[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill-nvidia-tao-train-mask-grounding-dino":3,"mdc--3tc2h1-key":31,"related-repo-nvidia-tao-train-mask-grounding-dino":2352,"related-org-nvidia-tao-train-mask-grounding-dino":2456},{"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-mask-grounding-dino","train Mask Grounding DINO models","Mask Grounding DINO for grounded instance segmentation. Extends Grounding DINO with a mask-prediction head for open-set segmentation guided by text prompts. Use when training, evaluating, exporting, quantizing, or running inference for a TAO Mask-Grounding-DINO model. Trigger phrases include \"train Mask Grounding DINO\", \"open-vocabulary segmentation\", \"text-prompted instance segmentation\", \"grounded mask DETR\".",{"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},"Deep Learning","deep-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","2026-07-14T05:29:56.871897","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-mask-grounding-dino","---\nname: tao-train-mask-grounding-dino\ndescription: Mask Grounding DINO for grounded instance segmentation. Extends Grounding DINO with a mask-prediction head for\n  open-set segmentation guided by text prompts. Use when training, evaluating, exporting, quantizing, or running inference for\n  a TAO Mask-Grounding-DINO model. Trigger phrases include \"train Mask Grounding DINO\", \"open-vocabulary segmentation\",\n  \"text-prompted instance segmentation\", \"grounded mask DETR\".\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- segmentation\n---\n\n# Mask Grounding DINO\n\nMask Grounding DINO for grounded instance segmentation. Extends Grounding DINO with mask prediction head for open-set segmentation guided by text prompts.\n\nSet train.pretrained_model_path for full model weights.\n\nFor TAO Deploy TensorRT actions (`gen_trt_engine`, TensorRT `evaluate`, and TensorRT `inference`), read `references\u002Ftao-deploy-mask-grounding-dino.md` first. Deploy spec templates live in this skill's `references\u002F` folder with the `spec_template_deploy_*.yaml` prefix.\n\n## Dataclass Schemas\n\nGenerated TAO Core schemas are packaged in `schemas\u002F\u003Caction>.schema.json`, with `schemas\u002Fmanifest.json` listing available actions. Each generated schema also emits `references\u002Fspec_template_\u003Caction>.yaml` from the schema top-level `default` field. AutoML enablement is declared at the model layer in `references\u002Fskill_info.yaml` via `automl_enabled`. Runnable AutoML still requires `schemas\u002Ftrain.schema.json` and `references\u002Fspec_template_train.yaml` to exist and parse. Use the packaged train schema for `automl_default_parameters`, `automl_disabled_parameters`, defaults, min\u002Fmax bounds, enums, option weights, math conditions, dependencies, and popular parameters. Do not expect `~\u002Ftao-core` at runtime; maintainers regenerate schemas\u002Ftemplates before packaging the skill bank.\n\n## Train Action Policy\n\nThis model is AutoML-enabled at the model layer. Before handling any train-stage request, read `references\u002Fskill_info.yaml` and resolve the run override from either an explicit `automl_policy` value or the user's workflow request. Use `automl_policy: on` by default and only expose `on` \u002F `off` in new launch prompts. Treat phrases like \"turn off AutoML\", \"disable AutoML\", \"no HPO\", or \"plain training\" as `automl_policy: off` for this run only. When `automl_policy: on`, `automl_enabled: true`, and both `schemas\u002Ftrain.schema.json` and `references\u002Fspec_template_train.yaml` are packaged, route the train action through `tao-skill-bank:tao-run-automl` by default with this model's `skill_dir`. Preserve workflow\u002Fapplication overrides for datasets, specs, output directories, GPU\u002Fplatform settings, parent checkpoints, and `automl_policy`. Use direct model training only when `automl_policy: off` or the packaged train schema\u002Ftemplate is missing; in the missing-schema case, report that AutoML is enabled but not runnable for this model until schemas are generated.\n\nNon-train actions such as `evaluate`, `inference`, `export`, and deploy flows stay in this model skill. The per-run `automl_policy` override does not change model metadata.\n\n## Training Requirements\n\n- **Dataset type:** segmentation\n- **Formats:** odvg, coco, coco_raw\n- **Monitoring metric:** val_loss\n\n### Per-Action Dataset Requirements\n\n| Action | Spec Key | Source | Files | List? |\n|---|---|---|---|---|\n| evaluate | dataset.test_data_sources | eval_dataset | image_dir: images.tar.gz, json_file: annotations.json | No |\n| evaluate | dataset.test_data_sources.data_type | eval_dataset | OD | No |\n| inference | dataset.infer_data_sources | inference_dataset | image_dir: images.tar.gz, captions: text prompts | No |\n| inference | dataset.infer_data_sources.data_type | inference_dataset | OD | No |\n| quantize | dataset.train_data_sources | train_datasets | image_dir: images.tar.gz, json_file: annotations_odvg.jsonl, label_map: annotations_odvg_labelmap.json | Yes |\n| quantize | dataset.val_data_sources | eval_dataset | image_dir: images.tar.gz, json_file: annotations.json | No |\n| quantize | dataset.val_data_sources.data_type | eval_dataset | OD | No |\n| quantize | dataset.quant_calibration_data_sources | train_datasets | image_dir: images.tar.gz, json_file: annotations_odvg.jsonl, label_map: annotations_odvg_labelmap.json | No |\n| train | dataset.train_data_sources | train_datasets | image_dir: images.tar.gz, json_file: annotations_odvg.jsonl, label_map: annotations_odvg_labelmap.json | Yes |\n| train | dataset.val_data_sources | eval_dataset | image_dir: images.tar.gz, json_file: annotations.json | No |\n| train | dataset.val_data_sources.data_type | eval_dataset | OD | No |\n\n### Typical Spec Overrides\n\nData source overrides are **mandatory for every action** — the agent MUST construct data source paths from the Per-Action Dataset Requirements table above and include them in `spec_overrides`.\n\n```python\nS3_TRAIN = \"s3:\u002F\u002Fbucket\u002Fdata\u002Ftrain\"\nS3_EVAL = \"s3:\u002F\u002Fbucket\u002Fdata\u002Feval\"\n```\n\n**train (mandatory data sources):**\n```python\n{\n    \"train.num_gpus\": 1,\n    \"train.num_epochs\": 10,\n    \"train.checkpoint_interval\": 10,\n    \"train.validation_interval\": 10,\n    \"dataset.val_data_sources.data_type\": \"OD\",\n    \"model.num_region_queries\": 100,\n    \"dataset.train_data_sources\": [{\"image_dir\": f\"{S3_TRAIN}\u002Fimages.tar.gz\", \"json_file\": f\"{S3_TRAIN}\u002Fannotations_odvg.jsonl\", \"label_map\": f\"{S3_TRAIN}\u002Fannotations_odvg_labelmap.json\"}],\n    \"dataset.val_data_sources\": {\"image_dir\": f\"{S3_EVAL}\u002Fimages.tar.gz\", \"json_file\": f\"{S3_EVAL}\u002Fannotations.json\"},\n}\n```\n\n**evaluate (mandatory data sources):**\n```python\n{\n    \"evaluate.checkpoint\": \"\u003Cselected train\u002FAutoML checkpoint>\",\n    \"dataset.test_data_sources.data_type\": \"OD\",\n    \"dataset.test_data_sources\": {\"image_dir\": f\"{S3_EVAL}\u002Fimages.tar.gz\", \"json_file\": f\"{S3_EVAL}\u002Fannotations.json\"},\n}\n```\n\n**inference (mandatory data sources):**\n```python\n{\n    \"inference.checkpoint\": \"\u003Cselected train\u002FAutoML checkpoint>\",\n    \"dataset.infer_data_sources.data_type\": \"OD\",\n    \"dataset.infer_data_sources\": {\"image_dir\": f\"{S3_EVAL}\u002Fimages.tar.gz\", \"captions\": [\"person\", \"bicycle\", \"car\"]},\n}\n```\n\n**quantize (mandatory data sources):**\n```python\n{\n    \"dataset.train_data_sources\": [{\"image_dir\": f\"{S3_TRAIN}\u002Fimages.tar.gz\", \"json_file\": f\"{S3_TRAIN}\u002Fannotations_odvg.jsonl\", \"label_map\": f\"{S3_TRAIN}\u002Fannotations_odvg_labelmap.json\"}],\n    \"dataset.val_data_sources\": {\"image_dir\": f\"{S3_EVAL}\u002Fimages.tar.gz\", \"json_file\": f\"{S3_EVAL}\u002Fannotations.json\"},\n    \"dataset.quant_calibration_data_sources\": {\"image_dir\": f\"{S3_TRAIN}\u002Fimages.tar.gz\", \"json_file\": f\"{S3_TRAIN}\u002Fannotations_odvg.jsonl\", \"label_map\": f\"{S3_TRAIN}\u002Fannotations_odvg_labelmap.json\"},\n}\n```\n## Eval Dataset\n\nOptional. Validation uses COCO-format annotations even when training uses ODVG.\n\n## Important Parameters\n\n- **model.backbone**: Default swin_tiny_224_1k. Same backbone options as Grounding DINO.\n- **train.optim.lr**: Learning rate. Default 2e-4. lr_backbone 2e-5. Reuses GDINOTrainExpConfig — same training setup as Grounding DINO.\n- **model.num_queries**: Object queries. Default 900.\n- **model.enc_layers \u002F model.dec_layers**: Keep both at 6 for train\u002FAutoML\n  runs. The mask head asserts six decoder outputs during validation, so\n  copying Grounding DINO smoke overrides that reduce transformer layers causes\n  an immediate failure.\n- **AutoML metric note**: Use `metric=\"val_loss\"` with\n  `direction=\"minimize\"` for train-stage AutoML. The packaged train loop logs\n  validation loss scalars; it does not emit `[bbox] val_mAP@50` during the\n  train job.\n- **model.has_mask**: Enables mask prediction head. Default True. Adds mask\u002Fdice\u002Frela loss coefficients.\n- **model.num_region_queries**: Number of region queries for mask prediction. Default 100.\n- **model.loss_types**: Loss components. Default [labels, boxes, masks]. Includes mask_loss_coef, dice_loss_coef, rela_loss_coef.\n- **evaluate.ioi_threshold**: IoI threshold for mask evaluation. Default 0.5.\n- **evaluate.nms_threshold**: NMS threshold. Default 0.2.\n- **evaluate.text_threshold**: Text matching threshold. Default 0.3.\n- **dataset.has_mask**: Dataset includes mask annotations. Default True. val_data_sources default data_type is \"VG\".\n\n## Multi-GPU \u002F Multi-Node\n\n**Launch method:** Lightning-managed. Same DDP\u002FFSDP behavior as Grounding DINO.\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.num_nodes` | Number of nodes | 1 |\n| `train.distributed_strategy` | `ddp` or `fsdp` | `ddp` |\n\n## Hardware\n\nMinimum 1 GPU(s), recommended 4 GPU(s). 24GB+ (A100 recommended) VRAM per GPU. Heavier than Grounding DINO due to mask prediction head. 24GB+ GPU memory recommended.\n\n## Error Patterns\n\n**CUDA out of memory**: Reduce batch_size. Mask prediction adds overhead on top of Grounding DINO.\n\n**Deploy schema error for `test_threshold`**: TAO Deploy uses\n`evaluate.text_threshold` and `inference.text_threshold`. Do not use\n`test_threshold` in deploy specs.\n\n**Deploy model shape mismatch**: Carry transformer and mask structure fields\nfrom export into deploy evaluate\u002Finference specs, including `model.num_queries`,\n`model.num_select`, `model.max_text_len`, `model.num_region_queries`, and\n`model.has_mask`. These values must match the ONNX model used to build the\nTensorRT engine.\n\n## Spec Param \u002F Parent Model Inference\n\nModel-specific inference mappings belong in this MD file, not in `config.json`. Generated runners should read this section and apply the mappings with SDK helpers before `create_job()`. This mirrors the old microservices `infer_params.py` flow.\n\nInference mappings from TAO Core `mask_grounding_dino.config.json`:\n\n| Action | Spec Field | Inference Function | Meaning |\n|---|---|---|---|\n| evaluate | `encryption_key` | `key` | encryption key |\n| evaluate | `evaluate.checkpoint` | `parent_model` | model file inferred from the parent job results folder |\n| evaluate | `evaluate.trt_engine` | `parent_model` | model file inferred from the parent job results folder |\n| evaluate | `results_dir` | `output_dir` | current job results directory |\n| export | `encryption_key` | `key` | encryption key |\n| export | `export.checkpoint` | `parent_model` | model file inferred from the parent job results folder |\n| export | `export.onnx_file` | `create_onnx_file` | output ONNX path |\n| export | `results_dir` | `output_dir` | current job results directory |\n| gen_trt_engine | `encryption_key` | `key` | encryption key |\n| gen_trt_engine | `gen_trt_engine.onnx_file` | `parent_model` | model file inferred from the parent job results folder |\n| gen_trt_engine | `gen_trt_engine.trt_engine` | `create_engine_file` | output TensorRT engine path |\n| gen_trt_engine | `results_dir` | `output_dir` | current job results directory |\n| inference | `encryption_key` | `key` | encryption key |\n| inference | `inference.checkpoint` | `parent_model` | model file inferred from the parent job results folder |\n| inference | `inference.trt_engine` | `parent_model` | model file inferred from the parent job results folder |\n| inference | `results_dir` | `output_dir` | current job results directory |\n| quantize | `encryption_key` | `key` | encryption key |\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| train | `encryption_key` | `key` | encryption key |\n| train | `model.pretrained_backbone_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.pretrained_model_path` | `ptm_if_no_resume_model` | PTM when no resume checkpoint exists |\n| train | `train.resume_training_checkpoint_path` | `resume_model` | model file inferred from the current job results folder |\n\nFor `parent_model` or `parent_model_folder`, pass the upstream train\u002Fexport\u002FAutoML child job id as `parent_job_id`. The SDK lists the parent result folder, filters checkpoint artifacts, and returns the selected model file or folder. Do not add these mappings back to `config.json` and do not patch generated runner scripts to guess checkpoint paths.\n\nWhen selecting a Mask Grounding DINO checkpoint outside the SDK resolver, match\nthe intended epoch\u002Fstep artifact exactly, for example\n`model_epoch_000_step_00049.pth`. The `mask_gdino_model_latest.pth` symlink is\nvalid only when latest is explicitly requested. The parent PyTorch\n`mask_grounding_dino` CLI supports `train`, `evaluate`, `inference`, `export`,\nand `quantize`; run TensorRT engine generation, TensorRT inference, and\nTensorRT evaluation through `references\u002Ftao-deploy-mask-grounding-dino.md`.\n\n## Deployment\n\n- [tao-deploy-mask-grounding-dino](references\u002Ftao-deploy-mask-grounding-dino.md)\n",{"data":32,"body":40},{"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],"segmentation",{"type":41,"children":42},"root",[43,52,58,63,117,124,217,223,332,364,370,406,413,728,734,754,785,793,888,896,941,949,994,1002,1045,1051,1056,1062,1216,1222,1232,1363,1369,1374,1380,1390,1427,1472,1478,1507,1520,2227,2261,2328,2334,2346],{"type":44,"tag":45,"props":46,"children":48},"element","h1",{"id":47},"mask-grounding-dino",[49],{"type":50,"value":51},"text","Mask Grounding DINO",{"type":44,"tag":53,"props":54,"children":55},"p",{},[56],{"type":50,"value":57},"Mask Grounding DINO for grounded instance segmentation. Extends Grounding DINO with mask prediction head for open-set segmentation guided by text prompts.",{"type":44,"tag":53,"props":59,"children":60},{},[61],{"type":50,"value":62},"Set train.pretrained_model_path for full model weights.",{"type":44,"tag":53,"props":64,"children":65},{},[66,68,75,77,83,85,91,93,99,101,107,109,115],{"type":50,"value":67},"For TAO Deploy TensorRT actions (",{"type":44,"tag":69,"props":70,"children":72},"code",{"className":71},[],[73],{"type":50,"value":74},"gen_trt_engine",{"type":50,"value":76},", TensorRT ",{"type":44,"tag":69,"props":78,"children":80},{"className":79},[],[81],{"type":50,"value":82},"evaluate",{"type":50,"value":84},", and TensorRT ",{"type":44,"tag":69,"props":86,"children":88},{"className":87},[],[89],{"type":50,"value":90},"inference",{"type":50,"value":92},"), read ",{"type":44,"tag":69,"props":94,"children":96},{"className":95},[],[97],{"type":50,"value":98},"references\u002Ftao-deploy-mask-grounding-dino.md",{"type":50,"value":100}," first. Deploy spec templates live in this skill's ",{"type":44,"tag":69,"props":102,"children":104},{"className":103},[],[105],{"type":50,"value":106},"references\u002F",{"type":50,"value":108}," folder with the ",{"type":44,"tag":69,"props":110,"children":112},{"className":111},[],[113],{"type":50,"value":114},"spec_template_deploy_*.yaml",{"type":50,"value":116}," prefix.",{"type":44,"tag":118,"props":119,"children":121},"h2",{"id":120},"dataclass-schemas",[122],{"type":50,"value":123},"Dataclass Schemas",{"type":44,"tag":53,"props":125,"children":126},{},[127,129,135,137,143,145,151,153,159,161,167,169,175,177,183,185,191,193,199,201,207,209,215],{"type":50,"value":128},"Generated TAO Core schemas are packaged in ",{"type":44,"tag":69,"props":130,"children":132},{"className":131},[],[133],{"type":50,"value":134},"schemas\u002F\u003Caction>.schema.json",{"type":50,"value":136},", with ",{"type":44,"tag":69,"props":138,"children":140},{"className":139},[],[141],{"type":50,"value":142},"schemas\u002Fmanifest.json",{"type":50,"value":144}," listing available actions. Each generated schema also emits ",{"type":44,"tag":69,"props":146,"children":148},{"className":147},[],[149],{"type":50,"value":150},"references\u002Fspec_template_\u003Caction>.yaml",{"type":50,"value":152}," from the schema top-level ",{"type":44,"tag":69,"props":154,"children":156},{"className":155},[],[157],{"type":50,"value":158},"default",{"type":50,"value":160}," field. AutoML enablement is declared at the model layer in ",{"type":44,"tag":69,"props":162,"children":164},{"className":163},[],[165],{"type":50,"value":166},"references\u002Fskill_info.yaml",{"type":50,"value":168}," via ",{"type":44,"tag":69,"props":170,"children":172},{"className":171},[],[173],{"type":50,"value":174},"automl_enabled",{"type":50,"value":176},". Runnable AutoML still requires ",{"type":44,"tag":69,"props":178,"children":180},{"className":179},[],[181],{"type":50,"value":182},"schemas\u002Ftrain.schema.json",{"type":50,"value":184}," and ",{"type":44,"tag":69,"props":186,"children":188},{"className":187},[],[189],{"type":50,"value":190},"references\u002Fspec_template_train.yaml",{"type":50,"value":192}," to exist and parse. Use the packaged train schema for ",{"type":44,"tag":69,"props":194,"children":196},{"className":195},[],[197],{"type":50,"value":198},"automl_default_parameters",{"type":50,"value":200},", ",{"type":44,"tag":69,"props":202,"children":204},{"className":203},[],[205],{"type":50,"value":206},"automl_disabled_parameters",{"type":50,"value":208},", defaults, min\u002Fmax bounds, enums, option weights, math conditions, dependencies, and popular parameters. Do not expect ",{"type":44,"tag":69,"props":210,"children":212},{"className":211},[],[213],{"type":50,"value":214},"~\u002Ftao-core",{"type":50,"value":216}," at runtime; maintainers regenerate schemas\u002Ftemplates before packaging the skill bank.",{"type":44,"tag":118,"props":218,"children":220},{"id":219},"train-action-policy",[221],{"type":50,"value":222},"Train Action Policy",{"type":44,"tag":53,"props":224,"children":225},{},[226,228,233,235,241,243,249,251,257,259,265,267,273,275,280,281,287,289,294,295,300,302,308,310,316,318,323,325,330],{"type":50,"value":227},"This model is AutoML-enabled at the model layer. Before handling any train-stage request, read ",{"type":44,"tag":69,"props":229,"children":231},{"className":230},[],[232],{"type":50,"value":166},{"type":50,"value":234}," and resolve the run override from either an explicit ",{"type":44,"tag":69,"props":236,"children":238},{"className":237},[],[239],{"type":50,"value":240},"automl_policy",{"type":50,"value":242}," value or the user's workflow request. Use ",{"type":44,"tag":69,"props":244,"children":246},{"className":245},[],[247],{"type":50,"value":248},"automl_policy: on",{"type":50,"value":250}," by default and only expose ",{"type":44,"tag":69,"props":252,"children":254},{"className":253},[],[255],{"type":50,"value":256},"on",{"type":50,"value":258}," \u002F ",{"type":44,"tag":69,"props":260,"children":262},{"className":261},[],[263],{"type":50,"value":264},"off",{"type":50,"value":266}," in new launch prompts. Treat phrases like \"turn off AutoML\", \"disable AutoML\", \"no HPO\", or \"plain training\" as ",{"type":44,"tag":69,"props":268,"children":270},{"className":269},[],[271],{"type":50,"value":272},"automl_policy: off",{"type":50,"value":274}," for this run only. When ",{"type":44,"tag":69,"props":276,"children":278},{"className":277},[],[279],{"type":50,"value":248},{"type":50,"value":200},{"type":44,"tag":69,"props":282,"children":284},{"className":283},[],[285],{"type":50,"value":286},"automl_enabled: true",{"type":50,"value":288},", and both ",{"type":44,"tag":69,"props":290,"children":292},{"className":291},[],[293],{"type":50,"value":182},{"type":50,"value":184},{"type":44,"tag":69,"props":296,"children":298},{"className":297},[],[299],{"type":50,"value":190},{"type":50,"value":301}," are packaged, route the train action through ",{"type":44,"tag":69,"props":303,"children":305},{"className":304},[],[306],{"type":50,"value":307},"tao-skill-bank:tao-run-automl",{"type":50,"value":309}," by default with this model's ",{"type":44,"tag":69,"props":311,"children":313},{"className":312},[],[314],{"type":50,"value":315},"skill_dir",{"type":50,"value":317},". Preserve workflow\u002Fapplication overrides for datasets, specs, output directories, GPU\u002Fplatform settings, parent checkpoints, and ",{"type":44,"tag":69,"props":319,"children":321},{"className":320},[],[322],{"type":50,"value":240},{"type":50,"value":324},". Use direct model training only when ",{"type":44,"tag":69,"props":326,"children":328},{"className":327},[],[329],{"type":50,"value":272},{"type":50,"value":331}," 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":44,"tag":53,"props":333,"children":334},{},[335,337,342,343,348,349,355,357,362],{"type":50,"value":336},"Non-train actions such as ",{"type":44,"tag":69,"props":338,"children":340},{"className":339},[],[341],{"type":50,"value":82},{"type":50,"value":200},{"type":44,"tag":69,"props":344,"children":346},{"className":345},[],[347],{"type":50,"value":90},{"type":50,"value":200},{"type":44,"tag":69,"props":350,"children":352},{"className":351},[],[353],{"type":50,"value":354},"export",{"type":50,"value":356},", and deploy flows stay in this model skill. The per-run ",{"type":44,"tag":69,"props":358,"children":360},{"className":359},[],[361],{"type":50,"value":240},{"type":50,"value":363}," override does not change model metadata.",{"type":44,"tag":118,"props":365,"children":367},{"id":366},"training-requirements",[368],{"type":50,"value":369},"Training Requirements",{"type":44,"tag":371,"props":372,"children":373},"ul",{},[374,386,396],{"type":44,"tag":375,"props":376,"children":377},"li",{},[378,384],{"type":44,"tag":379,"props":380,"children":381},"strong",{},[382],{"type":50,"value":383},"Dataset type:",{"type":50,"value":385}," segmentation",{"type":44,"tag":375,"props":387,"children":388},{},[389,394],{"type":44,"tag":379,"props":390,"children":391},{},[392],{"type":50,"value":393},"Formats:",{"type":50,"value":395}," odvg, coco, coco_raw",{"type":44,"tag":375,"props":397,"children":398},{},[399,404],{"type":44,"tag":379,"props":400,"children":401},{},[402],{"type":50,"value":403},"Monitoring metric:",{"type":50,"value":405}," val_loss",{"type":44,"tag":407,"props":408,"children":410},"h3",{"id":409},"per-action-dataset-requirements",[411],{"type":50,"value":412},"Per-Action Dataset Requirements",{"type":44,"tag":414,"props":415,"children":416},"table",{},[417,451],{"type":44,"tag":418,"props":419,"children":420},"thead",{},[421],{"type":44,"tag":422,"props":423,"children":424},"tr",{},[425,431,436,441,446],{"type":44,"tag":426,"props":427,"children":428},"th",{},[429],{"type":50,"value":430},"Action",{"type":44,"tag":426,"props":432,"children":433},{},[434],{"type":50,"value":435},"Spec Key",{"type":44,"tag":426,"props":437,"children":438},{},[439],{"type":50,"value":440},"Source",{"type":44,"tag":426,"props":442,"children":443},{},[444],{"type":50,"value":445},"Files",{"type":44,"tag":426,"props":447,"children":448},{},[449],{"type":50,"value":450},"List?",{"type":44,"tag":452,"props":453,"children":454},"tbody",{},[455,483,508,534,558,586,610,634,658,682,705],{"type":44,"tag":422,"props":456,"children":457},{},[458,463,468,473,478],{"type":44,"tag":459,"props":460,"children":461},"td",{},[462],{"type":50,"value":82},{"type":44,"tag":459,"props":464,"children":465},{},[466],{"type":50,"value":467},"dataset.test_data_sources",{"type":44,"tag":459,"props":469,"children":470},{},[471],{"type":50,"value":472},"eval_dataset",{"type":44,"tag":459,"props":474,"children":475},{},[476],{"type":50,"value":477},"image_dir: images.tar.gz, json_file: annotations.json",{"type":44,"tag":459,"props":479,"children":480},{},[481],{"type":50,"value":482},"No",{"type":44,"tag":422,"props":484,"children":485},{},[486,490,495,499,504],{"type":44,"tag":459,"props":487,"children":488},{},[489],{"type":50,"value":82},{"type":44,"tag":459,"props":491,"children":492},{},[493],{"type":50,"value":494},"dataset.test_data_sources.data_type",{"type":44,"tag":459,"props":496,"children":497},{},[498],{"type":50,"value":472},{"type":44,"tag":459,"props":500,"children":501},{},[502],{"type":50,"value":503},"OD",{"type":44,"tag":459,"props":505,"children":506},{},[507],{"type":50,"value":482},{"type":44,"tag":422,"props":509,"children":510},{},[511,515,520,525,530],{"type":44,"tag":459,"props":512,"children":513},{},[514],{"type":50,"value":90},{"type":44,"tag":459,"props":516,"children":517},{},[518],{"type":50,"value":519},"dataset.infer_data_sources",{"type":44,"tag":459,"props":521,"children":522},{},[523],{"type":50,"value":524},"inference_dataset",{"type":44,"tag":459,"props":526,"children":527},{},[528],{"type":50,"value":529},"image_dir: images.tar.gz, captions: text prompts",{"type":44,"tag":459,"props":531,"children":532},{},[533],{"type":50,"value":482},{"type":44,"tag":422,"props":535,"children":536},{},[537,541,546,550,554],{"type":44,"tag":459,"props":538,"children":539},{},[540],{"type":50,"value":90},{"type":44,"tag":459,"props":542,"children":543},{},[544],{"type":50,"value":545},"dataset.infer_data_sources.data_type",{"type":44,"tag":459,"props":547,"children":548},{},[549],{"type":50,"value":524},{"type":44,"tag":459,"props":551,"children":552},{},[553],{"type":50,"value":503},{"type":44,"tag":459,"props":555,"children":556},{},[557],{"type":50,"value":482},{"type":44,"tag":422,"props":559,"children":560},{},[561,566,571,576,581],{"type":44,"tag":459,"props":562,"children":563},{},[564],{"type":50,"value":565},"quantize",{"type":44,"tag":459,"props":567,"children":568},{},[569],{"type":50,"value":570},"dataset.train_data_sources",{"type":44,"tag":459,"props":572,"children":573},{},[574],{"type":50,"value":575},"train_datasets",{"type":44,"tag":459,"props":577,"children":578},{},[579],{"type":50,"value":580},"image_dir: images.tar.gz, json_file: annotations_odvg.jsonl, label_map: annotations_odvg_labelmap.json",{"type":44,"tag":459,"props":582,"children":583},{},[584],{"type":50,"value":585},"Yes",{"type":44,"tag":422,"props":587,"children":588},{},[589,593,598,602,606],{"type":44,"tag":459,"props":590,"children":591},{},[592],{"type":50,"value":565},{"type":44,"tag":459,"props":594,"children":595},{},[596],{"type":50,"value":597},"dataset.val_data_sources",{"type":44,"tag":459,"props":599,"children":600},{},[601],{"type":50,"value":472},{"type":44,"tag":459,"props":603,"children":604},{},[605],{"type":50,"value":477},{"type":44,"tag":459,"props":607,"children":608},{},[609],{"type":50,"value":482},{"type":44,"tag":422,"props":611,"children":612},{},[613,617,622,626,630],{"type":44,"tag":459,"props":614,"children":615},{},[616],{"type":50,"value":565},{"type":44,"tag":459,"props":618,"children":619},{},[620],{"type":50,"value":621},"dataset.val_data_sources.data_type",{"type":44,"tag":459,"props":623,"children":624},{},[625],{"type":50,"value":472},{"type":44,"tag":459,"props":627,"children":628},{},[629],{"type":50,"value":503},{"type":44,"tag":459,"props":631,"children":632},{},[633],{"type":50,"value":482},{"type":44,"tag":422,"props":635,"children":636},{},[637,641,646,650,654],{"type":44,"tag":459,"props":638,"children":639},{},[640],{"type":50,"value":565},{"type":44,"tag":459,"props":642,"children":643},{},[644],{"type":50,"value":645},"dataset.quant_calibration_data_sources",{"type":44,"tag":459,"props":647,"children":648},{},[649],{"type":50,"value":575},{"type":44,"tag":459,"props":651,"children":652},{},[653],{"type":50,"value":580},{"type":44,"tag":459,"props":655,"children":656},{},[657],{"type":50,"value":482},{"type":44,"tag":422,"props":659,"children":660},{},[661,666,670,674,678],{"type":44,"tag":459,"props":662,"children":663},{},[664],{"type":50,"value":665},"train",{"type":44,"tag":459,"props":667,"children":668},{},[669],{"type":50,"value":570},{"type":44,"tag":459,"props":671,"children":672},{},[673],{"type":50,"value":575},{"type":44,"tag":459,"props":675,"children":676},{},[677],{"type":50,"value":580},{"type":44,"tag":459,"props":679,"children":680},{},[681],{"type":50,"value":585},{"type":44,"tag":422,"props":683,"children":684},{},[685,689,693,697,701],{"type":44,"tag":459,"props":686,"children":687},{},[688],{"type":50,"value":665},{"type":44,"tag":459,"props":690,"children":691},{},[692],{"type":50,"value":597},{"type":44,"tag":459,"props":694,"children":695},{},[696],{"type":50,"value":472},{"type":44,"tag":459,"props":698,"children":699},{},[700],{"type":50,"value":477},{"type":44,"tag":459,"props":702,"children":703},{},[704],{"type":50,"value":482},{"type":44,"tag":422,"props":706,"children":707},{},[708,712,716,720,724],{"type":44,"tag":459,"props":709,"children":710},{},[711],{"type":50,"value":665},{"type":44,"tag":459,"props":713,"children":714},{},[715],{"type":50,"value":621},{"type":44,"tag":459,"props":717,"children":718},{},[719],{"type":50,"value":472},{"type":44,"tag":459,"props":721,"children":722},{},[723],{"type":50,"value":503},{"type":44,"tag":459,"props":725,"children":726},{},[727],{"type":50,"value":482},{"type":44,"tag":407,"props":729,"children":731},{"id":730},"typical-spec-overrides",[732],{"type":50,"value":733},"Typical Spec Overrides",{"type":44,"tag":53,"props":735,"children":736},{},[737,739,744,746,752],{"type":50,"value":738},"Data source overrides are ",{"type":44,"tag":379,"props":740,"children":741},{},[742],{"type":50,"value":743},"mandatory for every action",{"type":50,"value":745}," — the agent MUST construct data source paths from the Per-Action Dataset Requirements table above and include them in ",{"type":44,"tag":69,"props":747,"children":749},{"className":748},[],[750],{"type":50,"value":751},"spec_overrides",{"type":50,"value":753},".",{"type":44,"tag":755,"props":756,"children":761},"pre",{"className":757,"code":758,"language":759,"meta":760,"style":760},"language-python shiki shiki-themes material-theme-lighter material-theme material-theme-palenight","S3_TRAIN = \"s3:\u002F\u002Fbucket\u002Fdata\u002Ftrain\"\nS3_EVAL = \"s3:\u002F\u002Fbucket\u002Fdata\u002Feval\"\n","python","",[762],{"type":44,"tag":69,"props":763,"children":764},{"__ignoreMap":760},[765,776],{"type":44,"tag":766,"props":767,"children":770},"span",{"class":768,"line":769},"line",1,[771],{"type":44,"tag":766,"props":772,"children":773},{},[774],{"type":50,"value":775},"S3_TRAIN = \"s3:\u002F\u002Fbucket\u002Fdata\u002Ftrain\"\n",{"type":44,"tag":766,"props":777,"children":779},{"class":768,"line":778},2,[780],{"type":44,"tag":766,"props":781,"children":782},{},[783],{"type":50,"value":784},"S3_EVAL = \"s3:\u002F\u002Fbucket\u002Fdata\u002Feval\"\n",{"type":44,"tag":53,"props":786,"children":787},{},[788],{"type":44,"tag":379,"props":789,"children":790},{},[791],{"type":50,"value":792},"train (mandatory data sources):",{"type":44,"tag":755,"props":794,"children":796},{"className":757,"code":795,"language":759,"meta":760,"style":760},"{\n    \"train.num_gpus\": 1,\n    \"train.num_epochs\": 10,\n    \"train.checkpoint_interval\": 10,\n    \"train.validation_interval\": 10,\n    \"dataset.val_data_sources.data_type\": \"OD\",\n    \"model.num_region_queries\": 100,\n    \"dataset.train_data_sources\": [{\"image_dir\": f\"{S3_TRAIN}\u002Fimages.tar.gz\", \"json_file\": f\"{S3_TRAIN}\u002Fannotations_odvg.jsonl\", \"label_map\": f\"{S3_TRAIN}\u002Fannotations_odvg_labelmap.json\"}],\n    \"dataset.val_data_sources\": {\"image_dir\": f\"{S3_EVAL}\u002Fimages.tar.gz\", \"json_file\": f\"{S3_EVAL}\u002Fannotations.json\"},\n}\n",[797],{"type":44,"tag":69,"props":798,"children":799},{"__ignoreMap":760},[800,808,816,825,834,843,852,861,870,879],{"type":44,"tag":766,"props":801,"children":802},{"class":768,"line":769},[803],{"type":44,"tag":766,"props":804,"children":805},{},[806],{"type":50,"value":807},"{\n",{"type":44,"tag":766,"props":809,"children":810},{"class":768,"line":778},[811],{"type":44,"tag":766,"props":812,"children":813},{},[814],{"type":50,"value":815},"    \"train.num_gpus\": 1,\n",{"type":44,"tag":766,"props":817,"children":819},{"class":768,"line":818},3,[820],{"type":44,"tag":766,"props":821,"children":822},{},[823],{"type":50,"value":824},"    \"train.num_epochs\": 10,\n",{"type":44,"tag":766,"props":826,"children":828},{"class":768,"line":827},4,[829],{"type":44,"tag":766,"props":830,"children":831},{},[832],{"type":50,"value":833},"    \"train.checkpoint_interval\": 10,\n",{"type":44,"tag":766,"props":835,"children":837},{"class":768,"line":836},5,[838],{"type":44,"tag":766,"props":839,"children":840},{},[841],{"type":50,"value":842},"    \"train.validation_interval\": 10,\n",{"type":44,"tag":766,"props":844,"children":846},{"class":768,"line":845},6,[847],{"type":44,"tag":766,"props":848,"children":849},{},[850],{"type":50,"value":851},"    \"dataset.val_data_sources.data_type\": \"OD\",\n",{"type":44,"tag":766,"props":853,"children":855},{"class":768,"line":854},7,[856],{"type":44,"tag":766,"props":857,"children":858},{},[859],{"type":50,"value":860},"    \"model.num_region_queries\": 100,\n",{"type":44,"tag":766,"props":862,"children":864},{"class":768,"line":863},8,[865],{"type":44,"tag":766,"props":866,"children":867},{},[868],{"type":50,"value":869},"    \"dataset.train_data_sources\": [{\"image_dir\": f\"{S3_TRAIN}\u002Fimages.tar.gz\", \"json_file\": f\"{S3_TRAIN}\u002Fannotations_odvg.jsonl\", \"label_map\": f\"{S3_TRAIN}\u002Fannotations_odvg_labelmap.json\"}],\n",{"type":44,"tag":766,"props":871,"children":873},{"class":768,"line":872},9,[874],{"type":44,"tag":766,"props":875,"children":876},{},[877],{"type":50,"value":878},"    \"dataset.val_data_sources\": {\"image_dir\": f\"{S3_EVAL}\u002Fimages.tar.gz\", \"json_file\": f\"{S3_EVAL}\u002Fannotations.json\"},\n",{"type":44,"tag":766,"props":880,"children":882},{"class":768,"line":881},10,[883],{"type":44,"tag":766,"props":884,"children":885},{},[886],{"type":50,"value":887},"}\n",{"type":44,"tag":53,"props":889,"children":890},{},[891],{"type":44,"tag":379,"props":892,"children":893},{},[894],{"type":50,"value":895},"evaluate (mandatory data sources):",{"type":44,"tag":755,"props":897,"children":899},{"className":757,"code":898,"language":759,"meta":760,"style":760},"{\n    \"evaluate.checkpoint\": \"\u003Cselected train\u002FAutoML checkpoint>\",\n    \"dataset.test_data_sources.data_type\": \"OD\",\n    \"dataset.test_data_sources\": {\"image_dir\": f\"{S3_EVAL}\u002Fimages.tar.gz\", \"json_file\": f\"{S3_EVAL}\u002Fannotations.json\"},\n}\n",[900],{"type":44,"tag":69,"props":901,"children":902},{"__ignoreMap":760},[903,910,918,926,934],{"type":44,"tag":766,"props":904,"children":905},{"class":768,"line":769},[906],{"type":44,"tag":766,"props":907,"children":908},{},[909],{"type":50,"value":807},{"type":44,"tag":766,"props":911,"children":912},{"class":768,"line":778},[913],{"type":44,"tag":766,"props":914,"children":915},{},[916],{"type":50,"value":917},"    \"evaluate.checkpoint\": \"\u003Cselected train\u002FAutoML checkpoint>\",\n",{"type":44,"tag":766,"props":919,"children":920},{"class":768,"line":818},[921],{"type":44,"tag":766,"props":922,"children":923},{},[924],{"type":50,"value":925},"    \"dataset.test_data_sources.data_type\": \"OD\",\n",{"type":44,"tag":766,"props":927,"children":928},{"class":768,"line":827},[929],{"type":44,"tag":766,"props":930,"children":931},{},[932],{"type":50,"value":933},"    \"dataset.test_data_sources\": {\"image_dir\": f\"{S3_EVAL}\u002Fimages.tar.gz\", \"json_file\": f\"{S3_EVAL}\u002Fannotations.json\"},\n",{"type":44,"tag":766,"props":935,"children":936},{"class":768,"line":836},[937],{"type":44,"tag":766,"props":938,"children":939},{},[940],{"type":50,"value":887},{"type":44,"tag":53,"props":942,"children":943},{},[944],{"type":44,"tag":379,"props":945,"children":946},{},[947],{"type":50,"value":948},"inference (mandatory data sources):",{"type":44,"tag":755,"props":950,"children":952},{"className":757,"code":951,"language":759,"meta":760,"style":760},"{\n    \"inference.checkpoint\": \"\u003Cselected train\u002FAutoML checkpoint>\",\n    \"dataset.infer_data_sources.data_type\": \"OD\",\n    \"dataset.infer_data_sources\": {\"image_dir\": f\"{S3_EVAL}\u002Fimages.tar.gz\", \"captions\": [\"person\", \"bicycle\", \"car\"]},\n}\n",[953],{"type":44,"tag":69,"props":954,"children":955},{"__ignoreMap":760},[956,963,971,979,987],{"type":44,"tag":766,"props":957,"children":958},{"class":768,"line":769},[959],{"type":44,"tag":766,"props":960,"children":961},{},[962],{"type":50,"value":807},{"type":44,"tag":766,"props":964,"children":965},{"class":768,"line":778},[966],{"type":44,"tag":766,"props":967,"children":968},{},[969],{"type":50,"value":970},"    \"inference.checkpoint\": \"\u003Cselected train\u002FAutoML checkpoint>\",\n",{"type":44,"tag":766,"props":972,"children":973},{"class":768,"line":818},[974],{"type":44,"tag":766,"props":975,"children":976},{},[977],{"type":50,"value":978},"    \"dataset.infer_data_sources.data_type\": \"OD\",\n",{"type":44,"tag":766,"props":980,"children":981},{"class":768,"line":827},[982],{"type":44,"tag":766,"props":983,"children":984},{},[985],{"type":50,"value":986},"    \"dataset.infer_data_sources\": {\"image_dir\": f\"{S3_EVAL}\u002Fimages.tar.gz\", \"captions\": [\"person\", \"bicycle\", \"car\"]},\n",{"type":44,"tag":766,"props":988,"children":989},{"class":768,"line":836},[990],{"type":44,"tag":766,"props":991,"children":992},{},[993],{"type":50,"value":887},{"type":44,"tag":53,"props":995,"children":996},{},[997],{"type":44,"tag":379,"props":998,"children":999},{},[1000],{"type":50,"value":1001},"quantize (mandatory data sources):",{"type":44,"tag":755,"props":1003,"children":1005},{"className":757,"code":1004,"language":759,"meta":760,"style":760},"{\n    \"dataset.train_data_sources\": [{\"image_dir\": f\"{S3_TRAIN}\u002Fimages.tar.gz\", \"json_file\": f\"{S3_TRAIN}\u002Fannotations_odvg.jsonl\", \"label_map\": f\"{S3_TRAIN}\u002Fannotations_odvg_labelmap.json\"}],\n    \"dataset.val_data_sources\": {\"image_dir\": f\"{S3_EVAL}\u002Fimages.tar.gz\", \"json_file\": f\"{S3_EVAL}\u002Fannotations.json\"},\n    \"dataset.quant_calibration_data_sources\": {\"image_dir\": f\"{S3_TRAIN}\u002Fimages.tar.gz\", \"json_file\": f\"{S3_TRAIN}\u002Fannotations_odvg.jsonl\", \"label_map\": f\"{S3_TRAIN}\u002Fannotations_odvg_labelmap.json\"},\n}\n",[1006],{"type":44,"tag":69,"props":1007,"children":1008},{"__ignoreMap":760},[1009,1016,1023,1030,1038],{"type":44,"tag":766,"props":1010,"children":1011},{"class":768,"line":769},[1012],{"type":44,"tag":766,"props":1013,"children":1014},{},[1015],{"type":50,"value":807},{"type":44,"tag":766,"props":1017,"children":1018},{"class":768,"line":778},[1019],{"type":44,"tag":766,"props":1020,"children":1021},{},[1022],{"type":50,"value":869},{"type":44,"tag":766,"props":1024,"children":1025},{"class":768,"line":818},[1026],{"type":44,"tag":766,"props":1027,"children":1028},{},[1029],{"type":50,"value":878},{"type":44,"tag":766,"props":1031,"children":1032},{"class":768,"line":827},[1033],{"type":44,"tag":766,"props":1034,"children":1035},{},[1036],{"type":50,"value":1037},"    \"dataset.quant_calibration_data_sources\": {\"image_dir\": f\"{S3_TRAIN}\u002Fimages.tar.gz\", \"json_file\": f\"{S3_TRAIN}\u002Fannotations_odvg.jsonl\", \"label_map\": f\"{S3_TRAIN}\u002Fannotations_odvg_labelmap.json\"},\n",{"type":44,"tag":766,"props":1039,"children":1040},{"class":768,"line":836},[1041],{"type":44,"tag":766,"props":1042,"children":1043},{},[1044],{"type":50,"value":887},{"type":44,"tag":118,"props":1046,"children":1048},{"id":1047},"eval-dataset",[1049],{"type":50,"value":1050},"Eval Dataset",{"type":44,"tag":53,"props":1052,"children":1053},{},[1054],{"type":50,"value":1055},"Optional. Validation uses COCO-format annotations even when training uses ODVG.",{"type":44,"tag":118,"props":1057,"children":1059},{"id":1058},"important-parameters",[1060],{"type":50,"value":1061},"Important Parameters",{"type":44,"tag":371,"props":1063,"children":1064},{},[1065,1075,1085,1095,1105,1139,1149,1159,1176,1186,1196,1206],{"type":44,"tag":375,"props":1066,"children":1067},{},[1068,1073],{"type":44,"tag":379,"props":1069,"children":1070},{},[1071],{"type":50,"value":1072},"model.backbone",{"type":50,"value":1074},": Default swin_tiny_224_1k. Same backbone options as Grounding DINO.",{"type":44,"tag":375,"props":1076,"children":1077},{},[1078,1083],{"type":44,"tag":379,"props":1079,"children":1080},{},[1081],{"type":50,"value":1082},"train.optim.lr",{"type":50,"value":1084},": Learning rate. Default 2e-4. lr_backbone 2e-5. Reuses GDINOTrainExpConfig — same training setup as Grounding DINO.",{"type":44,"tag":375,"props":1086,"children":1087},{},[1088,1093],{"type":44,"tag":379,"props":1089,"children":1090},{},[1091],{"type":50,"value":1092},"model.num_queries",{"type":50,"value":1094},": Object queries. Default 900.",{"type":44,"tag":375,"props":1096,"children":1097},{},[1098,1103],{"type":44,"tag":379,"props":1099,"children":1100},{},[1101],{"type":50,"value":1102},"model.enc_layers \u002F model.dec_layers",{"type":50,"value":1104},": Keep both at 6 for train\u002FAutoML\nruns. The mask head asserts six decoder outputs during validation, so\ncopying Grounding DINO smoke overrides that reduce transformer layers causes\nan immediate failure.",{"type":44,"tag":375,"props":1106,"children":1107},{},[1108,1113,1115,1121,1123,1129,1131,1137],{"type":44,"tag":379,"props":1109,"children":1110},{},[1111],{"type":50,"value":1112},"AutoML metric note",{"type":50,"value":1114},": Use ",{"type":44,"tag":69,"props":1116,"children":1118},{"className":1117},[],[1119],{"type":50,"value":1120},"metric=\"val_loss\"",{"type":50,"value":1122}," with\n",{"type":44,"tag":69,"props":1124,"children":1126},{"className":1125},[],[1127],{"type":50,"value":1128},"direction=\"minimize\"",{"type":50,"value":1130}," for train-stage AutoML. The packaged train loop logs\nvalidation loss scalars; it does not emit ",{"type":44,"tag":69,"props":1132,"children":1134},{"className":1133},[],[1135],{"type":50,"value":1136},"[bbox] val_mAP@50",{"type":50,"value":1138}," during the\ntrain job.",{"type":44,"tag":375,"props":1140,"children":1141},{},[1142,1147],{"type":44,"tag":379,"props":1143,"children":1144},{},[1145],{"type":50,"value":1146},"model.has_mask",{"type":50,"value":1148},": Enables mask prediction head. Default True. Adds mask\u002Fdice\u002Frela loss coefficients.",{"type":44,"tag":375,"props":1150,"children":1151},{},[1152,1157],{"type":44,"tag":379,"props":1153,"children":1154},{},[1155],{"type":50,"value":1156},"model.num_region_queries",{"type":50,"value":1158},": Number of region queries for mask prediction. Default 100.",{"type":44,"tag":375,"props":1160,"children":1161},{},[1162,1167,1169,1174],{"type":44,"tag":379,"props":1163,"children":1164},{},[1165],{"type":50,"value":1166},"model.loss_types",{"type":50,"value":1168},": Loss components. Default ",{"type":44,"tag":766,"props":1170,"children":1171},{},[1172],{"type":50,"value":1173},"labels, boxes, masks",{"type":50,"value":1175},". Includes mask_loss_coef, dice_loss_coef, rela_loss_coef.",{"type":44,"tag":375,"props":1177,"children":1178},{},[1179,1184],{"type":44,"tag":379,"props":1180,"children":1181},{},[1182],{"type":50,"value":1183},"evaluate.ioi_threshold",{"type":50,"value":1185},": IoI threshold for mask evaluation. Default 0.5.",{"type":44,"tag":375,"props":1187,"children":1188},{},[1189,1194],{"type":44,"tag":379,"props":1190,"children":1191},{},[1192],{"type":50,"value":1193},"evaluate.nms_threshold",{"type":50,"value":1195},": NMS threshold. Default 0.2.",{"type":44,"tag":375,"props":1197,"children":1198},{},[1199,1204],{"type":44,"tag":379,"props":1200,"children":1201},{},[1202],{"type":50,"value":1203},"evaluate.text_threshold",{"type":50,"value":1205},": Text matching threshold. Default 0.3.",{"type":44,"tag":375,"props":1207,"children":1208},{},[1209,1214],{"type":44,"tag":379,"props":1210,"children":1211},{},[1212],{"type":50,"value":1213},"dataset.has_mask",{"type":50,"value":1215},": Dataset includes mask annotations. Default True. val_data_sources default data_type is \"VG\".",{"type":44,"tag":118,"props":1217,"children":1219},{"id":1218},"multi-gpu-multi-node",[1220],{"type":50,"value":1221},"Multi-GPU \u002F Multi-Node",{"type":44,"tag":53,"props":1223,"children":1224},{},[1225,1230],{"type":44,"tag":379,"props":1226,"children":1227},{},[1228],{"type":50,"value":1229},"Launch method:",{"type":50,"value":1231}," Lightning-managed. Same DDP\u002FFSDP behavior as Grounding DINO.",{"type":44,"tag":414,"props":1233,"children":1234},{},[1235,1255],{"type":44,"tag":418,"props":1236,"children":1237},{},[1238],{"type":44,"tag":422,"props":1239,"children":1240},{},[1241,1245,1250],{"type":44,"tag":426,"props":1242,"children":1243},{},[1244],{"type":50,"value":435},{"type":44,"tag":426,"props":1246,"children":1247},{},[1248],{"type":50,"value":1249},"Description",{"type":44,"tag":426,"props":1251,"children":1252},{},[1253],{"type":50,"value":1254},"Default",{"type":44,"tag":452,"props":1256,"children":1257},{},[1258,1280,1305,1326],{"type":44,"tag":422,"props":1259,"children":1260},{},[1261,1270,1275],{"type":44,"tag":459,"props":1262,"children":1263},{},[1264],{"type":44,"tag":69,"props":1265,"children":1267},{"className":1266},[],[1268],{"type":50,"value":1269},"train.num_gpus",{"type":44,"tag":459,"props":1271,"children":1272},{},[1273],{"type":50,"value":1274},"Number of GPUs",{"type":44,"tag":459,"props":1276,"children":1277},{},[1278],{"type":50,"value":1279},"1",{"type":44,"tag":422,"props":1281,"children":1282},{},[1283,1292,1297],{"type":44,"tag":459,"props":1284,"children":1285},{},[1286],{"type":44,"tag":69,"props":1287,"children":1289},{"className":1288},[],[1290],{"type":50,"value":1291},"train.gpu_ids",{"type":44,"tag":459,"props":1293,"children":1294},{},[1295],{"type":50,"value":1296},"GPU device indices",{"type":44,"tag":459,"props":1298,"children":1299},{},[1300],{"type":44,"tag":766,"props":1301,"children":1302},{},[1303],{"type":50,"value":1304},"0",{"type":44,"tag":422,"props":1306,"children":1307},{},[1308,1317,1322],{"type":44,"tag":459,"props":1309,"children":1310},{},[1311],{"type":44,"tag":69,"props":1312,"children":1314},{"className":1313},[],[1315],{"type":50,"value":1316},"train.num_nodes",{"type":44,"tag":459,"props":1318,"children":1319},{},[1320],{"type":50,"value":1321},"Number of nodes",{"type":44,"tag":459,"props":1323,"children":1324},{},[1325],{"type":50,"value":1279},{"type":44,"tag":422,"props":1327,"children":1328},{},[1329,1338,1355],{"type":44,"tag":459,"props":1330,"children":1331},{},[1332],{"type":44,"tag":69,"props":1333,"children":1335},{"className":1334},[],[1336],{"type":50,"value":1337},"train.distributed_strategy",{"type":44,"tag":459,"props":1339,"children":1340},{},[1341,1347,1349],{"type":44,"tag":69,"props":1342,"children":1344},{"className":1343},[],[1345],{"type":50,"value":1346},"ddp",{"type":50,"value":1348}," or ",{"type":44,"tag":69,"props":1350,"children":1352},{"className":1351},[],[1353],{"type":50,"value":1354},"fsdp",{"type":44,"tag":459,"props":1356,"children":1357},{},[1358],{"type":44,"tag":69,"props":1359,"children":1361},{"className":1360},[],[1362],{"type":50,"value":1346},{"type":44,"tag":118,"props":1364,"children":1366},{"id":1365},"hardware",[1367],{"type":50,"value":1368},"Hardware",{"type":44,"tag":53,"props":1370,"children":1371},{},[1372],{"type":50,"value":1373},"Minimum 1 GPU(s), recommended 4 GPU(s). 24GB+ (A100 recommended) VRAM per GPU. Heavier than Grounding DINO due to mask prediction head. 24GB+ GPU memory recommended.",{"type":44,"tag":118,"props":1375,"children":1377},{"id":1376},"error-patterns",[1378],{"type":50,"value":1379},"Error Patterns",{"type":44,"tag":53,"props":1381,"children":1382},{},[1383,1388],{"type":44,"tag":379,"props":1384,"children":1385},{},[1386],{"type":50,"value":1387},"CUDA out of memory",{"type":50,"value":1389},": Reduce batch_size. Mask prediction adds overhead on top of Grounding DINO.",{"type":44,"tag":53,"props":1391,"children":1392},{},[1393,1404,1406,1411,1412,1418,1420,1425],{"type":44,"tag":379,"props":1394,"children":1395},{},[1396,1398],{"type":50,"value":1397},"Deploy schema error for ",{"type":44,"tag":69,"props":1399,"children":1401},{"className":1400},[],[1402],{"type":50,"value":1403},"test_threshold",{"type":50,"value":1405},": TAO Deploy uses\n",{"type":44,"tag":69,"props":1407,"children":1409},{"className":1408},[],[1410],{"type":50,"value":1203},{"type":50,"value":184},{"type":44,"tag":69,"props":1413,"children":1415},{"className":1414},[],[1416],{"type":50,"value":1417},"inference.text_threshold",{"type":50,"value":1419},". Do not use\n",{"type":44,"tag":69,"props":1421,"children":1423},{"className":1422},[],[1424],{"type":50,"value":1403},{"type":50,"value":1426}," in deploy specs.",{"type":44,"tag":53,"props":1428,"children":1429},{},[1430,1435,1437,1442,1444,1450,1451,1457,1458,1463,1465,1470],{"type":44,"tag":379,"props":1431,"children":1432},{},[1433],{"type":50,"value":1434},"Deploy model shape mismatch",{"type":50,"value":1436},": Carry transformer and mask structure fields\nfrom export into deploy evaluate\u002Finference specs, including ",{"type":44,"tag":69,"props":1438,"children":1440},{"className":1439},[],[1441],{"type":50,"value":1092},{"type":50,"value":1443},",\n",{"type":44,"tag":69,"props":1445,"children":1447},{"className":1446},[],[1448],{"type":50,"value":1449},"model.num_select",{"type":50,"value":200},{"type":44,"tag":69,"props":1452,"children":1454},{"className":1453},[],[1455],{"type":50,"value":1456},"model.max_text_len",{"type":50,"value":200},{"type":44,"tag":69,"props":1459,"children":1461},{"className":1460},[],[1462],{"type":50,"value":1156},{"type":50,"value":1464},", and\n",{"type":44,"tag":69,"props":1466,"children":1468},{"className":1467},[],[1469],{"type":50,"value":1146},{"type":50,"value":1471},". These values must match the ONNX model used to build the\nTensorRT engine.",{"type":44,"tag":118,"props":1473,"children":1475},{"id":1474},"spec-param-parent-model-inference",[1476],{"type":50,"value":1477},"Spec Param \u002F Parent Model Inference",{"type":44,"tag":53,"props":1479,"children":1480},{},[1481,1483,1489,1491,1497,1499,1505],{"type":50,"value":1482},"Model-specific inference mappings belong in this MD file, not in ",{"type":44,"tag":69,"props":1484,"children":1486},{"className":1485},[],[1487],{"type":50,"value":1488},"config.json",{"type":50,"value":1490},". Generated runners should read this section and apply the mappings with SDK helpers before ",{"type":44,"tag":69,"props":1492,"children":1494},{"className":1493},[],[1495],{"type":50,"value":1496},"create_job()",{"type":50,"value":1498},". This mirrors the old microservices ",{"type":44,"tag":69,"props":1500,"children":1502},{"className":1501},[],[1503],{"type":50,"value":1504},"infer_params.py",{"type":50,"value":1506}," flow.",{"type":44,"tag":53,"props":1508,"children":1509},{},[1510,1512,1518],{"type":50,"value":1511},"Inference mappings from TAO Core ",{"type":44,"tag":69,"props":1513,"children":1515},{"className":1514},[],[1516],{"type":50,"value":1517},"mask_grounding_dino.config.json",{"type":50,"value":1519},":",{"type":44,"tag":414,"props":1521,"children":1522},{},[1523,1548],{"type":44,"tag":418,"props":1524,"children":1525},{},[1526],{"type":44,"tag":422,"props":1527,"children":1528},{},[1529,1533,1538,1543],{"type":44,"tag":426,"props":1530,"children":1531},{},[1532],{"type":50,"value":430},{"type":44,"tag":426,"props":1534,"children":1535},{},[1536],{"type":50,"value":1537},"Spec Field",{"type":44,"tag":426,"props":1539,"children":1540},{},[1541],{"type":50,"value":1542},"Inference Function",{"type":44,"tag":426,"props":1544,"children":1545},{},[1546],{"type":50,"value":1547},"Meaning",{"type":44,"tag":452,"props":1549,"children":1550},{},[1551,1581,1611,1639,1669,1696,1724,1754,1781,1808,1836,1866,1893,1920,1948,1976,2003,2030,2058,2085,2112,2142,2169,2197],{"type":44,"tag":422,"props":1552,"children":1553},{},[1554,1558,1567,1576],{"type":44,"tag":459,"props":1555,"children":1556},{},[1557],{"type":50,"value":82},{"type":44,"tag":459,"props":1559,"children":1560},{},[1561],{"type":44,"tag":69,"props":1562,"children":1564},{"className":1563},[],[1565],{"type":50,"value":1566},"encryption_key",{"type":44,"tag":459,"props":1568,"children":1569},{},[1570],{"type":44,"tag":69,"props":1571,"children":1573},{"className":1572},[],[1574],{"type":50,"value":1575},"key",{"type":44,"tag":459,"props":1577,"children":1578},{},[1579],{"type":50,"value":1580},"encryption key",{"type":44,"tag":422,"props":1582,"children":1583},{},[1584,1588,1597,1606],{"type":44,"tag":459,"props":1585,"children":1586},{},[1587],{"type":50,"value":82},{"type":44,"tag":459,"props":1589,"children":1590},{},[1591],{"type":44,"tag":69,"props":1592,"children":1594},{"className":1593},[],[1595],{"type":50,"value":1596},"evaluate.checkpoint",{"type":44,"tag":459,"props":1598,"children":1599},{},[1600],{"type":44,"tag":69,"props":1601,"children":1603},{"className":1602},[],[1604],{"type":50,"value":1605},"parent_model",{"type":44,"tag":459,"props":1607,"children":1608},{},[1609],{"type":50,"value":1610},"model file inferred from the parent job results folder",{"type":44,"tag":422,"props":1612,"children":1613},{},[1614,1618,1627,1635],{"type":44,"tag":459,"props":1615,"children":1616},{},[1617],{"type":50,"value":82},{"type":44,"tag":459,"props":1619,"children":1620},{},[1621],{"type":44,"tag":69,"props":1622,"children":1624},{"className":1623},[],[1625],{"type":50,"value":1626},"evaluate.trt_engine",{"type":44,"tag":459,"props":1628,"children":1629},{},[1630],{"type":44,"tag":69,"props":1631,"children":1633},{"className":1632},[],[1634],{"type":50,"value":1605},{"type":44,"tag":459,"props":1636,"children":1637},{},[1638],{"type":50,"value":1610},{"type":44,"tag":422,"props":1640,"children":1641},{},[1642,1646,1655,1664],{"type":44,"tag":459,"props":1643,"children":1644},{},[1645],{"type":50,"value":82},{"type":44,"tag":459,"props":1647,"children":1648},{},[1649],{"type":44,"tag":69,"props":1650,"children":1652},{"className":1651},[],[1653],{"type":50,"value":1654},"results_dir",{"type":44,"tag":459,"props":1656,"children":1657},{},[1658],{"type":44,"tag":69,"props":1659,"children":1661},{"className":1660},[],[1662],{"type":50,"value":1663},"output_dir",{"type":44,"tag":459,"props":1665,"children":1666},{},[1667],{"type":50,"value":1668},"current job results directory",{"type":44,"tag":422,"props":1670,"children":1671},{},[1672,1676,1684,1692],{"type":44,"tag":459,"props":1673,"children":1674},{},[1675],{"type":50,"value":354},{"type":44,"tag":459,"props":1677,"children":1678},{},[1679],{"type":44,"tag":69,"props":1680,"children":1682},{"className":1681},[],[1683],{"type":50,"value":1566},{"type":44,"tag":459,"props":1685,"children":1686},{},[1687],{"type":44,"tag":69,"props":1688,"children":1690},{"className":1689},[],[1691],{"type":50,"value":1575},{"type":44,"tag":459,"props":1693,"children":1694},{},[1695],{"type":50,"value":1580},{"type":44,"tag":422,"props":1697,"children":1698},{},[1699,1703,1712,1720],{"type":44,"tag":459,"props":1700,"children":1701},{},[1702],{"type":50,"value":354},{"type":44,"tag":459,"props":1704,"children":1705},{},[1706],{"type":44,"tag":69,"props":1707,"children":1709},{"className":1708},[],[1710],{"type":50,"value":1711},"export.checkpoint",{"type":44,"tag":459,"props":1713,"children":1714},{},[1715],{"type":44,"tag":69,"props":1716,"children":1718},{"className":1717},[],[1719],{"type":50,"value":1605},{"type":44,"tag":459,"props":1721,"children":1722},{},[1723],{"type":50,"value":1610},{"type":44,"tag":422,"props":1725,"children":1726},{},[1727,1731,1740,1749],{"type":44,"tag":459,"props":1728,"children":1729},{},[1730],{"type":50,"value":354},{"type":44,"tag":459,"props":1732,"children":1733},{},[1734],{"type":44,"tag":69,"props":1735,"children":1737},{"className":1736},[],[1738],{"type":50,"value":1739},"export.onnx_file",{"type":44,"tag":459,"props":1741,"children":1742},{},[1743],{"type":44,"tag":69,"props":1744,"children":1746},{"className":1745},[],[1747],{"type":50,"value":1748},"create_onnx_file",{"type":44,"tag":459,"props":1750,"children":1751},{},[1752],{"type":50,"value":1753},"output ONNX path",{"type":44,"tag":422,"props":1755,"children":1756},{},[1757,1761,1769,1777],{"type":44,"tag":459,"props":1758,"children":1759},{},[1760],{"type":50,"value":354},{"type":44,"tag":459,"props":1762,"children":1763},{},[1764],{"type":44,"tag":69,"props":1765,"children":1767},{"className":1766},[],[1768],{"type":50,"value":1654},{"type":44,"tag":459,"props":1770,"children":1771},{},[1772],{"type":44,"tag":69,"props":1773,"children":1775},{"className":1774},[],[1776],{"type":50,"value":1663},{"type":44,"tag":459,"props":1778,"children":1779},{},[1780],{"type":50,"value":1668},{"type":44,"tag":422,"props":1782,"children":1783},{},[1784,1788,1796,1804],{"type":44,"tag":459,"props":1785,"children":1786},{},[1787],{"type":50,"value":74},{"type":44,"tag":459,"props":1789,"children":1790},{},[1791],{"type":44,"tag":69,"props":1792,"children":1794},{"className":1793},[],[1795],{"type":50,"value":1566},{"type":44,"tag":459,"props":1797,"children":1798},{},[1799],{"type":44,"tag":69,"props":1800,"children":1802},{"className":1801},[],[1803],{"type":50,"value":1575},{"type":44,"tag":459,"props":1805,"children":1806},{},[1807],{"type":50,"value":1580},{"type":44,"tag":422,"props":1809,"children":1810},{},[1811,1815,1824,1832],{"type":44,"tag":459,"props":1812,"children":1813},{},[1814],{"type":50,"value":74},{"type":44,"tag":459,"props":1816,"children":1817},{},[1818],{"type":44,"tag":69,"props":1819,"children":1821},{"className":1820},[],[1822],{"type":50,"value":1823},"gen_trt_engine.onnx_file",{"type":44,"tag":459,"props":1825,"children":1826},{},[1827],{"type":44,"tag":69,"props":1828,"children":1830},{"className":1829},[],[1831],{"type":50,"value":1605},{"type":44,"tag":459,"props":1833,"children":1834},{},[1835],{"type":50,"value":1610},{"type":44,"tag":422,"props":1837,"children":1838},{},[1839,1843,1852,1861],{"type":44,"tag":459,"props":1840,"children":1841},{},[1842],{"type":50,"value":74},{"type":44,"tag":459,"props":1844,"children":1845},{},[1846],{"type":44,"tag":69,"props":1847,"children":1849},{"className":1848},[],[1850],{"type":50,"value":1851},"gen_trt_engine.trt_engine",{"type":44,"tag":459,"props":1853,"children":1854},{},[1855],{"type":44,"tag":69,"props":1856,"children":1858},{"className":1857},[],[1859],{"type":50,"value":1860},"create_engine_file",{"type":44,"tag":459,"props":1862,"children":1863},{},[1864],{"type":50,"value":1865},"output TensorRT engine path",{"type":44,"tag":422,"props":1867,"children":1868},{},[1869,1873,1881,1889],{"type":44,"tag":459,"props":1870,"children":1871},{},[1872],{"type":50,"value":74},{"type":44,"tag":459,"props":1874,"children":1875},{},[1876],{"type":44,"tag":69,"props":1877,"children":1879},{"className":1878},[],[1880],{"type":50,"value":1654},{"type":44,"tag":459,"props":1882,"children":1883},{},[1884],{"type":44,"tag":69,"props":1885,"children":1887},{"className":1886},[],[1888],{"type":50,"value":1663},{"type":44,"tag":459,"props":1890,"children":1891},{},[1892],{"type":50,"value":1668},{"type":44,"tag":422,"props":1894,"children":1895},{},[1896,1900,1908,1916],{"type":44,"tag":459,"props":1897,"children":1898},{},[1899],{"type":50,"value":90},{"type":44,"tag":459,"props":1901,"children":1902},{},[1903],{"type":44,"tag":69,"props":1904,"children":1906},{"className":1905},[],[1907],{"type":50,"value":1566},{"type":44,"tag":459,"props":1909,"children":1910},{},[1911],{"type":44,"tag":69,"props":1912,"children":1914},{"className":1913},[],[1915],{"type":50,"value":1575},{"type":44,"tag":459,"props":1917,"children":1918},{},[1919],{"type":50,"value":1580},{"type":44,"tag":422,"props":1921,"children":1922},{},[1923,1927,1936,1944],{"type":44,"tag":459,"props":1924,"children":1925},{},[1926],{"type":50,"value":90},{"type":44,"tag":459,"props":1928,"children":1929},{},[1930],{"type":44,"tag":69,"props":1931,"children":1933},{"className":1932},[],[1934],{"type":50,"value":1935},"inference.checkpoint",{"type":44,"tag":459,"props":1937,"children":1938},{},[1939],{"type":44,"tag":69,"props":1940,"children":1942},{"className":1941},[],[1943],{"type":50,"value":1605},{"type":44,"tag":459,"props":1945,"children":1946},{},[1947],{"type":50,"value":1610},{"type":44,"tag":422,"props":1949,"children":1950},{},[1951,1955,1964,1972],{"type":44,"tag":459,"props":1952,"children":1953},{},[1954],{"type":50,"value":90},{"type":44,"tag":459,"props":1956,"children":1957},{},[1958],{"type":44,"tag":69,"props":1959,"children":1961},{"className":1960},[],[1962],{"type":50,"value":1963},"inference.trt_engine",{"type":44,"tag":459,"props":1965,"children":1966},{},[1967],{"type":44,"tag":69,"props":1968,"children":1970},{"className":1969},[],[1971],{"type":50,"value":1605},{"type":44,"tag":459,"props":1973,"children":1974},{},[1975],{"type":50,"value":1610},{"type":44,"tag":422,"props":1977,"children":1978},{},[1979,1983,1991,1999],{"type":44,"tag":459,"props":1980,"children":1981},{},[1982],{"type":50,"value":90},{"type":44,"tag":459,"props":1984,"children":1985},{},[1986],{"type":44,"tag":69,"props":1987,"children":1989},{"className":1988},[],[1990],{"type":50,"value":1654},{"type":44,"tag":459,"props":1992,"children":1993},{},[1994],{"type":44,"tag":69,"props":1995,"children":1997},{"className":1996},[],[1998],{"type":50,"value":1663},{"type":44,"tag":459,"props":2000,"children":2001},{},[2002],{"type":50,"value":1668},{"type":44,"tag":422,"props":2004,"children":2005},{},[2006,2010,2018,2026],{"type":44,"tag":459,"props":2007,"children":2008},{},[2009],{"type":50,"value":565},{"type":44,"tag":459,"props":2011,"children":2012},{},[2013],{"type":44,"tag":69,"props":2014,"children":2016},{"className":2015},[],[2017],{"type":50,"value":1566},{"type":44,"tag":459,"props":2019,"children":2020},{},[2021],{"type":44,"tag":69,"props":2022,"children":2024},{"className":2023},[],[2025],{"type":50,"value":1575},{"type":44,"tag":459,"props":2027,"children":2028},{},[2029],{"type":50,"value":1580},{"type":44,"tag":422,"props":2031,"children":2032},{},[2033,2037,2046,2054],{"type":44,"tag":459,"props":2034,"children":2035},{},[2036],{"type":50,"value":565},{"type":44,"tag":459,"props":2038,"children":2039},{},[2040],{"type":44,"tag":69,"props":2041,"children":2043},{"className":2042},[],[2044],{"type":50,"value":2045},"quantize.model_path",{"type":44,"tag":459,"props":2047,"children":2048},{},[2049],{"type":44,"tag":69,"props":2050,"children":2052},{"className":2051},[],[2053],{"type":50,"value":1605},{"type":44,"tag":459,"props":2055,"children":2056},{},[2057],{"type":50,"value":1610},{"type":44,"tag":422,"props":2059,"children":2060},{},[2061,2065,2073,2081],{"type":44,"tag":459,"props":2062,"children":2063},{},[2064],{"type":50,"value":565},{"type":44,"tag":459,"props":2066,"children":2067},{},[2068],{"type":44,"tag":69,"props":2069,"children":2071},{"className":2070},[],[2072],{"type":50,"value":1654},{"type":44,"tag":459,"props":2074,"children":2075},{},[2076],{"type":44,"tag":69,"props":2077,"children":2079},{"className":2078},[],[2080],{"type":50,"value":1663},{"type":44,"tag":459,"props":2082,"children":2083},{},[2084],{"type":50,"value":1668},{"type":44,"tag":422,"props":2086,"children":2087},{},[2088,2092,2100,2108],{"type":44,"tag":459,"props":2089,"children":2090},{},[2091],{"type":50,"value":665},{"type":44,"tag":459,"props":2093,"children":2094},{},[2095],{"type":44,"tag":69,"props":2096,"children":2098},{"className":2097},[],[2099],{"type":50,"value":1566},{"type":44,"tag":459,"props":2101,"children":2102},{},[2103],{"type":44,"tag":69,"props":2104,"children":2106},{"className":2105},[],[2107],{"type":50,"value":1575},{"type":44,"tag":459,"props":2109,"children":2110},{},[2111],{"type":50,"value":1580},{"type":44,"tag":422,"props":2113,"children":2114},{},[2115,2119,2128,2137],{"type":44,"tag":459,"props":2116,"children":2117},{},[2118],{"type":50,"value":665},{"type":44,"tag":459,"props":2120,"children":2121},{},[2122],{"type":44,"tag":69,"props":2123,"children":2125},{"className":2124},[],[2126],{"type":50,"value":2127},"model.pretrained_backbone_path",{"type":44,"tag":459,"props":2129,"children":2130},{},[2131],{"type":44,"tag":69,"props":2132,"children":2134},{"className":2133},[],[2135],{"type":50,"value":2136},"ptm_if_no_resume_model",{"type":44,"tag":459,"props":2138,"children":2139},{},[2140],{"type":50,"value":2141},"PTM when no resume checkpoint exists",{"type":44,"tag":422,"props":2143,"children":2144},{},[2145,2149,2157,2165],{"type":44,"tag":459,"props":2146,"children":2147},{},[2148],{"type":50,"value":665},{"type":44,"tag":459,"props":2150,"children":2151},{},[2152],{"type":44,"tag":69,"props":2153,"children":2155},{"className":2154},[],[2156],{"type":50,"value":1654},{"type":44,"tag":459,"props":2158,"children":2159},{},[2160],{"type":44,"tag":69,"props":2161,"children":2163},{"className":2162},[],[2164],{"type":50,"value":1663},{"type":44,"tag":459,"props":2166,"children":2167},{},[2168],{"type":50,"value":1668},{"type":44,"tag":422,"props":2170,"children":2171},{},[2172,2176,2185,2193],{"type":44,"tag":459,"props":2173,"children":2174},{},[2175],{"type":50,"value":665},{"type":44,"tag":459,"props":2177,"children":2178},{},[2179],{"type":44,"tag":69,"props":2180,"children":2182},{"className":2181},[],[2183],{"type":50,"value":2184},"train.pretrained_model_path",{"type":44,"tag":459,"props":2186,"children":2187},{},[2188],{"type":44,"tag":69,"props":2189,"children":2191},{"className":2190},[],[2192],{"type":50,"value":2136},{"type":44,"tag":459,"props":2194,"children":2195},{},[2196],{"type":50,"value":2141},{"type":44,"tag":422,"props":2198,"children":2199},{},[2200,2204,2213,2222],{"type":44,"tag":459,"props":2201,"children":2202},{},[2203],{"type":50,"value":665},{"type":44,"tag":459,"props":2205,"children":2206},{},[2207],{"type":44,"tag":69,"props":2208,"children":2210},{"className":2209},[],[2211],{"type":50,"value":2212},"train.resume_training_checkpoint_path",{"type":44,"tag":459,"props":2214,"children":2215},{},[2216],{"type":44,"tag":69,"props":2217,"children":2219},{"className":2218},[],[2220],{"type":50,"value":2221},"resume_model",{"type":44,"tag":459,"props":2223,"children":2224},{},[2225],{"type":50,"value":2226},"model file inferred from the current job results folder",{"type":44,"tag":53,"props":2228,"children":2229},{},[2230,2232,2237,2238,2244,2246,2252,2254,2259],{"type":50,"value":2231},"For ",{"type":44,"tag":69,"props":2233,"children":2235},{"className":2234},[],[2236],{"type":50,"value":1605},{"type":50,"value":1348},{"type":44,"tag":69,"props":2239,"children":2241},{"className":2240},[],[2242],{"type":50,"value":2243},"parent_model_folder",{"type":50,"value":2245},", pass the upstream train\u002Fexport\u002FAutoML child job id as ",{"type":44,"tag":69,"props":2247,"children":2249},{"className":2248},[],[2250],{"type":50,"value":2251},"parent_job_id",{"type":50,"value":2253},". 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":44,"tag":69,"props":2255,"children":2257},{"className":2256},[],[2258],{"type":50,"value":1488},{"type":50,"value":2260}," and do not patch generated runner scripts to guess checkpoint paths.",{"type":44,"tag":53,"props":2262,"children":2263},{},[2264,2266,2272,2274,2280,2282,2288,2290,2295,2296,2301,2302,2307,2308,2313,2315,2320,2322,2327],{"type":50,"value":2265},"When selecting a Mask Grounding DINO checkpoint outside the SDK resolver, match\nthe intended epoch\u002Fstep artifact exactly, for example\n",{"type":44,"tag":69,"props":2267,"children":2269},{"className":2268},[],[2270],{"type":50,"value":2271},"model_epoch_000_step_00049.pth",{"type":50,"value":2273},". The ",{"type":44,"tag":69,"props":2275,"children":2277},{"className":2276},[],[2278],{"type":50,"value":2279},"mask_gdino_model_latest.pth",{"type":50,"value":2281}," symlink is\nvalid only when latest is explicitly requested. The parent PyTorch\n",{"type":44,"tag":69,"props":2283,"children":2285},{"className":2284},[],[2286],{"type":50,"value":2287},"mask_grounding_dino",{"type":50,"value":2289}," CLI supports ",{"type":44,"tag":69,"props":2291,"children":2293},{"className":2292},[],[2294],{"type":50,"value":665},{"type":50,"value":200},{"type":44,"tag":69,"props":2297,"children":2299},{"className":2298},[],[2300],{"type":50,"value":82},{"type":50,"value":200},{"type":44,"tag":69,"props":2303,"children":2305},{"className":2304},[],[2306],{"type":50,"value":90},{"type":50,"value":200},{"type":44,"tag":69,"props":2309,"children":2311},{"className":2310},[],[2312],{"type":50,"value":354},{"type":50,"value":2314},",\nand ",{"type":44,"tag":69,"props":2316,"children":2318},{"className":2317},[],[2319],{"type":50,"value":565},{"type":50,"value":2321},"; run TensorRT engine generation, TensorRT inference, and\nTensorRT evaluation through ",{"type":44,"tag":69,"props":2323,"children":2325},{"className":2324},[],[2326],{"type":50,"value":98},{"type":50,"value":753},{"type":44,"tag":118,"props":2329,"children":2331},{"id":2330},"deployment",[2332],{"type":50,"value":2333},"Deployment",{"type":44,"tag":371,"props":2335,"children":2336},{},[2337],{"type":44,"tag":375,"props":2338,"children":2339},{},[2340],{"type":44,"tag":2341,"props":2342,"children":2343},"a",{"href":98},[2344],{"type":50,"value":2345},"tao-deploy-mask-grounding-dino",{"type":44,"tag":2347,"props":2348,"children":2349},"style",{},[2350],{"type":50,"value":2351},"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":2353,"total":2455},[2354,2371,2383,2397,2409,2426,2441],{"slug":2355,"name":2355,"fn":2356,"description":2357,"org":2358,"tags":2359,"stars":20,"repoUrl":21,"updatedAt":2370},"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},[2360,2363,2366,2367],{"name":2361,"slug":2362,"type":15},"Data Analysis","data-analysis",{"name":2364,"slug":2365,"type":15},"Data Engineering","data-engineering",{"name":9,"slug":8,"type":15},{"name":2368,"slug":2369,"type":15},"Performance","performance","2026-07-14T05:28:43.176466",{"slug":2372,"name":2372,"fn":2373,"description":2374,"org":2375,"tags":2376,"stars":20,"repoUrl":21,"updatedAt":2382},"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},[2377,2378,2381],{"name":2333,"slug":2330,"type":15},{"name":2379,"slug":2380,"type":15},"Infrastructure","infrastructure",{"name":9,"slug":8,"type":15},"2026-07-14T05:29:06.667109",{"slug":2384,"name":2384,"fn":2385,"description":2386,"org":2387,"tags":2388,"stars":20,"repoUrl":21,"updatedAt":2396},"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},[2389,2392,2393],{"name":2390,"slug":2391,"type":15},"Agents","agents",{"name":9,"slug":8,"type":15},{"name":2394,"slug":2395,"type":15},"Research","research","2026-07-14T05:28:06.816956",{"slug":2398,"name":2398,"fn":2399,"description":2400,"org":2401,"tags":2402,"stars":20,"repoUrl":21,"updatedAt":2408},"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},[2403,2404,2405],{"name":2361,"slug":2362,"type":15},{"name":9,"slug":8,"type":15},{"name":2406,"slug":2407,"type":15},"Testing","testing","2026-07-17T05:29:03.913266",{"slug":2410,"name":2410,"fn":2411,"description":2412,"org":2413,"tags":2414,"stars":20,"repoUrl":21,"updatedAt":2425},"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},[2415,2418,2421,2422],{"name":2416,"slug":2417,"type":15},"Automation","automation",{"name":2419,"slug":2420,"type":15},"Imaging","imaging",{"name":9,"slug":8,"type":15},{"name":2423,"slug":2424,"type":15},"Video","video","2026-07-17T05:28:53.905004",{"slug":2427,"name":2427,"fn":2428,"description":2429,"org":2430,"tags":2431,"stars":20,"repoUrl":21,"updatedAt":2440},"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},[2432,2433,2436,2437],{"name":2333,"slug":2330,"type":15},{"name":2434,"slug":2435,"type":15},"Docker","docker",{"name":9,"slug":8,"type":15},{"name":2438,"slug":2439,"type":15},"Operations","operations","2026-07-17T05:28:56.913999",{"slug":2442,"name":2442,"fn":2443,"description":2444,"org":2445,"tags":2446,"stars":20,"repoUrl":21,"updatedAt":2454},"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},[2447,2448,2451],{"name":9,"slug":8,"type":15},{"name":2449,"slug":2450,"type":15},"Quantum Computing","quantum-computing",{"name":2452,"slug":2453,"type":15},"Simulation","simulation","2026-07-14T05:26:58.898253",305,{"items":2457,"total":2607},[2458,2476,2491,2502,2514,2528,2541,2555,2566,2575,2589,2598],{"slug":2459,"name":2459,"fn":2460,"description":2461,"org":2462,"tags":2463,"stars":2473,"repoUrl":2474,"updatedAt":2475},"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},[2464,2467,2470],{"name":2465,"slug":2466,"type":15},"Documentation","documentation",{"name":2468,"slug":2469,"type":15},"MCP","mcp",{"name":2471,"slug":2472,"type":15},"Search","search",21777,"https:\u002F\u002Fgithub.com\u002FNVIDIA\u002FNemoClaw","2026-07-20T06:00:01.461044",{"slug":2477,"name":2477,"fn":2478,"description":2479,"org":2480,"tags":2481,"stars":2488,"repoUrl":2489,"updatedAt":2490},"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},[2482,2485,2486],{"name":2483,"slug":2484,"type":15},"Containers","containers",{"name":2333,"slug":2330,"type":15},{"name":2487,"slug":759,"type":15},"Python",17049,"https:\u002F\u002Fgithub.com\u002FNVIDIA\u002FMegatron-LM","2026-07-27T06:06:11.249662",{"slug":2492,"name":2492,"fn":2493,"description":2494,"org":2495,"tags":2496,"stars":2488,"repoUrl":2489,"updatedAt":2501},"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},[2497,2500],{"name":2498,"slug":2499,"type":15},"CI\u002FCD","ci-cd",{"name":2333,"slug":2330,"type":15},"2026-07-14T05:25:59.97109",{"slug":2503,"name":2503,"fn":2504,"description":2505,"org":2506,"tags":2507,"stars":2488,"repoUrl":2489,"updatedAt":2513},"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},[2508,2509,2510],{"name":2498,"slug":2499,"type":15},{"name":2333,"slug":2330,"type":15},{"name":2511,"slug":2512,"type":15},"GitHub","github","2026-07-27T06:06:12.278222",{"slug":2515,"name":2515,"fn":2516,"description":2517,"org":2518,"tags":2519,"stars":2488,"repoUrl":2489,"updatedAt":2527},"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},[2520,2523,2524],{"name":2521,"slug":2522,"type":15},"Debugging","debugging",{"name":2511,"slug":2512,"type":15},{"name":2525,"slug":2526,"type":15},"Triage","triage","2026-07-14T05:25:57.442089",{"slug":2529,"name":2529,"fn":2530,"description":2531,"org":2532,"tags":2533,"stars":2488,"repoUrl":2489,"updatedAt":2540},"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},[2534,2537],{"name":2535,"slug":2536,"type":15},"Best Practices","best-practices",{"name":2538,"slug":2539,"type":15},"Code Analysis","code-analysis","2026-07-14T05:25:56.18433",{"slug":2542,"name":2542,"fn":2543,"description":2544,"org":2545,"tags":2546,"stars":2488,"repoUrl":2489,"updatedAt":2554},"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},[2547,2550,2553],{"name":2548,"slug":2549,"type":15},"Machine Learning","machine-learning",{"name":2551,"slug":2552,"type":15},"Migration","migration",{"name":9,"slug":8,"type":15},"2026-07-17T06:07:11.777011",{"slug":2556,"name":2556,"fn":2557,"description":2558,"org":2559,"tags":2560,"stars":2488,"repoUrl":2489,"updatedAt":2565},"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},[2561,2564],{"name":2562,"slug":2563,"type":15},"QA","qa",{"name":2406,"slug":2407,"type":15},"2026-07-14T05:25:53.673039",{"slug":2567,"name":2567,"fn":2568,"description":2569,"org":2570,"tags":2571,"stars":2488,"repoUrl":2489,"updatedAt":2574},"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},[2572,2573],{"name":2333,"slug":2330,"type":15},{"name":2379,"slug":2380,"type":15},"2026-07-14T05:25:49.362534",{"slug":2576,"name":2576,"fn":2577,"description":2578,"org":2579,"tags":2580,"stars":2488,"repoUrl":2489,"updatedAt":2588},"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},[2581,2584,2585],{"name":2582,"slug":2583,"type":15},"Code Review","code-review",{"name":2511,"slug":2512,"type":15},{"name":2586,"slug":2587,"type":15},"Pull Requests","pull-requests","2026-07-14T05:26:01.226578",{"slug":2590,"name":2590,"fn":2591,"description":2592,"org":2593,"tags":2594,"stars":2488,"repoUrl":2489,"updatedAt":2597},"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},[2595,2596],{"name":2562,"slug":2563,"type":15},{"name":2406,"slug":2407,"type":15},"2026-07-14T05:25:54.928983",{"slug":2599,"name":2599,"fn":2600,"description":2601,"org":2602,"tags":2603,"stars":2488,"repoUrl":2489,"updatedAt":2606},"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},[2604,2605],{"name":2416,"slug":2417,"type":15},{"name":2498,"slug":2499,"type":15},"2026-07-30T05:29:03.275638",496]