nEVAL_ROOT = \"\u002Fpath\u002Fto\u002Fextracted\u002Ftest\"\nINFER_IMG_DIR = \"\u002Fpath\u002Fto\u002Fextracted\u002Ftest\u002Fimg\"\nCALIB_IMG_DIR = \"\u002Fpath\u002Fto\u002Fextracted\u002Ftrain\u002Fimg\"\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.train_dataset.loader.batch_size\": 16,\n    \"dataset.train_dataset.data_path\": [TRAIN_ROOT],\n    \"dataset.validate_dataset.data_path\": [EVAL_ROOT],\n}\n```\n\n**evaluate (mandatory data sources):**\n```python\n{\n    \"evaluate.checkpoint\": \"\u003Cselected train\u002FAutoML checkpoint>\",\n    \"dataset.validate_dataset.data_path\": [EVAL_ROOT],\n}\n```\n\n**inference (mandatory data sources):**\n```python\n{\n    \"inference.checkpoint\": \"\u003Cselected train\u002FAutoML checkpoint>\",\n    \"inference.input_folder\": INFER_IMG_DIR,\n}\n```\n\n**prune (mandatory data sources):**\n```python\n{\n    \"prune.checkpoint\": \"\u003Cselected train\u002FAutoML checkpoint>\",\n    \"dataset.validate_dataset.data_path\": [EVAL_ROOT],\n}\n```\n\n**quantize (mandatory data sources):**\n```python\n{\n    \"quantize.model_path\": \"\u003Cselected train checkpoint or exported ONNX>\",\n    \"dataset.train_dataset.data_path\": [TRAIN_ROOT],\n    \"dataset.validate_dataset.data_path\": [EVAL_ROOT],\n    \"dataset.quant_calibration_dataset.images_dir\": CALIB_IMG_DIR,\n}\n```\n\n**resume training (mandatory data sources):**\n```python\n{\n    \"train.resume_training_checkpoint_path\": \"\u003Cexact model_epoch checkpoint>\",\n    \"dataset.train_dataset.data_path\": [TRAIN_ROOT],\n    \"dataset.validate_dataset.data_path\": [EVAL_ROOT],\n}\n```\n\n**retrain from prune output (mandatory data sources):**\n```python\n{\n    \"model.load_pruned_graph\": True,\n    \"model.pruned_graph_path\": \"\u003Cselected prune output>\",\n    \"dataset.train_dataset.data_path\": [TRAIN_ROOT],\n    \"dataset.validate_dataset.data_path\": [EVAL_ROOT],\n}\n```\n\n**default_specs:**\n```python\n{\n    \"results_dir\": \"\u003Cwritable output directory>\",\n}\n```\n## Eval Dataset\n\nOptional. Test dataset provided as separate tarball.\n\n## Important Parameters\n\n- **model.backbone**: Default deformable_resnet18. Deformable convolutions improve text region detection for irregular text.\n- **train.optimizer.args.lr**: Learning rate. Default 0.001 (Adam).\n- **postprocess.thresh**: Binarization threshold for text region extraction.\n- **postprocess.box_thresh**: Box confidence threshold for filtering detections.\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| `train.distributed_strategy` | `ddp`, `fsdp`, or `deepspeed_stage_3_offload` | `ddp` |\n\n- `ddp` with activation checkpointing: `find_unused_parameters=False`\n- `ddp` without: `find_unused_parameters=True`\n- `fsdp` forces FP16\n- **`deepspeed_stage_3_offload`** is uniquely supported for OCDNet (forces FP16)\n- FAN backbones auto-enable `sync_batchnorm`\n\n## Hardware\n\nMinimum 1 GPU(s), recommended 1 GPU(s). 8GB+ VRAM per GPU. OCDNet is lightweight. Single GPU is sufficient for most datasets.\n\n## Error Patterns\n\n**Low detection rate**: Tune postprocess.thresh and box_thresh. Default thresholds may be too aggressive for some datasets.\n\n**One-epoch smoke train with default scheduler**: `train.num_epochs` must not equal `train.lr_scheduler.args.warmup_epoch`. For one-epoch validation, set `warmup_epoch: 0`; for normal starter runs, keep `num_epochs > warmup_epoch`.\n\n**Archive passed as dataset path**: `dataset.*.data_path` is not an archive path for OCDNet. Passing `train.tar.gz` or `test.tar.gz` directly causes the dataloader to open the gzip as a UTF-8 datalist. Extract the archive and pass the split folder containing `img\u002F` and `gt\u002F`, or pass a real UTF-8 datalist file.\n\n**Quantize checkpoint type**: Do not pass `model_best.pth` to the PyTorch quantize path. Some older PyT runtimes wrote `model_best.pth` without full Lightning checkpoint metadata. The default `torchao` quantize path should use the intended full `model_epoch_\u003Cepoch>_step_\u003Cstep>.pth` checkpoint and write `quantized_model_torchao.pth`.\n\n**Default specs output directory**: `ocdnet default_specs` requires a writable `results_dir` override, for example `results_dir=\u002Fworkspace\u002Frun\u002Fresults\u002Fdefault_specs`.\n\n## Checkpoint Handoff\n\nOCDNet train writes `model_best.pth` plus full Lightning epoch checkpoints such as `model_epoch_001_step_00046.pth`; it may also write `ocd_model_latest.pth` as a latest symlink. Use `model_best.pth` for `evaluate.checkpoint`, `inference.checkpoint`, `export.checkpoint`, and `prune.checkpoint` when the user asks for the best checkpoint. Use a specific `model_epoch_\u003Cepoch>_step_\u003Cstep>.pth` for `train.resume_training_checkpoint_path` and for any action that explicitly needs a full Lightning checkpoint. Prune writes artifacts such as `pruned_\u003Cch_sparsity>.pth`; use the exact pruned `.pth` artifact for `model.pruned_graph_path` when retraining from a pruned graph. Use a latest checkpoint only when the user explicitly asks for latest.\n\nIf quantize is retried with a PyTorch backend, resolve the full `model_epoch_\u003Cepoch>_step_\u003Cstep>.pth` that corresponds to the intended best epoch or requested epoch; do not pass `model_best.pth` to the PyTorch quantize path. If quantize is retried with `modelopt.onnx`, pass the exported ONNX as `quantize.model_path` and verify that the runtime image actually contains `modelopt.onnx.quantization`.\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\nModel handoff mappings:\n\n| Action | Spec Field | Inference Function | Meaning |\n|---|---|---|---|\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 | `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 | `inference.checkpoint` | `parent_model` | model file inferred from the parent job results folder |\n| inference | `results_dir` | `output_dir` | current job results directory |\n| prune | `prune.checkpoint` | `parent_model` | model file inferred from the parent job results folder |\n| prune | `results_dir` | `output_dir` | current job results directory |\n| quantize | `quantize.model_path` | `parent_model` | model file inferred from the parent job results folder |\n| quantize | `results_dir` | `output_dir` | current job results directory |\n| retrain from prune | `model.pruned_graph_path` | `parent_model` | exact pruned model file inferred from the parent prune results folder |\n| retrain from prune | `results_dir` | `output_dir` | current job results directory |\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\n## Deployment\n\n- [tao-deploy-ocdnet](references\u002Ftao-deploy-ocdnet.md)\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],"text","detection",{"type":42,"children":43},"root",[44,52,58,72,125,241,248,339,345,454,538,544,580,587,964,970,1026,1075,1083,1167,1175,1211,1219,1256,1264,1300,1308,1359,1367,1410,1418,1469,1477,1506,1512,1517,1523,1566,1572,1589,1705,1774,1780,1785,1791,1801,1843,1886,1934,1966,1972,2072,2114,2120,2149,2154,2630,2664,2670,2682],{"type":45,"tag":46,"props":47,"children":49},"element","h1",{"id":48},"ocdnet",[50],{"type":39,"value":51},"OCDNet",{"type":45,"tag":53,"props":54,"children":55},"p",{},[56],{"type":39,"value":57},"OCDNet for scene text detection. Detects arbitrary-oriented text regions in natural images using a differentiable binarization approach.",{"type":45,"tag":53,"props":59,"children":60},{},[61,63,70],{"type":39,"value":62},"Set ",{"type":45,"tag":64,"props":65,"children":67},"code",{"className":66},[],[68],{"type":39,"value":69},"model.pretrained_model_path",{"type":39,"value":71}," for pretrained weights.",{"type":45,"tag":53,"props":73,"children":74},{},[75,77,83,85,91,93,99,101,107,109,115,117,123],{"type":39,"value":76},"For TAO Deploy TensorRT actions (",{"type":45,"tag":64,"props":78,"children":80},{"className":79},[],[81],{"type":39,"value":82},"gen_trt_engine",{"type":39,"value":84},", TensorRT ",{"type":45,"tag":64,"props":86,"children":88},{"className":87},[],[89],{"type":39,"value":90},"evaluate",{"type":39,"value":92},", and TensorRT ",{"type":45,"tag":64,"props":94,"children":96},{"className":95},[],[97],{"type":39,"value":98},"inference",{"type":39,"value":100},"), read ",{"type":45,"tag":64,"props":102,"children":104},{"className":103},[],[105],{"type":39,"value":106},"references\u002Ftao-deploy-ocdnet.md",{"type":39,"value":108}," first. Deploy spec templates live in this skill's ",{"type":45,"tag":64,"props":110,"children":112},{"className":111},[],[113],{"type":39,"value":114},"references\u002F",{"type":39,"value":116}," folder with the ",{"type":45,"tag":64,"props":118,"children":120},{"className":119},[],[121],{"type":39,"value":122},"spec_template_deploy_*.yaml",{"type":39,"value":124}," prefix.",{"type":45,"tag":53,"props":126,"children":127},{},[128,130,136,138,143,144,150,151,156,157,163,164,170,172,178,180,186,188,193,195,200,202,208,210,216,218,224,226,231,233,239],{"type":39,"value":129},"The PyT OCDNet CLI supports ",{"type":45,"tag":64,"props":131,"children":133},{"className":132},[],[134],{"type":39,"value":135},"train",{"type":39,"value":137},", ",{"type":45,"tag":64,"props":139,"children":141},{"className":140},[],[142],{"type":39,"value":90},{"type":39,"value":137},{"type":45,"tag":64,"props":145,"children":147},{"className":146},[],[148],{"type":39,"value":149},"export",{"type":39,"value":137},{"type":45,"tag":64,"props":152,"children":154},{"className":153},[],[155],{"type":39,"value":98},{"type":39,"value":137},{"type":45,"tag":64,"props":158,"children":160},{"className":159},[],[161],{"type":39,"value":162},"prune",{"type":39,"value":137},{"type":45,"tag":64,"props":165,"children":167},{"className":166},[],[168],{"type":39,"value":169},"quantize",{"type":39,"value":171},", and ",{"type":45,"tag":64,"props":173,"children":175},{"className":174},[],[176],{"type":39,"value":177},"default_specs",{"type":39,"value":179},". It does not expose PyT-side ",{"type":45,"tag":64,"props":181,"children":183},{"className":182},[],[184],{"type":39,"value":185},"retrain",{"type":39,"value":187}," or ",{"type":45,"tag":64,"props":189,"children":191},{"className":190},[],[192],{"type":39,"value":82},{"type":39,"value":194}," subcommands. The model skill exposes ",{"type":45,"tag":64,"props":196,"children":198},{"className":197},[],[199],{"type":39,"value":185},{"type":39,"value":201}," by running ",{"type":45,"tag":64,"props":203,"children":205},{"className":204},[],[206],{"type":39,"value":207},"ocdnet train",{"type":39,"value":209}," with ",{"type":45,"tag":64,"props":211,"children":213},{"className":212},[],[214],{"type":39,"value":215},"model.load_pruned_graph: true",{"type":39,"value":217}," and ",{"type":45,"tag":64,"props":219,"children":221},{"className":220},[],[222],{"type":39,"value":223},"model.pruned_graph_path",{"type":39,"value":225},". Resume from an epoch checkpoint uses ",{"type":45,"tag":64,"props":227,"children":229},{"className":228},[],[230],{"type":39,"value":207},{"type":39,"value":232}," plus ",{"type":45,"tag":64,"props":234,"children":236},{"className":235},[],[237],{"type":39,"value":238},"train.resume_training_checkpoint_path",{"type":39,"value":240},". TensorRT engine generation is owned by the deploy workflow.",{"type":45,"tag":242,"props":243,"children":245},"h2",{"id":244},"dataclass-schemas",[246],{"type":39,"value":247},"Dataclass Schemas",{"type":45,"tag":53,"props":249,"children":250},{},[251,253,259,261,267,269,275,277,283,285,291,293,299,301,307,308,314,316,322,323,329,331,337],{"type":39,"value":252},"Generated TAO Core schemas are packaged in ",{"type":45,"tag":64,"props":254,"children":256},{"className":255},[],[257],{"type":39,"value":258},"schemas\u002F\u003Caction>.schema.json",{"type":39,"value":260},", with ",{"type":45,"tag":64,"props":262,"children":264},{"className":263},[],[265],{"type":39,"value":266},"schemas\u002Fmanifest.json",{"type":39,"value":268}," listing available actions. Each generated schema also emits ",{"type":45,"tag":64,"props":270,"children":272},{"className":271},[],[273],{"type":39,"value":274},"references\u002Fspec_template_\u003Caction>.yaml",{"type":39,"value":276}," from the schema top-level ",{"type":45,"tag":64,"props":278,"children":280},{"className":279},[],[281],{"type":39,"value":282},"default",{"type":39,"value":284}," field. AutoML enablement is declared at the model layer in ",{"type":45,"tag":64,"props":286,"children":288},{"className":287},[],[289],{"type":39,"value":290},"references\u002Fskill_info.yaml",{"type":39,"value":292}," via ",{"type":45,"tag":64,"props":294,"children":296},{"className":295},[],[297],{"type":39,"value":298},"automl_enabled",{"type":39,"value":300},". Runnable AutoML still requires ",{"type":45,"tag":64,"props":302,"children":304},{"className":303},[],[305],{"type":39,"value":306},"schemas\u002Ftrain.schema.json",{"type":39,"value":217},{"type":45,"tag":64,"props":309,"children":311},{"className":310},[],[312],{"type":39,"value":313},"references\u002Fspec_template_train.yaml",{"type":39,"value":315}," to exist and parse. Use the packaged train schema for ",{"type":45,"tag":64,"props":317,"children":319},{"className":318},[],[320],{"type":39,"value":321},"automl_default_parameters",{"type":39,"value":137},{"type":45,"tag":64,"props":324,"children":326},{"className":325},[],[327],{"type":39,"value":328},"automl_disabled_parameters",{"type":39,"value":330},", defaults, min\u002Fmax bounds, enums, option weights, math conditions, dependencies, and popular parameters. Do not expect ",{"type":45,"tag":64,"props":332,"children":334},{"className":333},[],[335],{"type":39,"value":336},"~\u002Ftao-core",{"type":39,"value":338}," at runtime; maintainers regenerate schemas\u002Ftemplates before packaging the skill bank.",{"type":45,"tag":242,"props":340,"children":342},{"id":341},"train-action-policy",[343],{"type":39,"value":344},"Train Action Policy",{"type":45,"tag":53,"props":346,"children":347},{},[348,350,355,357,363,365,371,373,379,381,387,389,395,397,402,403,409,411,416,417,422,424,430,432,438,440,445,447,452],{"type":39,"value":349},"This model is AutoML-enabled at the model layer. Before handling any train-stage request, read ",{"type":45,"tag":64,"props":351,"children":353},{"className":352},[],[354],{"type":39,"value":290},{"type":39,"value":356}," and resolve the run override from either an explicit ",{"type":45,"tag":64,"props":358,"children":360},{"className":359},[],[361],{"type":39,"value":362},"automl_policy",{"type":39,"value":364}," value or the user's workflow request. Use ",{"type":45,"tag":64,"props":366,"children":368},{"className":367},[],[369],{"type":39,"value":370},"automl_policy: on",{"type":39,"value":372}," by default and only expose ",{"type":45,"tag":64,"props":374,"children":376},{"className":375},[],[377],{"type":39,"value":378},"on",{"type":39,"value":380}," \u002F ",{"type":45,"tag":64,"props":382,"children":384},{"className":383},[],[385],{"type":39,"value":386},"off",{"type":39,"value":388}," in new launch prompts. Treat phrases like \"turn off AutoML\", \"disable AutoML\", \"no HPO\", or \"plain training\" as ",{"type":45,"tag":64,"props":390,"children":392},{"className":391},[],[393],{"type":39,"value":394},"automl_policy: off",{"type":39,"value":396}," for this run only. When ",{"type":45,"tag":64,"props":398,"children":400},{"className":399},[],[401],{"type":39,"value":370},{"type":39,"value":137},{"type":45,"tag":64,"props":404,"children":406},{"className":405},[],[407],{"type":39,"value":408},"automl_enabled: true",{"type":39,"value":410},", and both ",{"type":45,"tag":64,"props":412,"children":414},{"className":413},[],[415],{"type":39,"value":306},{"type":39,"value":217},{"type":45,"tag":64,"props":418,"children":420},{"className":419},[],[421],{"type":39,"value":313},{"type":39,"value":423}," are packaged, route the train action through ",{"type":45,"tag":64,"props":425,"children":427},{"className":426},[],[428],{"type":39,"value":429},"tao-skill-bank:tao-run-automl",{"type":39,"value":431}," by default with this model's ",{"type":45,"tag":64,"props":433,"children":435},{"className":434},[],[436],{"type":39,"value":437},"skill_dir",{"type":39,"value":439},". Preserve workflow\u002Fapplication overrides for datasets, specs, output directories, GPU\u002Fplatform settings, parent checkpoints, and ",{"type":45,"tag":64,"props":441,"children":443},{"className":442},[],[444],{"type":39,"value":362},{"type":39,"value":446},". Use direct model training only when ",{"type":45,"tag":64,"props":448,"children":450},{"className":449},[],[451],{"type":39,"value":394},{"type":39,"value":453}," 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":53,"props":455,"children":456},{},[457,459,465,466,472,474,480,482,487,489,495,497,502,504,510,512,517,518,523,524,529,531,536],{"type":39,"value":458},"For AutoML train, use ",{"type":45,"tag":64,"props":460,"children":462},{"className":461},[],[463],{"type":39,"value":464},"train_loss_epoch",{"type":39,"value":187},{"type":45,"tag":64,"props":467,"children":469},{"className":468},[],[470],{"type":39,"value":471},"train_loss",{"type":39,"value":473}," as the optimization\nmetric with ",{"type":45,"tag":64,"props":475,"children":477},{"className":476},[],[478],{"type":39,"value":479},"direction=minimize",{"type":39,"value":481},". The Lightning progress log emits\n",{"type":45,"tag":64,"props":483,"children":485},{"className":484},[],[486],{"type":39,"value":464},{"type":39,"value":488},", and TAO ",{"type":45,"tag":64,"props":490,"children":492},{"className":491},[],[493],{"type":39,"value":494},"status.json",{"type":39,"value":496}," records the same final value under\n",{"type":45,"tag":64,"props":498,"children":500},{"className":499},[],[501],{"type":39,"value":471},{"type":39,"value":503},". For one-epoch local AutoML smoke runs, set\n",{"type":45,"tag":64,"props":505,"children":507},{"className":506},[],[508],{"type":39,"value":509},"train.lr_scheduler.args.warmup_epoch: 0",{"type":39,"value":511},"; leaving warmup equal to the epoch\nbudget causes the trainer to fail before a recommendation can report a metric.\nNon-train actions such as ",{"type":45,"tag":64,"props":513,"children":515},{"className":514},[],[516],{"type":39,"value":90},{"type":39,"value":137},{"type":45,"tag":64,"props":519,"children":521},{"className":520},[],[522],{"type":39,"value":98},{"type":39,"value":137},{"type":45,"tag":64,"props":525,"children":527},{"className":526},[],[528],{"type":39,"value":149},{"type":39,"value":530},", and deploy flows stay in this model skill. The per-run ",{"type":45,"tag":64,"props":532,"children":534},{"className":533},[],[535],{"type":39,"value":362},{"type":39,"value":537}," override does not change model metadata.",{"type":45,"tag":242,"props":539,"children":541},{"id":540},"training-requirements",[542],{"type":39,"value":543},"Training Requirements",{"type":45,"tag":545,"props":546,"children":547},"ul",{},[548,560,570],{"type":45,"tag":549,"props":550,"children":551},"li",{},[552,558],{"type":45,"tag":553,"props":554,"children":555},"strong",{},[556],{"type":39,"value":557},"Dataset type:",{"type":39,"value":559}," ocdnet",{"type":45,"tag":549,"props":561,"children":562},{},[563,568],{"type":45,"tag":553,"props":564,"children":565},{},[566],{"type":39,"value":567},"Formats:",{"type":39,"value":569}," default",{"type":45,"tag":549,"props":571,"children":572},{},[573,578],{"type":45,"tag":553,"props":574,"children":575},{},[576],{"type":39,"value":577},"Monitoring metric:",{"type":39,"value":579}," hmean",{"type":45,"tag":581,"props":582,"children":584},"h3",{"id":583},"per-action-dataset-requirements",[585],{"type":39,"value":586},"Per-Action Dataset Requirements",{"type":45,"tag":588,"props":589,"children":590},"table",{},[591,625],{"type":45,"tag":592,"props":593,"children":594},"thead",{},[595],{"type":45,"tag":596,"props":597,"children":598},"tr",{},[599,605,610,615,620],{"type":45,"tag":600,"props":601,"children":602},"th",{},[603],{"type":39,"value":604},"Action",{"type":45,"tag":600,"props":606,"children":607},{},[608],{"type":39,"value":609},"Spec Key",{"type":45,"tag":600,"props":611,"children":612},{},[613],{"type":39,"value":614},"Source",{"type":45,"tag":600,"props":616,"children":617},{},[618],{"type":39,"value":619},"Runtime value",{"type":45,"tag":600,"props":621,"children":622},{},[623],{"type":39,"value":624},"List?",{"type":45,"tag":626,"props":627,"children":628},"tbody",{},[629,670,697,731,768,802,828,862,896,930],{"type":45,"tag":596,"props":630,"children":631},{},[632,637,642,647,665],{"type":45,"tag":633,"props":634,"children":635},"td",{},[636],{"type":39,"value":90},{"type":45,"tag":633,"props":638,"children":639},{},[640],{"type":39,"value":641},"dataset.validate_dataset.data_path",{"type":45,"tag":633,"props":643,"children":644},{},[645],{"type":39,"value":646},"eval_dataset",{"type":45,"tag":633,"props":648,"children":649},{},[650,652,658,659],{"type":39,"value":651},"extracted validation split folder with ",{"type":45,"tag":64,"props":653,"children":655},{"className":654},[],[656],{"type":39,"value":657},"img\u002F",{"type":39,"value":217},{"type":45,"tag":64,"props":660,"children":662},{"className":661},[],[663],{"type":39,"value":664},"gt\u002F",{"type":45,"tag":633,"props":666,"children":667},{},[668],{"type":39,"value":669},"Yes",{"type":45,"tag":596,"props":671,"children":672},{},[673,677,682,687,692],{"type":45,"tag":633,"props":674,"children":675},{},[676],{"type":39,"value":98},{"type":45,"tag":633,"props":678,"children":679},{},[680],{"type":39,"value":681},"inference.input_folder",{"type":45,"tag":633,"props":683,"children":684},{},[685],{"type":39,"value":686},"inference_dataset or eval_dataset",{"type":45,"tag":633,"props":688,"children":689},{},[690],{"type":39,"value":691},"extracted image folder",{"type":45,"tag":633,"props":693,"children":694},{},[695],{"type":39,"value":696},"No",{"type":45,"tag":596,"props":698,"children":699},{},[700,704,708,712,727],{"type":45,"tag":633,"props":701,"children":702},{},[703],{"type":39,"value":162},{"type":45,"tag":633,"props":705,"children":706},{},[707],{"type":39,"value":641},{"type":45,"tag":633,"props":709,"children":710},{},[711],{"type":39,"value":646},{"type":45,"tag":633,"props":713,"children":714},{},[715,716,721,722],{"type":39,"value":651},{"type":45,"tag":64,"props":717,"children":719},{"className":718},[],[720],{"type":39,"value":657},{"type":39,"value":217},{"type":45,"tag":64,"props":723,"children":725},{"className":724},[],[726],{"type":39,"value":664},{"type":45,"tag":633,"props":728,"children":729},{},[730],{"type":39,"value":669},{"type":45,"tag":596,"props":732,"children":733},{},[734,738,743,748,764],{"type":45,"tag":633,"props":735,"children":736},{},[737],{"type":39,"value":169},{"type":45,"tag":633,"props":739,"children":740},{},[741],{"type":39,"value":742},"dataset.train_dataset.data_path",{"type":45,"tag":633,"props":744,"children":745},{},[746],{"type":39,"value":747},"train_datasets",{"type":45,"tag":633,"props":749,"children":750},{},[751,753,758,759],{"type":39,"value":752},"extracted train split folder with ",{"type":45,"tag":64,"props":754,"children":756},{"className":755},[],[757],{"type":39,"value":657},{"type":39,"value":217},{"type":45,"tag":64,"props":760,"children":762},{"className":761},[],[763],{"type":39,"value":664},{"type":45,"tag":633,"props":765,"children":766},{},[767],{"type":39,"value":669},{"type":45,"tag":596,"props":769,"children":770},{},[771,775,779,783,798],{"type":45,"tag":633,"props":772,"children":773},{},[774],{"type":39,"value":169},{"type":45,"tag":633,"props":776,"children":777},{},[778],{"type":39,"value":641},{"type":45,"tag":633,"props":780,"children":781},{},[782],{"type":39,"value":646},{"type":45,"tag":633,"props":784,"children":785},{},[786,787,792,793],{"type":39,"value":651},{"type":45,"tag":64,"props":788,"children":790},{"className":789},[],[791],{"type":39,"value":657},{"type":39,"value":217},{"type":45,"tag":64,"props":794,"children":796},{"className":795},[],[797],{"type":39,"value":664},{"type":45,"tag":633,"props":799,"children":800},{},[801],{"type":39,"value":669},{"type":45,"tag":596,"props":803,"children":804},{},[805,809,814,819,824],{"type":45,"tag":633,"props":806,"children":807},{},[808],{"type":39,"value":169},{"type":45,"tag":633,"props":810,"children":811},{},[812],{"type":39,"value":813},"dataset.quant_calibration_dataset.images_dir",{"type":45,"tag":633,"props":815,"children":816},{},[817],{"type":39,"value":818},"train_datasets or calibration_dataset",{"type":45,"tag":633,"props":820,"children":821},{},[822],{"type":39,"value":823},"extracted calibration image folder",{"type":45,"tag":633,"props":825,"children":826},{},[827],{"type":39,"value":696},{"type":45,"tag":596,"props":829,"children":830},{},[831,835,839,843,858],{"type":45,"tag":633,"props":832,"children":833},{},[834],{"type":39,"value":135},{"type":45,"tag":633,"props":836,"children":837},{},[838],{"type":39,"value":742},{"type":45,"tag":633,"props":840,"children":841},{},[842],{"type":39,"value":747},{"type":45,"tag":633,"props":844,"children":845},{},[846,847,852,853],{"type":39,"value":752},{"type":45,"tag":64,"props":848,"children":850},{"className":849},[],[851],{"type":39,"value":657},{"type":39,"value":217},{"type":45,"tag":64,"props":854,"children":856},{"className":855},[],[857],{"type":39,"value":664},{"type":45,"tag":633,"props":859,"children":860},{},[861],{"type":39,"value":669},{"type":45,"tag":596,"props":863,"children":864},{},[865,869,873,877,892],{"type":45,"tag":633,"props":866,"children":867},{},[868],{"type":39,"value":135},{"type":45,"tag":633,"props":870,"children":871},{},[872],{"type":39,"value":641},{"type":45,"tag":633,"props":874,"children":875},{},[876],{"type":39,"value":646},{"type":45,"tag":633,"props":878,"children":879},{},[880,881,886,887],{"type":39,"value":651},{"type":45,"tag":64,"props":882,"children":884},{"className":883},[],[885],{"type":39,"value":657},{"type":39,"value":217},{"type":45,"tag":64,"props":888,"children":890},{"className":889},[],[891],{"type":39,"value":664},{"type":45,"tag":633,"props":893,"children":894},{},[895],{"type":39,"value":669},{"type":45,"tag":596,"props":897,"children":898},{},[899,903,907,911,926],{"type":45,"tag":633,"props":900,"children":901},{},[902],{"type":39,"value":185},{"type":45,"tag":633,"props":904,"children":905},{},[906],{"type":39,"value":742},{"type":45,"tag":633,"props":908,"children":909},{},[910],{"type":39,"value":747},{"type":45,"tag":633,"props":912,"children":913},{},[914,915,920,921],{"type":39,"value":752},{"type":45,"tag":64,"props":916,"children":918},{"className":917},[],[919],{"type":39,"value":657},{"type":39,"value":217},{"type":45,"tag":64,"props":922,"children":924},{"className":923},[],[925],{"type":39,"value":664},{"type":45,"tag":633,"props":927,"children":928},{},[929],{"type":39,"value":669},{"type":45,"tag":596,"props":931,"children":932},{},[933,937,941,945,960],{"type":45,"tag":633,"props":934,"children":935},{},[936],{"type":39,"value":185},{"type":45,"tag":633,"props":938,"children":939},{},[940],{"type":39,"value":641},{"type":45,"tag":633,"props":942,"children":943},{},[944],{"type":39,"value":646},{"type":45,"tag":633,"props":946,"children":947},{},[948,949,954,955],{"type":39,"value":651},{"type":45,"tag":64,"props":950,"children":952},{"className":951},[],[953],{"type":39,"value":657},{"type":39,"value":217},{"type":45,"tag":64,"props":956,"children":958},{"className":957},[],[959],{"type":39,"value":664},{"type":45,"tag":633,"props":961,"children":962},{},[963],{"type":39,"value":669},{"type":45,"tag":581,"props":965,"children":967},{"id":966},"typical-spec-overrides",[968],{"type":39,"value":969},"Typical Spec Overrides",{"type":45,"tag":53,"props":971,"children":972},{},[973,975,980,982,988,990,996,997,1003,1005,1011,1013,1018,1019,1024],{"type":39,"value":974},"Data source overrides are ",{"type":45,"tag":553,"props":976,"children":977},{},[978],{"type":39,"value":979},"mandatory for every action",{"type":39,"value":981}," — the agent MUST construct data source paths from the Per-Action Dataset Requirements table above and include them in ",{"type":45,"tag":64,"props":983,"children":985},{"className":984},[],[986],{"type":39,"value":987},"spec_overrides",{"type":39,"value":989},". OCDNet does not unpack dataset archives at runtime. If the source is ",{"type":45,"tag":64,"props":991,"children":993},{"className":992},[],[994],{"type":39,"value":995},"train.tar.gz",{"type":39,"value":137},{"type":45,"tag":64,"props":998,"children":1000},{"className":999},[],[1001],{"type":39,"value":1002},"test.tar.gz",{"type":39,"value":1004},", or ",{"type":45,"tag":64,"props":1006,"children":1008},{"className":1007},[],[1009],{"type":39,"value":1010},"img.tar.gz",{"type":39,"value":1012},", extract it first and pass the split folder or image folder into the spec. The split folder must contain ",{"type":45,"tag":64,"props":1014,"children":1016},{"className":1015},[],[1017],{"type":39,"value":657},{"type":39,"value":217},{"type":45,"tag":64,"props":1020,"children":1022},{"className":1021},[],[1023],{"type":39,"value":664},{"type":39,"value":1025},"; alternatively, pass a UTF-8 datalist text file whose lines map image paths to label paths.",{"type":45,"tag":1027,"props":1028,"children":1033},"pre",{"className":1029,"code":1030,"language":1031,"meta":1032,"style":1032},"language-python shiki shiki-themes material-theme-lighter material-theme material-theme-palenight","TRAIN_ROOT = \"\u002Fpath\u002Fto\u002Fextracted\u002Ftrain\"\nEVAL_ROOT = \"\u002Fpath\u002Fto\u002Fextracted\u002Ftest\"\nINFER_IMG_DIR = \"\u002Fpath\u002Fto\u002Fextracted\u002Ftest\u002Fimg\"\nCALIB_IMG_DIR = \"\u002Fpath\u002Fto\u002Fextracted\u002Ftrain\u002Fimg\"\n","python","",[1034],{"type":45,"tag":64,"props":1035,"children":1036},{"__ignoreMap":1032},[1037,1048,1057,1066],{"type":45,"tag":1038,"props":1039,"children":1042},"span",{"class":1040,"line":1041},"line",1,[1043],{"type":45,"tag":1038,"props":1044,"children":1045},{},[1046],{"type":39,"value":1047},"TRAIN_ROOT = \"\u002Fpath\u002Fto\u002Fextracted\u002Ftrain\"\n",{"type":45,"tag":1038,"props":1049,"children":1051},{"class":1040,"line":1050},2,[1052],{"type":45,"tag":1038,"props":1053,"children":1054},{},[1055],{"type":39,"value":1056},"EVAL_ROOT = \"\u002Fpath\u002Fto\u002Fextracted\u002Ftest\"\n",{"type":45,"tag":1038,"props":1058,"children":1060},{"class":1040,"line":1059},3,[1061],{"type":45,"tag":1038,"props":1062,"children":1063},{},[1064],{"type":39,"value":1065},"INFER_IMG_DIR = \"\u002Fpath\u002Fto\u002Fextracted\u002Ftest\u002Fimg\"\n",{"type":45,"tag":1038,"props":1067,"children":1069},{"class":1040,"line":1068},4,[1070],{"type":45,"tag":1038,"props":1071,"children":1072},{},[1073],{"type":39,"value":1074},"CALIB_IMG_DIR = \"\u002Fpath\u002Fto\u002Fextracted\u002Ftrain\u002Fimg\"\n",{"type":45,"tag":53,"props":1076,"children":1077},{},[1078],{"type":45,"tag":553,"props":1079,"children":1080},{},[1081],{"type":39,"value":1082},"train (mandatory data sources):",{"type":45,"tag":1027,"props":1084,"children":1086},{"className":1029,"code":1085,"language":1031,"meta":1032,"style":1032},"{\n    \"train.num_epochs\": 30,\n    \"train.checkpoint_interval\": 10,\n    \"train.validation_interval\": 10,\n    \"train.num_gpus\": 1,\n    \"dataset.train_dataset.loader.batch_size\": 16,\n    \"dataset.train_dataset.data_path\": [TRAIN_ROOT],\n    \"dataset.validate_dataset.data_path\": [EVAL_ROOT],\n}\n",[1087],{"type":45,"tag":64,"props":1088,"children":1089},{"__ignoreMap":1032},[1090,1098,1106,1114,1122,1131,1140,1149,1158],{"type":45,"tag":1038,"props":1091,"children":1092},{"class":1040,"line":1041},[1093],{"type":45,"tag":1038,"props":1094,"children":1095},{},[1096],{"type":39,"value":1097},"{\n",{"type":45,"tag":1038,"props":1099,"children":1100},{"class":1040,"line":1050},[1101],{"type":45,"tag":1038,"props":1102,"children":1103},{},[1104],{"type":39,"value":1105},"    \"train.num_epochs\": 30,\n",{"type":45,"tag":1038,"props":1107,"children":1108},{"class":1040,"line":1059},[1109],{"type":45,"tag":1038,"props":1110,"children":1111},{},[1112],{"type":39,"value":1113},"    \"train.checkpoint_interval\": 10,\n",{"type":45,"tag":1038,"props":1115,"children":1116},{"class":1040,"line":1068},[1117],{"type":45,"tag":1038,"props":1118,"children":1119},{},[1120],{"type":39,"value":1121},"    \"train.validation_interval\": 10,\n",{"type":45,"tag":1038,"props":1123,"children":1125},{"class":1040,"line":1124},5,[1126],{"type":45,"tag":1038,"props":1127,"children":1128},{},[1129],{"type":39,"value":1130},"    \"train.num_gpus\": 1,\n",{"type":45,"tag":1038,"props":1132,"children":1134},{"class":1040,"line":1133},6,[1135],{"type":45,"tag":1038,"props":1136,"children":1137},{},[1138],{"type":39,"value":1139},"    \"dataset.train_dataset.loader.batch_size\": 16,\n",{"type":45,"tag":1038,"props":1141,"children":1143},{"class":1040,"line":1142},7,[1144],{"type":45,"tag":1038,"props":1145,"children":1146},{},[1147],{"type":39,"value":1148},"    \"dataset.train_dataset.data_path\": [TRAIN_ROOT],\n",{"type":45,"tag":1038,"props":1150,"children":1152},{"class":1040,"line":1151},8,[1153],{"type":45,"tag":1038,"props":1154,"children":1155},{},[1156],{"type":39,"value":1157},"    \"dataset.validate_dataset.data_path\": [EVAL_ROOT],\n",{"type":45,"tag":1038,"props":1159,"children":1161},{"class":1040,"line":1160},9,[1162],{"type":45,"tag":1038,"props":1163,"children":1164},{},[1165],{"type":39,"value":1166},"}\n",{"type":45,"tag":53,"props":1168,"children":1169},{},[1170],{"type":45,"tag":553,"props":1171,"children":1172},{},[1173],{"type":39,"value":1174},"evaluate (mandatory data sources):",{"type":45,"tag":1027,"props":1176,"children":1178},{"className":1029,"code":1177,"language":1031,"meta":1032,"style":1032},"{\n    \"evaluate.checkpoint\": \"\u003Cselected train\u002FAutoML checkpoint>\",\n    \"dataset.validate_dataset.data_path\": [EVAL_ROOT],\n}\n",[1179],{"type":45,"tag":64,"props":1180,"children":1181},{"__ignoreMap":1032},[1182,1189,1197,1204],{"type":45,"tag":1038,"props":1183,"children":1184},{"class":1040,"line":1041},[1185],{"type":45,"tag":1038,"props":1186,"children":1187},{},[1188],{"type":39,"value":1097},{"type":45,"tag":1038,"props":1190,"children":1191},{"class":1040,"line":1050},[1192],{"type":45,"tag":1038,"props":1193,"children":1194},{},[1195],{"type":39,"value":1196},"    \"evaluate.checkpoint\": \"\u003Cselected train\u002FAutoML checkpoint>\",\n",{"type":45,"tag":1038,"props":1198,"children":1199},{"class":1040,"line":1059},[1200],{"type":45,"tag":1038,"props":1201,"children":1202},{},[1203],{"type":39,"value":1157},{"type":45,"tag":1038,"props":1205,"children":1206},{"class":1040,"line":1068},[1207],{"type":45,"tag":1038,"props":1208,"children":1209},{},[1210],{"type":39,"value":1166},{"type":45,"tag":53,"props":1212,"children":1213},{},[1214],{"type":45,"tag":553,"props":1215,"children":1216},{},[1217],{"type":39,"value":1218},"inference (mandatory data sources):",{"type":45,"tag":1027,"props":1220,"children":1222},{"className":1029,"code":1221,"language":1031,"meta":1032,"style":1032},"{\n    \"inference.checkpoint\": \"\u003Cselected train\u002FAutoML checkpoint>\",\n    \"inference.input_folder\": INFER_IMG_DIR,\n}\n",[1223],{"type":45,"tag":64,"props":1224,"children":1225},{"__ignoreMap":1032},[1226,1233,1241,1249],{"type":45,"tag":1038,"props":1227,"children":1228},{"class":1040,"line":1041},[1229],{"type":45,"tag":1038,"props":1230,"children":1231},{},[1232],{"type":39,"value":1097},{"type":45,"tag":1038,"props":1234,"children":1235},{"class":1040,"line":1050},[1236],{"type":45,"tag":1038,"props":1237,"children":1238},{},[1239],{"type":39,"value":1240},"    \"inference.checkpoint\": \"\u003Cselected train\u002FAutoML checkpoint>\",\n",{"type":45,"tag":1038,"props":1242,"children":1243},{"class":1040,"line":1059},[1244],{"type":45,"tag":1038,"props":1245,"children":1246},{},[1247],{"type":39,"value":1248},"    \"inference.input_folder\": INFER_IMG_DIR,\n",{"type":45,"tag":1038,"props":1250,"children":1251},{"class":1040,"line":1068},[1252],{"type":45,"tag":1038,"props":1253,"children":1254},{},[1255],{"type":39,"value":1166},{"type":45,"tag":53,"props":1257,"children":1258},{},[1259],{"type":45,"tag":553,"props":1260,"children":1261},{},[1262],{"type":39,"value":1263},"prune (mandatory data sources):",{"type":45,"tag":1027,"props":1265,"children":1267},{"className":1029,"code":1266,"language":1031,"meta":1032,"style":1032},"{\n    \"prune.checkpoint\": \"\u003Cselected train\u002FAutoML checkpoint>\",\n    \"dataset.validate_dataset.data_path\": [EVAL_ROOT],\n}\n",[1268],{"type":45,"tag":64,"props":1269,"children":1270},{"__ignoreMap":1032},[1271,1278,1286,1293],{"type":45,"tag":1038,"props":1272,"children":1273},{"class":1040,"line":1041},[1274],{"type":45,"tag":1038,"props":1275,"children":1276},{},[1277],{"type":39,"value":1097},{"type":45,"tag":1038,"props":1279,"children":1280},{"class":1040,"line":1050},[1281],{"type":45,"tag":1038,"props":1282,"children":1283},{},[1284],{"type":39,"value":1285},"    \"prune.checkpoint\": \"\u003Cselected train\u002FAutoML checkpoint>\",\n",{"type":45,"tag":1038,"props":1287,"children":1288},{"class":1040,"line":1059},[1289],{"type":45,"tag":1038,"props":1290,"children":1291},{},[1292],{"type":39,"value":1157},{"type":45,"tag":1038,"props":1294,"children":1295},{"class":1040,"line":1068},[1296],{"type":45,"tag":1038,"props":1297,"children":1298},{},[1299],{"type":39,"value":1166},{"type":45,"tag":53,"props":1301,"children":1302},{},[1303],{"type":45,"tag":553,"props":1304,"children":1305},{},[1306],{"type":39,"value":1307},"quantize (mandatory data sources):",{"type":45,"tag":1027,"props":1309,"children":1311},{"className":1029,"code":1310,"language":1031,"meta":1032,"style":1032},"{\n    \"quantize.model_path\": \"\u003Cselected train checkpoint or exported ONNX>\",\n    \"dataset.train_dataset.data_path\": [TRAIN_ROOT],\n    \"dataset.validate_dataset.data_path\": [EVAL_ROOT],\n    \"dataset.quant_calibration_dataset.images_dir\": CALIB_IMG_DIR,\n}\n",[1312],{"type":45,"tag":64,"props":1313,"children":1314},{"__ignoreMap":1032},[1315,1322,1330,1337,1344,1352],{"type":45,"tag":1038,"props":1316,"children":1317},{"class":1040,"line":1041},[1318],{"type":45,"tag":1038,"props":1319,"children":1320},{},[1321],{"type":39,"value":1097},{"type":45,"tag":1038,"props":1323,"children":1324},{"class":1040,"line":1050},[1325],{"type":45,"tag":1038,"props":1326,"children":1327},{},[1328],{"type":39,"value":1329},"    \"quantize.model_path\": \"\u003Cselected train checkpoint or exported ONNX>\",\n",{"type":45,"tag":1038,"props":1331,"children":1332},{"class":1040,"line":1059},[1333],{"type":45,"tag":1038,"props":1334,"children":1335},{},[1336],{"type":39,"value":1148},{"type":45,"tag":1038,"props":1338,"children":1339},{"class":1040,"line":1068},[1340],{"type":45,"tag":1038,"props":1341,"children":1342},{},[1343],{"type":39,"value":1157},{"type":45,"tag":1038,"props":1345,"children":1346},{"class":1040,"line":1124},[1347],{"type":45,"tag":1038,"props":1348,"children":1349},{},[1350],{"type":39,"value":1351},"    \"dataset.quant_calibration_dataset.images_dir\": CALIB_IMG_DIR,\n",{"type":45,"tag":1038,"props":1353,"children":1354},{"class":1040,"line":1133},[1355],{"type":45,"tag":1038,"props":1356,"children":1357},{},[1358],{"type":39,"value":1166},{"type":45,"tag":53,"props":1360,"children":1361},{},[1362],{"type":45,"tag":553,"props":1363,"children":1364},{},[1365],{"type":39,"value":1366},"resume training (mandatory data sources):",{"type":45,"tag":1027,"props":1368,"children":1370},{"className":1029,"code":1369,"language":1031,"meta":1032,"style":1032},"{\n    \"train.resume_training_checkpoint_path\": \"\u003Cexact model_epoch checkpoint>\",\n    \"dataset.train_dataset.data_path\": [TRAIN_ROOT],\n    \"dataset.validate_dataset.data_path\": [EVAL_ROOT],\n}\n",[1371],{"type":45,"tag":64,"props":1372,"children":1373},{"__ignoreMap":1032},[1374,1381,1389,1396,1403],{"type":45,"tag":1038,"props":1375,"children":1376},{"class":1040,"line":1041},[1377],{"type":45,"tag":1038,"props":1378,"children":1379},{},[1380],{"type":39,"value":1097},{"type":45,"tag":1038,"props":1382,"children":1383},{"class":1040,"line":1050},[1384],{"type":45,"tag":1038,"props":1385,"children":1386},{},[1387],{"type":39,"value":1388},"    \"train.resume_training_checkpoint_path\": \"\u003Cexact model_epoch checkpoint>\",\n",{"type":45,"tag":1038,"props":1390,"children":1391},{"class":1040,"line":1059},[1392],{"type":45,"tag":1038,"props":1393,"children":1394},{},[1395],{"type":39,"value":1148},{"type":45,"tag":1038,"props":1397,"children":1398},{"class":1040,"line":1068},[1399],{"type":45,"tag":1038,"props":1400,"children":1401},{},[1402],{"type":39,"value":1157},{"type":45,"tag":1038,"props":1404,"children":1405},{"class":1040,"line":1124},[1406],{"type":45,"tag":1038,"props":1407,"children":1408},{},[1409],{"type":39,"value":1166},{"type":45,"tag":53,"props":1411,"children":1412},{},[1413],{"type":45,"tag":553,"props":1414,"children":1415},{},[1416],{"type":39,"value":1417},"retrain from prune output (mandatory data sources):",{"type":45,"tag":1027,"props":1419,"children":1421},{"className":1029,"code":1420,"language":1031,"meta":1032,"style":1032},"{\n    \"model.load_pruned_graph\": True,\n    \"model.pruned_graph_path\": \"\u003Cselected prune output>\",\n    \"dataset.train_dataset.data_path\": [TRAIN_ROOT],\n    \"dataset.validate_dataset.data_path\": [EVAL_ROOT],\n}\n",[1422],{"type":45,"tag":64,"props":1423,"children":1424},{"__ignoreMap":1032},[1425,1432,1440,1448,1455,1462],{"type":45,"tag":1038,"props":1426,"children":1427},{"class":1040,"line":1041},[1428],{"type":45,"tag":1038,"props":1429,"children":1430},{},[1431],{"type":39,"value":1097},{"type":45,"tag":1038,"props":1433,"children":1434},{"class":1040,"line":1050},[1435],{"type":45,"tag":1038,"props":1436,"children":1437},{},[1438],{"type":39,"value":1439},"    \"model.load_pruned_graph\": True,\n",{"type":45,"tag":1038,"props":1441,"children":1442},{"class":1040,"line":1059},[1443],{"type":45,"tag":1038,"props":1444,"children":1445},{},[1446],{"type":39,"value":1447},"    \"model.pruned_graph_path\": \"\u003Cselected prune output>\",\n",{"type":45,"tag":1038,"props":1449,"children":1450},{"class":1040,"line":1068},[1451],{"type":45,"tag":1038,"props":1452,"children":1453},{},[1454],{"type":39,"value":1148},{"type":45,"tag":1038,"props":1456,"children":1457},{"class":1040,"line":1124},[1458],{"type":45,"tag":1038,"props":1459,"children":1460},{},[1461],{"type":39,"value":1157},{"type":45,"tag":1038,"props":1463,"children":1464},{"class":1040,"line":1133},[1465],{"type":45,"tag":1038,"props":1466,"children":1467},{},[1468],{"type":39,"value":1166},{"type":45,"tag":53,"props":1470,"children":1471},{},[1472],{"type":45,"tag":553,"props":1473,"children":1474},{},[1475],{"type":39,"value":1476},"default_specs:",{"type":45,"tag":1027,"props":1478,"children":1480},{"className":1029,"code":1479,"language":1031,"meta":1032,"style":1032},"{\n    \"results_dir\": \"\u003Cwritable output directory>\",\n}\n",[1481],{"type":45,"tag":64,"props":1482,"children":1483},{"__ignoreMap":1032},[1484,1491,1499],{"type":45,"tag":1038,"props":1485,"children":1486},{"class":1040,"line":1041},[1487],{"type":45,"tag":1038,"props":1488,"children":1489},{},[1490],{"type":39,"value":1097},{"type":45,"tag":1038,"props":1492,"children":1493},{"class":1040,"line":1050},[1494],{"type":45,"tag":1038,"props":1495,"children":1496},{},[1497],{"type":39,"value":1498},"    \"results_dir\": \"\u003Cwritable output directory>\",\n",{"type":45,"tag":1038,"props":1500,"children":1501},{"class":1040,"line":1059},[1502],{"type":45,"tag":1038,"props":1503,"children":1504},{},[1505],{"type":39,"value":1166},{"type":45,"tag":242,"props":1507,"children":1509},{"id":1508},"eval-dataset",[1510],{"type":39,"value":1511},"Eval Dataset",{"type":45,"tag":53,"props":1513,"children":1514},{},[1515],{"type":39,"value":1516},"Optional. Test dataset provided as separate tarball.",{"type":45,"tag":242,"props":1518,"children":1520},{"id":1519},"important-parameters",[1521],{"type":39,"value":1522},"Important Parameters",{"type":45,"tag":545,"props":1524,"children":1525},{},[1526,1536,1546,1556],{"type":45,"tag":549,"props":1527,"children":1528},{},[1529,1534],{"type":45,"tag":553,"props":1530,"children":1531},{},[1532],{"type":39,"value":1533},"model.backbone",{"type":39,"value":1535},": Default deformable_resnet18. Deformable convolutions improve text region detection for irregular text.",{"type":45,"tag":549,"props":1537,"children":1538},{},[1539,1544],{"type":45,"tag":553,"props":1540,"children":1541},{},[1542],{"type":39,"value":1543},"train.optimizer.args.lr",{"type":39,"value":1545},": Learning rate. Default 0.001 (Adam).",{"type":45,"tag":549,"props":1547,"children":1548},{},[1549,1554],{"type":45,"tag":553,"props":1550,"children":1551},{},[1552],{"type":39,"value":1553},"postprocess.thresh",{"type":39,"value":1555},": Binarization threshold for text region extraction.",{"type":45,"tag":549,"props":1557,"children":1558},{},[1559,1564],{"type":45,"tag":553,"props":1560,"children":1561},{},[1562],{"type":39,"value":1563},"postprocess.box_thresh",{"type":39,"value":1565},": Box confidence threshold for filtering detections.",{"type":45,"tag":242,"props":1567,"children":1569},{"id":1568},"multi-gpu-multi-node",[1570],{"type":39,"value":1571},"Multi-GPU \u002F Multi-Node",{"type":45,"tag":53,"props":1573,"children":1574},{},[1575,1580,1582,1587],{"type":45,"tag":553,"props":1576,"children":1577},{},[1578],{"type":39,"value":1579},"Launch method:",{"type":39,"value":1581}," Lightning-managed (single ",{"type":45,"tag":64,"props":1583,"children":1585},{"className":1584},[],[1586],{"type":39,"value":1031},{"type":39,"value":1588}," process, Lightning spawns workers).",{"type":45,"tag":588,"props":1590,"children":1591},{},[1592,1612],{"type":45,"tag":592,"props":1593,"children":1594},{},[1595],{"type":45,"tag":596,"props":1596,"children":1597},{},[1598,1602,1607],{"type":45,"tag":600,"props":1599,"children":1600},{},[1601],{"type":39,"value":609},{"type":45,"tag":600,"props":1603,"children":1604},{},[1605],{"type":39,"value":1606},"Description",{"type":45,"tag":600,"props":1608,"children":1609},{},[1610],{"type":39,"value":1611},"Default",{"type":45,"tag":626,"props":1613,"children":1614},{},[1615,1637,1662],{"type":45,"tag":596,"props":1616,"children":1617},{},[1618,1627,1632],{"type":45,"tag":633,"props":1619,"children":1620},{},[1621],{"type":45,"tag":64,"props":1622,"children":1624},{"className":1623},[],[1625],{"type":39,"value":1626},"train.num_gpus",{"type":45,"tag":633,"props":1628,"children":1629},{},[1630],{"type":39,"value":1631},"Number of GPUs",{"type":45,"tag":633,"props":1633,"children":1634},{},[1635],{"type":39,"value":1636},"1",{"type":45,"tag":596,"props":1638,"children":1639},{},[1640,1649,1654],{"type":45,"tag":633,"props":1641,"children":1642},{},[1643],{"type":45,"tag":64,"props":1644,"children":1646},{"className":1645},[],[1647],{"type":39,"value":1648},"train.gpu_ids",{"type":45,"tag":633,"props":1650,"children":1651},{},[1652],{"type":39,"value":1653},"GPU device indices",{"type":45,"tag":633,"props":1655,"children":1656},{},[1657],{"type":45,"tag":1038,"props":1658,"children":1659},{},[1660],{"type":39,"value":1661},"0",{"type":45,"tag":596,"props":1663,"children":1664},{},[1665,1674,1697],{"type":45,"tag":633,"props":1666,"children":1667},{},[1668],{"type":45,"tag":64,"props":1669,"children":1671},{"className":1670},[],[1672],{"type":39,"value":1673},"train.distributed_strategy",{"type":45,"tag":633,"props":1675,"children":1676},{},[1677,1683,1684,1690,1691],{"type":45,"tag":64,"props":1678,"children":1680},{"className":1679},[],[1681],{"type":39,"value":1682},"ddp",{"type":39,"value":137},{"type":45,"tag":64,"props":1685,"children":1687},{"className":1686},[],[1688],{"type":39,"value":1689},"fsdp",{"type":39,"value":1004},{"type":45,"tag":64,"props":1692,"children":1694},{"className":1693},[],[1695],{"type":39,"value":1696},"deepspeed_stage_3_offload",{"type":45,"tag":633,"props":1698,"children":1699},{},[1700],{"type":45,"tag":64,"props":1701,"children":1703},{"className":1702},[],[1704],{"type":39,"value":1682},{"type":45,"tag":545,"props":1706,"children":1707},{},[1708,1724,1740,1750,1763],{"type":45,"tag":549,"props":1709,"children":1710},{},[1711,1716,1718],{"type":45,"tag":64,"props":1712,"children":1714},{"className":1713},[],[1715],{"type":39,"value":1682},{"type":39,"value":1717}," with activation checkpointing: ",{"type":45,"tag":64,"props":1719,"children":1721},{"className":1720},[],[1722],{"type":39,"value":1723},"find_unused_parameters=False",{"type":45,"tag":549,"props":1725,"children":1726},{},[1727,1732,1734],{"type":45,"tag":64,"props":1728,"children":1730},{"className":1729},[],[1731],{"type":39,"value":1682},{"type":39,"value":1733}," without: ",{"type":45,"tag":64,"props":1735,"children":1737},{"className":1736},[],[1738],{"type":39,"value":1739},"find_unused_parameters=True",{"type":45,"tag":549,"props":1741,"children":1742},{},[1743,1748],{"type":45,"tag":64,"props":1744,"children":1746},{"className":1745},[],[1747],{"type":39,"value":1689},{"type":39,"value":1749}," forces FP16",{"type":45,"tag":549,"props":1751,"children":1752},{},[1753,1761],{"type":45,"tag":553,"props":1754,"children":1755},{},[1756],{"type":45,"tag":64,"props":1757,"children":1759},{"className":1758},[],[1760],{"type":39,"value":1696},{"type":39,"value":1762}," is uniquely supported for OCDNet (forces FP16)",{"type":45,"tag":549,"props":1764,"children":1765},{},[1766,1768],{"type":39,"value":1767},"FAN backbones auto-enable ",{"type":45,"tag":64,"props":1769,"children":1771},{"className":1770},[],[1772],{"type":39,"value":1773},"sync_batchnorm",{"type":45,"tag":242,"props":1775,"children":1777},{"id":1776},"hardware",[1778],{"type":39,"value":1779},"Hardware",{"type":45,"tag":53,"props":1781,"children":1782},{},[1783],{"type":39,"value":1784},"Minimum 1 GPU(s), recommended 1 GPU(s). 8GB+ VRAM per GPU. OCDNet is lightweight. Single GPU is sufficient for most datasets.",{"type":45,"tag":242,"props":1786,"children":1788},{"id":1787},"error-patterns",[1789],{"type":39,"value":1790},"Error Patterns",{"type":45,"tag":53,"props":1792,"children":1793},{},[1794,1799],{"type":45,"tag":553,"props":1795,"children":1796},{},[1797],{"type":39,"value":1798},"Low detection rate",{"type":39,"value":1800},": Tune postprocess.thresh and box_thresh. Default thresholds may be too aggressive for some datasets.",{"type":45,"tag":53,"props":1802,"children":1803},{},[1804,1809,1811,1817,1819,1825,1827,1833,1835,1841],{"type":45,"tag":553,"props":1805,"children":1806},{},[1807],{"type":39,"value":1808},"One-epoch smoke train with default scheduler",{"type":39,"value":1810},": ",{"type":45,"tag":64,"props":1812,"children":1814},{"className":1813},[],[1815],{"type":39,"value":1816},"train.num_epochs",{"type":39,"value":1818}," must not equal ",{"type":45,"tag":64,"props":1820,"children":1822},{"className":1821},[],[1823],{"type":39,"value":1824},"train.lr_scheduler.args.warmup_epoch",{"type":39,"value":1826},". For one-epoch validation, set ",{"type":45,"tag":64,"props":1828,"children":1830},{"className":1829},[],[1831],{"type":39,"value":1832},"warmup_epoch: 0",{"type":39,"value":1834},"; for normal starter runs, keep ",{"type":45,"tag":64,"props":1836,"children":1838},{"className":1837},[],[1839],{"type":39,"value":1840},"num_epochs > warmup_epoch",{"type":39,"value":1842},".",{"type":45,"tag":53,"props":1844,"children":1845},{},[1846,1851,1852,1858,1860,1865,1866,1871,1873,1878,1879,1884],{"type":45,"tag":553,"props":1847,"children":1848},{},[1849],{"type":39,"value":1850},"Archive passed as dataset path",{"type":39,"value":1810},{"type":45,"tag":64,"props":1853,"children":1855},{"className":1854},[],[1856],{"type":39,"value":1857},"dataset.*.data_path",{"type":39,"value":1859}," is not an archive path for OCDNet. Passing ",{"type":45,"tag":64,"props":1861,"children":1863},{"className":1862},[],[1864],{"type":39,"value":995},{"type":39,"value":187},{"type":45,"tag":64,"props":1867,"children":1869},{"className":1868},[],[1870],{"type":39,"value":1002},{"type":39,"value":1872}," directly causes the dataloader to open the gzip as a UTF-8 datalist. Extract the archive and pass the split folder containing ",{"type":45,"tag":64,"props":1874,"children":1876},{"className":1875},[],[1877],{"type":39,"value":657},{"type":39,"value":217},{"type":45,"tag":64,"props":1880,"children":1882},{"className":1881},[],[1883],{"type":39,"value":664},{"type":39,"value":1885},", or pass a real UTF-8 datalist file.",{"type":45,"tag":53,"props":1887,"children":1888},{},[1889,1894,1896,1902,1904,1909,1911,1917,1919,1925,1927,1933],{"type":45,"tag":553,"props":1890,"children":1891},{},[1892],{"type":39,"value":1893},"Quantize checkpoint type",{"type":39,"value":1895},": Do not pass ",{"type":45,"tag":64,"props":1897,"children":1899},{"className":1898},[],[1900],{"type":39,"value":1901},"model_best.pth",{"type":39,"value":1903}," to the PyTorch quantize path. Some older PyT runtimes wrote ",{"type":45,"tag":64,"props":1905,"children":1907},{"className":1906},[],[1908],{"type":39,"value":1901},{"type":39,"value":1910}," without full Lightning checkpoint metadata. The default ",{"type":45,"tag":64,"props":1912,"children":1914},{"className":1913},[],[1915],{"type":39,"value":1916},"torchao",{"type":39,"value":1918}," quantize path should use the intended full ",{"type":45,"tag":64,"props":1920,"children":1922},{"className":1921},[],[1923],{"type":39,"value":1924},"model_epoch_\u003Cepoch>_step_\u003Cstep>.pth",{"type":39,"value":1926}," checkpoint and write ",{"type":45,"tag":64,"props":1928,"children":1930},{"className":1929},[],[1931],{"type":39,"value":1932},"quantized_model_torchao.pth",{"type":39,"value":1842},{"type":45,"tag":53,"props":1935,"children":1936},{},[1937,1942,1943,1949,1951,1957,1959,1965],{"type":45,"tag":553,"props":1938,"children":1939},{},[1940],{"type":39,"value":1941},"Default specs output directory",{"type":39,"value":1810},{"type":45,"tag":64,"props":1944,"children":1946},{"className":1945},[],[1947],{"type":39,"value":1948},"ocdnet default_specs",{"type":39,"value":1950}," requires a writable ",{"type":45,"tag":64,"props":1952,"children":1954},{"className":1953},[],[1955],{"type":39,"value":1956},"results_dir",{"type":39,"value":1958}," override, for example ",{"type":45,"tag":64,"props":1960,"children":1962},{"className":1961},[],[1963],{"type":39,"value":1964},"results_dir=\u002Fworkspace\u002Frun\u002Fresults\u002Fdefault_specs",{"type":39,"value":1842},{"type":45,"tag":242,"props":1967,"children":1969},{"id":1968},"checkpoint-handoff",[1970],{"type":39,"value":1971},"Checkpoint Handoff",{"type":45,"tag":53,"props":1973,"children":1974},{},[1975,1977,1982,1984,1990,1992,1998,2000,2005,2007,2013,2014,2020,2021,2027,2028,2034,2036,2041,2042,2047,2049,2055,2057,2063,2065,2070],{"type":39,"value":1976},"OCDNet train writes ",{"type":45,"tag":64,"props":1978,"children":1980},{"className":1979},[],[1981],{"type":39,"value":1901},{"type":39,"value":1983}," plus full Lightning epoch checkpoints such as ",{"type":45,"tag":64,"props":1985,"children":1987},{"className":1986},[],[1988],{"type":39,"value":1989},"model_epoch_001_step_00046.pth",{"type":39,"value":1991},"; it may also write ",{"type":45,"tag":64,"props":1993,"children":1995},{"className":1994},[],[1996],{"type":39,"value":1997},"ocd_model_latest.pth",{"type":39,"value":1999}," as a latest symlink. Use ",{"type":45,"tag":64,"props":2001,"children":2003},{"className":2002},[],[2004],{"type":39,"value":1901},{"type":39,"value":2006}," for ",{"type":45,"tag":64,"props":2008,"children":2010},{"className":2009},[],[2011],{"type":39,"value":2012},"evaluate.checkpoint",{"type":39,"value":137},{"type":45,"tag":64,"props":2015,"children":2017},{"className":2016},[],[2018],{"type":39,"value":2019},"inference.checkpoint",{"type":39,"value":137},{"type":45,"tag":64,"props":2022,"children":2024},{"className":2023},[],[2025],{"type":39,"value":2026},"export.checkpoint",{"type":39,"value":171},{"type":45,"tag":64,"props":2029,"children":2031},{"className":2030},[],[2032],{"type":39,"value":2033},"prune.checkpoint",{"type":39,"value":2035}," when the user asks for the best checkpoint. Use a specific ",{"type":45,"tag":64,"props":2037,"children":2039},{"className":2038},[],[2040],{"type":39,"value":1924},{"type":39,"value":2006},{"type":45,"tag":64,"props":2043,"children":2045},{"className":2044},[],[2046],{"type":39,"value":238},{"type":39,"value":2048}," and for any action that explicitly needs a full Lightning checkpoint. Prune writes artifacts such as ",{"type":45,"tag":64,"props":2050,"children":2052},{"className":2051},[],[2053],{"type":39,"value":2054},"pruned_\u003Cch_sparsity>.pth",{"type":39,"value":2056},"; use the exact pruned ",{"type":45,"tag":64,"props":2058,"children":2060},{"className":2059},[],[2061],{"type":39,"value":2062},".pth",{"type":39,"value":2064}," artifact for ",{"type":45,"tag":64,"props":2066,"children":2068},{"className":2067},[],[2069],{"type":39,"value":223},{"type":39,"value":2071}," when retraining from a pruned graph. Use a latest checkpoint only when the user explicitly asks for latest.",{"type":45,"tag":53,"props":2073,"children":2074},{},[2075,2077,2082,2084,2089,2091,2097,2099,2105,2107,2113],{"type":39,"value":2076},"If quantize is retried with a PyTorch backend, resolve the full ",{"type":45,"tag":64,"props":2078,"children":2080},{"className":2079},[],[2081],{"type":39,"value":1924},{"type":39,"value":2083}," that corresponds to the intended best epoch or requested epoch; do not pass ",{"type":45,"tag":64,"props":2085,"children":2087},{"className":2086},[],[2088],{"type":39,"value":1901},{"type":39,"value":2090}," to the PyTorch quantize path. If quantize is retried with ",{"type":45,"tag":64,"props":2092,"children":2094},{"className":2093},[],[2095],{"type":39,"value":2096},"modelopt.onnx",{"type":39,"value":2098},", pass the exported ONNX as ",{"type":45,"tag":64,"props":2100,"children":2102},{"className":2101},[],[2103],{"type":39,"value":2104},"quantize.model_path",{"type":39,"value":2106}," and verify that the runtime image actually contains ",{"type":45,"tag":64,"props":2108,"children":2110},{"className":2109},[],[2111],{"type":39,"value":2112},"modelopt.onnx.quantization",{"type":39,"value":1842},{"type":45,"tag":242,"props":2115,"children":2117},{"id":2116},"spec-param-parent-model-inference",[2118],{"type":39,"value":2119},"Spec Param \u002F Parent Model Inference",{"type":45,"tag":53,"props":2121,"children":2122},{},[2123,2125,2131,2133,2139,2141,2147],{"type":39,"value":2124},"Model-specific inference mappings belong in this MD file, not in ",{"type":45,"tag":64,"props":2126,"children":2128},{"className":2127},[],[2129],{"type":39,"value":2130},"config.json",{"type":39,"value":2132},". Generated runners should read this section and apply the mappings with SDK helpers before ",{"type":45,"tag":64,"props":2134,"children":2136},{"className":2135},[],[2137],{"type":39,"value":2138},"create_job()",{"type":39,"value":2140},". This mirrors the old microservices ",{"type":45,"tag":64,"props":2142,"children":2144},{"className":2143},[],[2145],{"type":39,"value":2146},"infer_params.py",{"type":39,"value":2148}," flow.",{"type":45,"tag":53,"props":2150,"children":2151},{},[2152],{"type":39,"value":2153},"Model handoff mappings:",{"type":45,"tag":588,"props":2155,"children":2156},{},[2157,2182],{"type":45,"tag":592,"props":2158,"children":2159},{},[2160],{"type":45,"tag":596,"props":2161,"children":2162},{},[2163,2167,2172,2177],{"type":45,"tag":600,"props":2164,"children":2165},{},[2166],{"type":39,"value":604},{"type":45,"tag":600,"props":2168,"children":2169},{},[2170],{"type":39,"value":2171},"Spec Field",{"type":45,"tag":600,"props":2173,"children":2174},{},[2175],{"type":39,"value":2176},"Inference Function",{"type":45,"tag":600,"props":2178,"children":2179},{},[2180],{"type":39,"value":2181},"Meaning",{"type":45,"tag":626,"props":2183,"children":2184},{},[2185,2214,2243,2270,2300,2327,2354,2381,2408,2435,2462,2489,2518,2545,2574,2601],{"type":45,"tag":596,"props":2186,"children":2187},{},[2188,2192,2200,2209],{"type":45,"tag":633,"props":2189,"children":2190},{},[2191],{"type":39,"value":90},{"type":45,"tag":633,"props":2193,"children":2194},{},[2195],{"type":45,"tag":64,"props":2196,"children":2198},{"className":2197},[],[2199],{"type":39,"value":2012},{"type":45,"tag":633,"props":2201,"children":2202},{},[2203],{"type":45,"tag":64,"props":2204,"children":2206},{"className":2205},[],[2207],{"type":39,"value":2208},"parent_model",{"type":45,"tag":633,"props":2210,"children":2211},{},[2212],{"type":39,"value":2213},"model file inferred from the parent job results folder",{"type":45,"tag":596,"props":2215,"children":2216},{},[2217,2221,2229,2238],{"type":45,"tag":633,"props":2218,"children":2219},{},[2220],{"type":39,"value":90},{"type":45,"tag":633,"props":2222,"children":2223},{},[2224],{"type":45,"tag":64,"props":2225,"children":2227},{"className":2226},[],[2228],{"type":39,"value":1956},{"type":45,"tag":633,"props":2230,"children":2231},{},[2232],{"type":45,"tag":64,"props":2233,"children":2235},{"className":2234},[],[2236],{"type":39,"value":2237},"output_dir",{"type":45,"tag":633,"props":2239,"children":2240},{},[2241],{"type":39,"value":2242},"current job results directory",{"type":45,"tag":596,"props":2244,"children":2245},{},[2246,2250,2258,2266],{"type":45,"tag":633,"props":2247,"children":2248},{},[2249],{"type":39,"value":149},{"type":45,"tag":633,"props":2251,"children":2252},{},[2253],{"type":45,"tag":64,"props":2254,"children":2256},{"className":2255},[],[2257],{"type":39,"value":2026},{"type":45,"tag":633,"props":2259,"children":2260},{},[2261],{"type":45,"tag":64,"props":2262,"children":2264},{"className":2263},[],[2265],{"type":39,"value":2208},{"type":45,"tag":633,"props":2267,"children":2268},{},[2269],{"type":39,"value":2213},{"type":45,"tag":596,"props":2271,"children":2272},{},[2273,2277,2286,2295],{"type":45,"tag":633,"props":2274,"children":2275},{},[2276],{"type":39,"value":149},{"type":45,"tag":633,"props":2278,"children":2279},{},[2280],{"type":45,"tag":64,"props":2281,"children":2283},{"className":2282},[],[2284],{"type":39,"value":2285},"export.onnx_file",{"type":45,"tag":633,"props":2287,"children":2288},{},[2289],{"type":45,"tag":64,"props":2290,"children":2292},{"className":2291},[],[2293],{"type":39,"value":2294},"create_onnx_file",{"type":45,"tag":633,"props":2296,"children":2297},{},[2298],{"type":39,"value":2299},"output ONNX path",{"type":45,"tag":596,"props":2301,"children":2302},{},[2303,2307,2315,2323],{"type":45,"tag":633,"props":2304,"children":2305},{},[2306],{"type":39,"value":149},{"type":45,"tag":633,"props":2308,"children":2309},{},[2310],{"type":45,"tag":64,"props":2311,"children":2313},{"className":2312},[],[2314],{"type":39,"value":1956},{"type":45,"tag":633,"props":2316,"children":2317},{},[2318],{"type":45,"tag":64,"props":2319,"children":2321},{"className":2320},[],[2322],{"type":39,"value":2237},{"type":45,"tag":633,"props":2324,"children":2325},{},[2326],{"type":39,"value":2242},{"type":45,"tag":596,"props":2328,"children":2329},{},[2330,2334,2342,2350],{"type":45,"tag":633,"props":2331,"children":2332},{},[2333],{"type":39,"value":98},{"type":45,"tag":633,"props":2335,"children":2336},{},[2337],{"type":45,"tag":64,"props":2338,"children":2340},{"className":2339},[],[2341],{"type":39,"value":2019},{"type":45,"tag":633,"props":2343,"children":2344},{},[2345],{"type":45,"tag":64,"props":2346,"children":2348},{"className":2347},[],[2349],{"type":39,"value":2208},{"type":45,"tag":633,"props":2351,"children":2352},{},[2353],{"type":39,"value":2213},{"type":45,"tag":596,"props":2355,"children":2356},{},[2357,2361,2369,2377],{"type":45,"tag":633,"props":2358,"children":2359},{},[2360],{"type":39,"value":98},{"type":45,"tag":633,"props":2362,"children":2363},{},[2364],{"type":45,"tag":64,"props":2365,"children":2367},{"className":2366},[],[2368],{"type":39,"value":1956},{"type":45,"tag":633,"props":2370,"children":2371},{},[2372],{"type":45,"tag":64,"props":2373,"children":2375},{"className":2374},[],[2376],{"type":39,"value":2237},{"type":45,"tag":633,"props":2378,"children":2379},{},[2380],{"type":39,"value":2242},{"type":45,"tag":596,"props":2382,"children":2383},{},[2384,2388,2396,2404],{"type":45,"tag":633,"props":2385,"children":2386},{},[2387],{"type":39,"value":162},{"type":45,"tag":633,"props":2389,"children":2390},{},[2391],{"type":45,"tag":64,"props":2392,"children":2394},{"className":2393},[],[2395],{"type":39,"value":2033},{"type":45,"tag":633,"props":2397,"children":2398},{},[2399],{"type":45,"tag":64,"props":2400,"children":2402},{"className":2401},[],[2403],{"type":39,"value":2208},{"type":45,"tag":633,"props":2405,"children":2406},{},[2407],{"type":39,"value":2213},{"type":45,"tag":596,"props":2409,"children":2410},{},[2411,2415,2423,2431],{"type":45,"tag":633,"props":2412,"children":2413},{},[2414],{"type":39,"value":162},{"type":45,"tag":633,"props":2416,"children":2417},{},[2418],{"type":45,"tag":64,"props":2419,"children":2421},{"className":2420},[],[2422],{"type":39,"value":1956},{"type":45,"tag":633,"props":2424,"children":2425},{},[2426],{"type":45,"tag":64,"props":2427,"children":2429},{"className":2428},[],[2430],{"type":39,"value":2237},{"type":45,"tag":633,"props":2432,"children":2433},{},[2434],{"type":39,"value":2242},{"type":45,"tag":596,"props":2436,"children":2437},{},[2438,2442,2450,2458],{"type":45,"tag":633,"props":2439,"children":2440},{},[2441],{"type":39,"value":169},{"type":45,"tag":633,"props":2443,"children":2444},{},[2445],{"type":45,"tag":64,"props":2446,"children":2448},{"className":2447},[],[2449],{"type":39,"value":2104},{"type":45,"tag":633,"props":2451,"children":2452},{},[2453],{"type":45,"tag":64,"props":2454,"children":2456},{"className":2455},[],[2457],{"type":39,"value":2208},{"type":45,"tag":633,"props":2459,"children":2460},{},[2461],{"type":39,"value":2213},{"type":45,"tag":596,"props":2463,"children":2464},{},[2465,2469,2477,2485],{"type":45,"tag":633,"props":2466,"children":2467},{},[2468],{"type":39,"value":169},{"type":45,"tag":633,"props":2470,"children":2471},{},[2472],{"type":45,"tag":64,"props":2473,"children":2475},{"className":2474},[],[2476],{"type":39,"value":1956},{"type":45,"tag":633,"props":2478,"children":2479},{},[2480],{"type":45,"tag":64,"props":2481,"children":2483},{"className":2482},[],[2484],{"type":39,"value":2237},{"type":45,"tag":633,"props":2486,"children":2487},{},[2488],{"type":39,"value":2242},{"type":45,"tag":596,"props":2490,"children":2491},{},[2492,2497,2505,2513],{"type":45,"tag":633,"props":2493,"children":2494},{},[2495],{"type":39,"value":2496},"retrain from prune",{"type":45,"tag":633,"props":2498,"children":2499},{},[2500],{"type":45,"tag":64,"props":2501,"children":2503},{"className":2502},[],[2504],{"type":39,"value":223},{"type":45,"tag":633,"props":2506,"children":2507},{},[2508],{"type":45,"tag":64,"props":2509,"children":2511},{"className":2510},[],[2512],{"type":39,"value":2208},{"type":45,"tag":633,"props":2514,"children":2515},{},[2516],{"type":39,"value":2517},"exact pruned model file inferred from the parent prune results folder",{"type":45,"tag":596,"props":2519,"children":2520},{},[2521,2525,2533,2541],{"type":45,"tag":633,"props":2522,"children":2523},{},[2524],{"type":39,"value":2496},{"type":45,"tag":633,"props":2526,"children":2527},{},[2528],{"type":45,"tag":64,"props":2529,"children":2531},{"className":2530},[],[2532],{"type":39,"value":1956},{"type":45,"tag":633,"props":2534,"children":2535},{},[2536],{"type":45,"tag":64,"props":2537,"children":2539},{"className":2538},[],[2540],{"type":39,"value":2237},{"type":45,"tag":633,"props":2542,"children":2543},{},[2544],{"type":39,"value":2242},{"type":45,"tag":596,"props":2546,"children":2547},{},[2548,2552,2560,2569],{"type":45,"tag":633,"props":2549,"children":2550},{},[2551],{"type":39,"value":135},{"type":45,"tag":633,"props":2553,"children":2554},{},[2555],{"type":45,"tag":64,"props":2556,"children":2558},{"className":2557},[],[2559],{"type":39,"value":69},{"type":45,"tag":633,"props":2561,"children":2562},{},[2563],{"type":45,"tag":64,"props":2564,"children":2566},{"className":2565},[],[2567],{"type":39,"value":2568},"ptm_if_no_resume_model",{"type":45,"tag":633,"props":2570,"children":2571},{},[2572],{"type":39,"value":2573},"PTM when no resume checkpoint exists",{"type":45,"tag":596,"props":2575,"children":2576},{},[2577,2581,2589,2597],{"type":45,"tag":633,"props":2578,"children":2579},{},[2580],{"type":39,"value":135},{"type":45,"tag":633,"props":2582,"children":2583},{},[2584],{"type":45,"tag":64,"props":2585,"children":2587},{"className":2586},[],[2588],{"type":39,"value":1956},{"type":45,"tag":633,"props":2590,"children":2591},{},[2592],{"type":45,"tag":64,"props":2593,"children":2595},{"className":2594},[],[2596],{"type":39,"value":2237},{"type":45,"tag":633,"props":2598,"children":2599},{},[2600],{"type":39,"value":2242},{"type":45,"tag":596,"props":2602,"children":2603},{},[2604,2608,2616,2625],{"type":45,"tag":633,"props":2605,"children":2606},{},[2607],{"type":39,"value":135},{"type":45,"tag":633,"props":2609,"children":2610},{},[2611],{"type":45,"tag":64,"props":2612,"children":2614},{"className":2613},[],[2615],{"type":39,"value":238},{"type":45,"tag":633,"props":2617,"children":2618},{},[2619],{"type":45,"tag":64,"props":2620,"children":2622},{"className":2621},[],[2623],{"type":39,"value":2624},"resume_model",{"type":45,"tag":633,"props":2626,"children":2627},{},[2628],{"type":39,"value":2629},"model file inferred from the current job results folder",{"type":45,"tag":53,"props":2631,"children":2632},{},[2633,2635,2640,2641,2647,2649,2655,2657,2662],{"type":39,"value":2634},"For ",{"type":45,"tag":64,"props":2636,"children":2638},{"className":2637},[],[2639],{"type":39,"value":2208},{"type":39,"value":187},{"type":45,"tag":64,"props":2642,"children":2644},{"className":2643},[],[2645],{"type":39,"value":2646},"parent_model_folder",{"type":39,"value":2648},", pass the upstream train\u002Fexport\u002FAutoML child job id as ",{"type":45,"tag":64,"props":2650,"children":2652},{"className":2651},[],[2653],{"type":39,"value":2654},"parent_job_id",{"type":39,"value":2656},". 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":64,"props":2658,"children":2660},{"className":2659},[],[2661],{"type":39,"value":2130},{"type":39,"value":2663}," and do not patch generated runner scripts to guess checkpoint paths.",{"type":45,"tag":242,"props":2665,"children":2667},{"id":2666},"deployment",[2668],{"type":39,"value":2669},"Deployment",{"type":45,"tag":545,"props":2671,"children":2672},{},[2673],{"type":45,"tag":549,"props":2674,"children":2675},{},[2676],{"type":45,"tag":2677,"props":2678,"children":2679},"a",{"href":106},[2680],{"type":39,"value":2681},"tao-deploy-ocdnet",{"type":45,"tag":2683,"props":2684,"children":2685},"style",{},[2686],{"type":39,"value":2687},"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":2689,"total":2791},[2690,2707,2719,2733,2745,2762,2777],{"slug":2691,"name":2691,"fn":2692,"description":2693,"org":2694,"tags":2695,"stars":20,"repoUrl":21,"updatedAt":2706},"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},[2696,2699,2702,2703],{"name":2697,"slug":2698,"type":15},"Data Analysis","data-analysis",{"name":2700,"slug":2701,"type":15},"Data Engineering","data-engineering",{"name":9,"slug":8,"type":15},{"name":2704,"slug":2705,"type":15},"Performance","performance","2026-07-14T05:28:43.176466",{"slug":2708,"name":2708,"fn":2709,"description":2710,"org":2711,"tags":2712,"stars":20,"repoUrl":21,"updatedAt":2718},"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},[2713,2714,2717],{"name":2669,"slug":2666,"type":15},{"name":2715,"slug":2716,"type":15},"Infrastructure","infrastructure",{"name":9,"slug":8,"type":15},"2026-07-14T05:29:06.667109",{"slug":2720,"name":2720,"fn":2721,"description":2722,"org":2723,"tags":2724,"stars":20,"repoUrl":21,"updatedAt":2732},"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},[2725,2728,2729],{"name":2726,"slug":2727,"type":15},"Agents","agents",{"name":9,"slug":8,"type":15},{"name":2730,"slug":2731,"type":15},"Research","research","2026-07-14T05:28:06.816956",{"slug":2734,"name":2734,"fn":2735,"description":2736,"org":2737,"tags":2738,"stars":20,"repoUrl":21,"updatedAt":2744},"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},[2739,2740,2741],{"name":2697,"slug":2698,"type":15},{"name":9,"slug":8,"type":15},{"name":2742,"slug":2743,"type":15},"Testing","testing","2026-07-17T05:29:03.913266",{"slug":2746,"name":2746,"fn":2747,"description":2748,"org":2749,"tags":2750,"stars":20,"repoUrl":21,"updatedAt":2761},"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},[2751,2754,2757,2758],{"name":2752,"slug":2753,"type":15},"Automation","automation",{"name":2755,"slug":2756,"type":15},"Imaging","imaging",{"name":9,"slug":8,"type":15},{"name":2759,"slug":2760,"type":15},"Video","video","2026-07-17T05:28:53.905004",{"slug":2763,"name":2763,"fn":2764,"description":2765,"org":2766,"tags":2767,"stars":20,"repoUrl":21,"updatedAt":2776},"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},[2768,2769,2772,2773],{"name":2669,"slug":2666,"type":15},{"name":2770,"slug":2771,"type":15},"Docker","docker",{"name":9,"slug":8,"type":15},{"name":2774,"slug":2775,"type":15},"Operations","operations","2026-07-17T05:28:56.913999",{"slug":2778,"name":2778,"fn":2779,"description":2780,"org":2781,"tags":2782,"stars":20,"repoUrl":21,"updatedAt":2790},"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},[2783,2784,2787],{"name":9,"slug":8,"type":15},{"name":2785,"slug":2786,"type":15},"Quantum Computing","quantum-computing",{"name":2788,"slug":2789,"type":15},"Simulation","simulation","2026-07-14T05:26:58.898253",305,{"items":2793,"total":2941},[2794,2812,2827,2838,2850,2864,2877,2889,2900,2909,2923,2932],{"slug":2795,"name":2795,"fn":2796,"description":2797,"org":2798,"tags":2799,"stars":2809,"repoUrl":2810,"updatedAt":2811},"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},[2800,2803,2806],{"name":2801,"slug":2802,"type":15},"Documentation","documentation",{"name":2804,"slug":2805,"type":15},"MCP","mcp",{"name":2807,"slug":2808,"type":15},"Search","search",21777,"https:\u002F\u002Fgithub.com\u002FNVIDIA\u002FNemoClaw","2026-07-20T06:00:01.461044",{"slug":2813,"name":2813,"fn":2814,"description":2815,"org":2816,"tags":2817,"stars":2824,"repoUrl":2825,"updatedAt":2826},"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},[2818,2821,2822],{"name":2819,"slug":2820,"type":15},"Containers","containers",{"name":2669,"slug":2666,"type":15},{"name":2823,"slug":1031,"type":15},"Python",17049,"https:\u002F\u002Fgithub.com\u002FNVIDIA\u002FMegatron-LM","2026-07-27T06:06:11.249662",{"slug":2828,"name":2828,"fn":2829,"description":2830,"org":2831,"tags":2832,"stars":2824,"repoUrl":2825,"updatedAt":2837},"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},[2833,2836],{"name":2834,"slug":2835,"type":15},"CI\u002FCD","ci-cd",{"name":2669,"slug":2666,"type":15},"2026-07-14T05:25:59.97109",{"slug":2839,"name":2839,"fn":2840,"description":2841,"org":2842,"tags":2843,"stars":2824,"repoUrl":2825,"updatedAt":2849},"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},[2844,2845,2846],{"name":2834,"slug":2835,"type":15},{"name":2669,"slug":2666,"type":15},{"name":2847,"slug":2848,"type":15},"GitHub","github","2026-07-27T06:06:12.278222",{"slug":2851,"name":2851,"fn":2852,"description":2853,"org":2854,"tags":2855,"stars":2824,"repoUrl":2825,"updatedAt":2863},"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},[2856,2859,2860],{"name":2857,"slug":2858,"type":15},"Debugging","debugging",{"name":2847,"slug":2848,"type":15},{"name":2861,"slug":2862,"type":15},"Triage","triage","2026-07-14T05:25:57.442089",{"slug":2865,"name":2865,"fn":2866,"description":2867,"org":2868,"tags":2869,"stars":2824,"repoUrl":2825,"updatedAt":2876},"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},[2870,2873],{"name":2871,"slug":2872,"type":15},"Best Practices","best-practices",{"name":2874,"slug":2875,"type":15},"Code Analysis","code-analysis","2026-07-14T05:25:56.18433",{"slug":2878,"name":2878,"fn":2879,"description":2880,"org":2881,"tags":2882,"stars":2824,"repoUrl":2825,"updatedAt":2888},"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},[2883,2884,2887],{"name":13,"slug":14,"type":15},{"name":2885,"slug":2886,"type":15},"Migration","migration",{"name":9,"slug":8,"type":15},"2026-07-17T06:07:11.777011",{"slug":2890,"name":2890,"fn":2891,"description":2892,"org":2893,"tags":2894,"stars":2824,"repoUrl":2825,"updatedAt":2899},"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},[2895,2898],{"name":2896,"slug":2897,"type":15},"QA","qa",{"name":2742,"slug":2743,"type":15},"2026-07-14T05:25:53.673039",{"slug":2901,"name":2901,"fn":2902,"description":2903,"org":2904,"tags":2905,"stars":2824,"repoUrl":2825,"updatedAt":2908},"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},[2906,2907],{"name":2669,"slug":2666,"type":15},{"name":2715,"slug":2716,"type":15},"2026-07-14T05:25:49.362534",{"slug":2910,"name":2910,"fn":2911,"description":2912,"org":2913,"tags":2914,"stars":2824,"repoUrl":2825,"updatedAt":2922},"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},[2915,2918,2919],{"name":2916,"slug":2917,"type":15},"Code Review","code-review",{"name":2847,"slug":2848,"type":15},{"name":2920,"slug":2921,"type":15},"Pull Requests","pull-requests","2026-07-14T05:26:01.226578",{"slug":2924,"name":2924,"fn":2925,"description":2926,"org":2927,"tags":2928,"stars":2824,"repoUrl":2825,"updatedAt":2931},"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},[2929,2930],{"name":2896,"slug":2897,"type":15},{"name":2742,"slug":2743,"type":15},"2026-07-14T05:25:54.928983",{"slug":2933,"name":2933,"fn":2934,"description":2935,"org":2936,"tags":2937,"stars":2824,"repoUrl":2825,"updatedAt":2940},"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},[2938,2939],{"name":2752,"slug":2753,"type":15},{"name":2834,"slug":2835,"type":15},"2026-07-30T05:29:03.275638",496]