[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill-nvidia-tao-train-grounding-dino":3,"mdc--radn3i-key":34,"related-repo-nvidia-tao-train-grounding-dino":2419,"related-org-nvidia-tao-train-grounding-dino":2523},{"slug":4,"name":4,"fn":5,"description":6,"org":7,"tags":11,"stars":23,"repoUrl":24,"updatedAt":25,"license":26,"forks":27,"topics":28,"repo":29,"sourceUrl":32,"mdContent":33},"tao-train-grounding-dino","detect objects with Grounding DINO","Grounding DINO for open-set object detection. Combines DINO-style detection with a BERT text encoder for language-guided detection — detects objects described by text prompts without a fixed class vocabulary. Use when training, evaluating, exporting, quantizing, or running inference for a TAO Grounding DINO model. Trigger phrases include \"train Grounding DINO\", \"open-vocabulary detection\", \"text-prompted detector\", \"language-guided object detection\".",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},"nvidia","NVIDIA","https:\u002F\u002Fpexgzepcugksgbtrxkhf.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Forg-logos\u002Fnvidia.png",[12,16,19,20],{"name":13,"slug":14,"type":15},"Deep Learning","deep-learning","tag",{"name":17,"slug":18,"type":15},"Machine Learning","machine-learning",{"name":9,"slug":8,"type":15},{"name":21,"slug":22,"type":15},"Computer Vision","computer-vision",2473,"https:\u002F\u002Fgithub.com\u002FNVIDIA\u002Fskills","2026-07-14T05:27:26.850145","Apache-2.0",281,[],{"repoUrl":24,"stars":23,"forks":27,"topics":30,"description":31},[],"AI agent skills published by NVIDIA","https:\u002F\u002Fgithub.com\u002FNVIDIA\u002Fskills\u002Ftree\u002FHEAD\u002Fskills\u002Ftao-train-grounding-dino","---\nname: tao-train-grounding-dino\ndescription: Grounding DINO for open-set object detection. Combines DINO-style detection with a BERT text encoder for\n  language-guided detection — detects objects described by text prompts without a fixed class vocabulary. Use when training,\n  evaluating, exporting, quantizing, or running inference for a TAO Grounding DINO model. Trigger phrases include \"train\n  Grounding DINO\", \"open-vocabulary detection\", \"text-prompted detector\", \"language-guided object detection\".\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- object\n- detection\n---\n\n# Grounding DINO\n\nGrounding DINO for open-set object detection. Combines DINO-style detection with BERT text encoder for language-guided detection. Detects objects described by text prompts without fixed class vocabulary.\n\nSet train.pretrained_model_path for full Grounding DINO weights or model.pretrained_backbone_path for backbone-only.\n\nFor TAO Deploy TensorRT actions (`gen_trt_engine`, TensorRT `evaluate`, and TensorRT `inference`), read `references\u002Ftao-deploy-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:** object_detection\n- **Formats:** odvg, coco, raw\n- **Monitoring metric:** val_mAP50\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| inference | dataset.infer_data_sources.image_dir | inference_dataset | images.tar.gz | Yes |\n| inference | dataset.infer_data_sources.captions | workflow prompts | prompt list | Yes |\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.quant_calibration_data_sources | calibration\u002Feval dataset | image_dir: images.tar.gz, json_file: annotations.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\nThe runner may source image archives as `images.tar.gz`, but direct local\nDocker TAO CLI specs must point `image_dir` to an extracted image directory.\nSkill metadata marks these archive-backed image sources with\n`runtime: extracted_folder` so a fresh runner can unpack the archive before\nlaunching TAO.\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_epochs\": 10,\n    \"train.checkpoint_interval\": 10,\n    \"train.validation_interval\": 10,\n    \"train.num_gpus\": 1,\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**deploy\u002Fgen_trt_engine (use `references\u002Ftao-deploy-grounding-dino.md`):**\n```python\n{\n    \"gen_trt_engine.onnx_file\": \"\u003Cexported_onnx_uri>\",\n    \"gen_trt_engine.trt_engine\": \"\u003Coutput_engine_path>\",\n    \"gen_trt_engine.tensorrt.data_type\": \"FP16\",\n}\n```\n\n**inference (mandatory data sources):**\n```python\n{\n    \"inference.checkpoint\": \"\u003Cselected train\u002FAutoML checkpoint>\",\n    \"dataset.infer_data_sources.image_dir\": [f\"{S3_EVAL}\u002Fimages.tar.gz\"],\n    \"dataset.infer_data_sources.captions\": [\n        \"fire extinguisher\",\n        \"cone\",\n        \"cart\",\n        \"forklift\"\n    ],\n}\n```\n\n**evaluate (mandatory data sources):**\n```python\n{\n    \"evaluate.checkpoint\": \"\u003Cselected train\u002FAutoML checkpoint>\",\n    \"dataset.test_data_sources\": {\"image_dir\": f\"{S3_EVAL}\u002Fimages.tar.gz\", \"json_file\": f\"{S3_EVAL}\u002Fannotations.json\"},\n}\n```\n\n**quantize (mandatory data sources):**\n```python\n{\n    \"quantize.model_path\": \"\u003Cselected train checkpoint or exported ONNX model>\",\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_EVAL}\u002Fimages.tar.gz\", \"json_file\": f\"{S3_EVAL}\u002Fannotations.json\"},\n}\n```\n## Eval Dataset\n\nOptional. Validation uses COCO-format annotations for mAP even though training can use ODVG format.\n\n## Important Parameters\n\n- **model.backbone**: Default swin_tiny_224_1k. Also supports resnet_50 and other Swin variants. Swin generally performs better for grounding tasks.\n- **model.text_encoder_type**: BERT model for text encoding. Default bert-base-uncased. max_text_len defaults to 256.\n- **model.max_text_len**: Keep this aligned with the dataset label\u002Ftoken\n  position maps. Do not shrink it for smoke tests unless the corresponding\n  label maps are regenerated with the same length; otherwise validation can\n  fail with a matrix shape mismatch between token probabilities and position\n  maps.\n- **train.optim.lr**: Learning rate. Default 2e-4. lr_backbone 2e-5. Supports bf16 precision in addition to fp16\u002Ffp32.\n- **dataset.max_labels**: Maximum labels per image during training. Default 50. Increase for dense annotation datasets.\n- **model.num_queries**: Object queries. Default 900 (higher than DINO's 300) due to open-vocabulary nature.\n- **model.num_queries \u002F model.num_select**: Keep `num_queries` high enough\n  for the number of matched ODVG targets in a batch. Very small smoke values\n  such as 20 can fail during Hungarian target indexing on dense images; use at\n  least 100 for minimal Grounding DINO smoke runs unless the dataset is known\n  to have fewer objects per image.\n- **train.optim.lr_steps**: MultiStep LR schedule. Default [10].\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.num_nodes` | Number of nodes | 1 |\n| `train.distributed_strategy` | `ddp` or `fsdp` | `ddp` |\n\nSame DDP\u002FFSDP behavior as DINO. Multi-node requires `WORLD_SIZE`, `NODE_RANK`, `MASTER_ADDR`, `MASTER_PORT` env vars set by orchestrator.\n\n## Export \u002F TRT Defaults\n\n- Export input: 960x544 (larger than other OD models), opset 17. Keep\n  Grounding-DINO export specs at the template export resolution for smoke tests;\n  reducing export to very small image sizes such as 128x128 can trigger a\n  PyTorch ONNX shape-inference assertion in the contrastive text head during\n  `torch.onnx.export`.\n- The parent PyTorch `grounding_dino` CLI supports `train`, `evaluate`,\n  `inference`, `export`, and `quantize`. Run TensorRT engine generation,\n  TensorRT inference, and TensorRT evaluation through `references\u002Ftao-deploy-grounding-dino.md`.\n- TRT data types: FP32, FP16 only — **INT8 is NOT supported**\n- TRT workspace: 8192 MB (8x larger than other OD models)\n- TRT max_batch_size: 4\n\n## Hardware\n\nMinimum 1 GPU(s), recommended 4 GPU(s). 24GB+ (A100 recommended) VRAM per GPU. Grounding DINO is heavier than standard DINO due to the text encoder (BERT). 24GB+ GPU memory recommended. Reduce batch_size for 16GB GPUs.\n\n## Error Patterns\n\n**CUDA out of memory**: Reduce batch_size (4 -> 2 -> 1). The BERT text encoder adds significant memory overhead on top of the vision backbone.\n\n**Val annotation category IDs**: Validation annotations should have category IDs starting from 0 for correct loss computation. Use annotation format conversion if needed.\n\n**Text encoder loading error**: Ensure the container has access to download bert-base-uncased weights or provide a local path.\n\n**Quantize with a PyTorch checkpoint fails in TAO Toolkit 7.0.0-rc-226**:\nThe container's Grounding-DINO quantize script passes `cap_lists=None` when\nloading a checkpoint, which fails in `post_process.py`. ONNX quantization uses\nthe exported ONNX artifact and COCO calibration data, but the default rc-226\nPyTorch image also lacks the `modelopt.onnx.quantization` module. Treat this as\nan image\u002FSDK blocker, not a checkpoint resolver issue.\n\n**mat1 and mat2 shapes cannot be multiplied in `post_process.py`**: The text\ntoken length and label position maps are inconsistent, commonly because\n`model.max_text_len` was overridden below the default 256 while the dataset\nlabel maps still use 256-length position maps. Restore `model.max_text_len` or\nregenerate the label maps with the same length.\n\n**index is out of bounds for dimension 0 in `criterion.py`**: `model.num_queries`\nis too small for the matched ODVG targets in the current batch. Increase\n`model.num_queries` and keep `model.num_select` compatible with it.\n\n**NotADirectoryError with `images.tar.gz\u002F\u003Cimage>.jpg`**: The direct TAO CLI is\ntrying to traverse an archive path as a directory. Extract the archive and set\nthe relevant `image_dir` field to the extracted image folder; archive-backed\nskill data sources use `runtime: extracted_folder` for this reason.\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 `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| 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 Grounding-DINO checkpoint outside the SDK resolver, match the\nintended epoch\u002Fstep artifact exactly, for example\n`model_epoch_000_step_00046.pth`. The `gdino_model_latest.pth` symlink is valid\nonly when latest is explicitly requested. Carry structural model settings such\nas `model.backbone`, `model.num_queries`, `model.num_select`,\n`model.num_feature_levels`, `model.max_text_len`, and export input resolution\nforward into evaluate, inference, export, and deploy specs so checkpoint and\nengine shapes match.\n\n## Deployment\n\n- [tao-deploy-grounding-dino](references\u002Ftao-deploy-grounding-dino.md)\n",{"data":35,"body":44},{"name":4,"description":6,"license":26,"compatibility":36,"metadata":37,"allowed-tools":40,"tags":41},"Requires docker + nvidia-container-toolkit.",{"version":38,"author":39},"0.1.0","NVIDIA Corporation","Read Bash",[42,43],"object","detection",{"type":45,"children":46},"root",[47,56,62,67,121,128,221,227,336,368,374,410,417,663,691,697,717,748,756,833,848,893,901,988,996,1033,1041,1092,1098,1103,1109,1206,1212,1229,1360,1394,1400,1487,1493,1498,1504,1514,1524,1534,1568,1597,1635,1665,1671,1700,1713,2308,2342,2395,2401,2413],{"type":48,"tag":49,"props":50,"children":52},"element","h1",{"id":51},"grounding-dino",[53],{"type":54,"value":55},"text","Grounding DINO",{"type":48,"tag":57,"props":58,"children":59},"p",{},[60],{"type":54,"value":61},"Grounding DINO for open-set object detection. Combines DINO-style detection with BERT text encoder for language-guided detection. Detects objects described by text prompts without fixed class vocabulary.",{"type":48,"tag":57,"props":63,"children":64},{},[65],{"type":54,"value":66},"Set train.pretrained_model_path for full Grounding DINO weights or model.pretrained_backbone_path for backbone-only.",{"type":48,"tag":57,"props":68,"children":69},{},[70,72,79,81,87,89,95,97,103,105,111,113,119],{"type":54,"value":71},"For TAO Deploy TensorRT actions (",{"type":48,"tag":73,"props":74,"children":76},"code",{"className":75},[],[77],{"type":54,"value":78},"gen_trt_engine",{"type":54,"value":80},", TensorRT ",{"type":48,"tag":73,"props":82,"children":84},{"className":83},[],[85],{"type":54,"value":86},"evaluate",{"type":54,"value":88},", and TensorRT ",{"type":48,"tag":73,"props":90,"children":92},{"className":91},[],[93],{"type":54,"value":94},"inference",{"type":54,"value":96},"), read ",{"type":48,"tag":73,"props":98,"children":100},{"className":99},[],[101],{"type":54,"value":102},"references\u002Ftao-deploy-grounding-dino.md",{"type":54,"value":104}," first. Deploy spec templates live in this skill's ",{"type":48,"tag":73,"props":106,"children":108},{"className":107},[],[109],{"type":54,"value":110},"references\u002F",{"type":54,"value":112}," folder with the ",{"type":48,"tag":73,"props":114,"children":116},{"className":115},[],[117],{"type":54,"value":118},"spec_template_deploy_*.yaml",{"type":54,"value":120}," prefix.",{"type":48,"tag":122,"props":123,"children":125},"h2",{"id":124},"dataclass-schemas",[126],{"type":54,"value":127},"Dataclass Schemas",{"type":48,"tag":57,"props":129,"children":130},{},[131,133,139,141,147,149,155,157,163,165,171,173,179,181,187,189,195,197,203,205,211,213,219],{"type":54,"value":132},"Generated TAO Core schemas are packaged in ",{"type":48,"tag":73,"props":134,"children":136},{"className":135},[],[137],{"type":54,"value":138},"schemas\u002F\u003Caction>.schema.json",{"type":54,"value":140},", with ",{"type":48,"tag":73,"props":142,"children":144},{"className":143},[],[145],{"type":54,"value":146},"schemas\u002Fmanifest.json",{"type":54,"value":148}," listing available actions. Each generated schema also emits ",{"type":48,"tag":73,"props":150,"children":152},{"className":151},[],[153],{"type":54,"value":154},"references\u002Fspec_template_\u003Caction>.yaml",{"type":54,"value":156}," from the schema top-level ",{"type":48,"tag":73,"props":158,"children":160},{"className":159},[],[161],{"type":54,"value":162},"default",{"type":54,"value":164}," field. AutoML enablement is declared at the model layer in ",{"type":48,"tag":73,"props":166,"children":168},{"className":167},[],[169],{"type":54,"value":170},"references\u002Fskill_info.yaml",{"type":54,"value":172}," via ",{"type":48,"tag":73,"props":174,"children":176},{"className":175},[],[177],{"type":54,"value":178},"automl_enabled",{"type":54,"value":180},". Runnable AutoML still requires ",{"type":48,"tag":73,"props":182,"children":184},{"className":183},[],[185],{"type":54,"value":186},"schemas\u002Ftrain.schema.json",{"type":54,"value":188}," and ",{"type":48,"tag":73,"props":190,"children":192},{"className":191},[],[193],{"type":54,"value":194},"references\u002Fspec_template_train.yaml",{"type":54,"value":196}," to exist and parse. Use the packaged train schema for ",{"type":48,"tag":73,"props":198,"children":200},{"className":199},[],[201],{"type":54,"value":202},"automl_default_parameters",{"type":54,"value":204},", ",{"type":48,"tag":73,"props":206,"children":208},{"className":207},[],[209],{"type":54,"value":210},"automl_disabled_parameters",{"type":54,"value":212},", defaults, min\u002Fmax bounds, enums, option weights, math conditions, dependencies, and popular parameters. Do not expect ",{"type":48,"tag":73,"props":214,"children":216},{"className":215},[],[217],{"type":54,"value":218},"~\u002Ftao-core",{"type":54,"value":220}," at runtime; maintainers regenerate schemas\u002Ftemplates before packaging the skill bank.",{"type":48,"tag":122,"props":222,"children":224},{"id":223},"train-action-policy",[225],{"type":54,"value":226},"Train Action Policy",{"type":48,"tag":57,"props":228,"children":229},{},[230,232,237,239,245,247,253,255,261,263,269,271,277,279,284,285,291,293,298,299,304,306,312,314,320,322,327,329,334],{"type":54,"value":231},"This model is AutoML-enabled at the model layer. Before handling any train-stage request, read ",{"type":48,"tag":73,"props":233,"children":235},{"className":234},[],[236],{"type":54,"value":170},{"type":54,"value":238}," and resolve the run override from either an explicit ",{"type":48,"tag":73,"props":240,"children":242},{"className":241},[],[243],{"type":54,"value":244},"automl_policy",{"type":54,"value":246}," value or the user's workflow request. Use ",{"type":48,"tag":73,"props":248,"children":250},{"className":249},[],[251],{"type":54,"value":252},"automl_policy: on",{"type":54,"value":254}," by default and only expose ",{"type":48,"tag":73,"props":256,"children":258},{"className":257},[],[259],{"type":54,"value":260},"on",{"type":54,"value":262}," \u002F ",{"type":48,"tag":73,"props":264,"children":266},{"className":265},[],[267],{"type":54,"value":268},"off",{"type":54,"value":270}," in new launch prompts. Treat phrases like \"turn off AutoML\", \"disable AutoML\", \"no HPO\", or \"plain training\" as ",{"type":48,"tag":73,"props":272,"children":274},{"className":273},[],[275],{"type":54,"value":276},"automl_policy: off",{"type":54,"value":278}," for this run only. When ",{"type":48,"tag":73,"props":280,"children":282},{"className":281},[],[283],{"type":54,"value":252},{"type":54,"value":204},{"type":48,"tag":73,"props":286,"children":288},{"className":287},[],[289],{"type":54,"value":290},"automl_enabled: true",{"type":54,"value":292},", and both ",{"type":48,"tag":73,"props":294,"children":296},{"className":295},[],[297],{"type":54,"value":186},{"type":54,"value":188},{"type":48,"tag":73,"props":300,"children":302},{"className":301},[],[303],{"type":54,"value":194},{"type":54,"value":305}," are packaged, route the train action through ",{"type":48,"tag":73,"props":307,"children":309},{"className":308},[],[310],{"type":54,"value":311},"tao-skill-bank:tao-run-automl",{"type":54,"value":313}," by default with this model's ",{"type":48,"tag":73,"props":315,"children":317},{"className":316},[],[318],{"type":54,"value":319},"skill_dir",{"type":54,"value":321},". Preserve workflow\u002Fapplication overrides for datasets, specs, output directories, GPU\u002Fplatform settings, parent checkpoints, and ",{"type":48,"tag":73,"props":323,"children":325},{"className":324},[],[326],{"type":54,"value":244},{"type":54,"value":328},". Use direct model training only when ",{"type":48,"tag":73,"props":330,"children":332},{"className":331},[],[333],{"type":54,"value":276},{"type":54,"value":335}," or the packaged train schema\u002Ftemplate is missing; in the missing-schema case, report that AutoML is enabled but not runnable for this model until schemas are generated.",{"type":48,"tag":57,"props":337,"children":338},{},[339,341,346,347,352,353,359,361,366],{"type":54,"value":340},"Non-train actions such as ",{"type":48,"tag":73,"props":342,"children":344},{"className":343},[],[345],{"type":54,"value":86},{"type":54,"value":204},{"type":48,"tag":73,"props":348,"children":350},{"className":349},[],[351],{"type":54,"value":94},{"type":54,"value":204},{"type":48,"tag":73,"props":354,"children":356},{"className":355},[],[357],{"type":54,"value":358},"export",{"type":54,"value":360},", and deploy flows stay in this model skill. The per-run ",{"type":48,"tag":73,"props":362,"children":364},{"className":363},[],[365],{"type":54,"value":244},{"type":54,"value":367}," override does not change model metadata.",{"type":48,"tag":122,"props":369,"children":371},{"id":370},"training-requirements",[372],{"type":54,"value":373},"Training Requirements",{"type":48,"tag":375,"props":376,"children":377},"ul",{},[378,390,400],{"type":48,"tag":379,"props":380,"children":381},"li",{},[382,388],{"type":48,"tag":383,"props":384,"children":385},"strong",{},[386],{"type":54,"value":387},"Dataset type:",{"type":54,"value":389}," object_detection",{"type":48,"tag":379,"props":391,"children":392},{},[393,398],{"type":48,"tag":383,"props":394,"children":395},{},[396],{"type":54,"value":397},"Formats:",{"type":54,"value":399}," odvg, coco, raw",{"type":48,"tag":379,"props":401,"children":402},{},[403,408],{"type":48,"tag":383,"props":404,"children":405},{},[406],{"type":54,"value":407},"Monitoring metric:",{"type":54,"value":409}," val_mAP50",{"type":48,"tag":411,"props":412,"children":414},"h3",{"id":413},"per-action-dataset-requirements",[415],{"type":54,"value":416},"Per-Action Dataset Requirements",{"type":48,"tag":418,"props":419,"children":420},"table",{},[421,455],{"type":48,"tag":422,"props":423,"children":424},"thead",{},[425],{"type":48,"tag":426,"props":427,"children":428},"tr",{},[429,435,440,445,450],{"type":48,"tag":430,"props":431,"children":432},"th",{},[433],{"type":54,"value":434},"Action",{"type":48,"tag":430,"props":436,"children":437},{},[438],{"type":54,"value":439},"Spec Key",{"type":48,"tag":430,"props":441,"children":442},{},[443],{"type":54,"value":444},"Source",{"type":48,"tag":430,"props":446,"children":447},{},[448],{"type":54,"value":449},"Files",{"type":48,"tag":430,"props":451,"children":452},{},[453],{"type":54,"value":454},"List?",{"type":48,"tag":456,"props":457,"children":458},"tbody",{},[459,487,514,540,567,591,616,640],{"type":48,"tag":426,"props":460,"children":461},{},[462,467,472,477,482],{"type":48,"tag":463,"props":464,"children":465},"td",{},[466],{"type":54,"value":86},{"type":48,"tag":463,"props":468,"children":469},{},[470],{"type":54,"value":471},"dataset.test_data_sources",{"type":48,"tag":463,"props":473,"children":474},{},[475],{"type":54,"value":476},"eval_dataset",{"type":48,"tag":463,"props":478,"children":479},{},[480],{"type":54,"value":481},"image_dir: images.tar.gz, json_file: annotations.json",{"type":48,"tag":463,"props":483,"children":484},{},[485],{"type":54,"value":486},"No",{"type":48,"tag":426,"props":488,"children":489},{},[490,494,499,504,509],{"type":48,"tag":463,"props":491,"children":492},{},[493],{"type":54,"value":94},{"type":48,"tag":463,"props":495,"children":496},{},[497],{"type":54,"value":498},"dataset.infer_data_sources.image_dir",{"type":48,"tag":463,"props":500,"children":501},{},[502],{"type":54,"value":503},"inference_dataset",{"type":48,"tag":463,"props":505,"children":506},{},[507],{"type":54,"value":508},"images.tar.gz",{"type":48,"tag":463,"props":510,"children":511},{},[512],{"type":54,"value":513},"Yes",{"type":48,"tag":426,"props":515,"children":516},{},[517,521,526,531,536],{"type":48,"tag":463,"props":518,"children":519},{},[520],{"type":54,"value":94},{"type":48,"tag":463,"props":522,"children":523},{},[524],{"type":54,"value":525},"dataset.infer_data_sources.captions",{"type":48,"tag":463,"props":527,"children":528},{},[529],{"type":54,"value":530},"workflow prompts",{"type":48,"tag":463,"props":532,"children":533},{},[534],{"type":54,"value":535},"prompt list",{"type":48,"tag":463,"props":537,"children":538},{},[539],{"type":54,"value":513},{"type":48,"tag":426,"props":541,"children":542},{},[543,548,553,558,563],{"type":48,"tag":463,"props":544,"children":545},{},[546],{"type":54,"value":547},"quantize",{"type":48,"tag":463,"props":549,"children":550},{},[551],{"type":54,"value":552},"dataset.train_data_sources",{"type":48,"tag":463,"props":554,"children":555},{},[556],{"type":54,"value":557},"train_datasets",{"type":48,"tag":463,"props":559,"children":560},{},[561],{"type":54,"value":562},"image_dir: images.tar.gz, json_file: annotations_odvg.jsonl, label_map: annotations_odvg_labelmap.json",{"type":48,"tag":463,"props":564,"children":565},{},[566],{"type":54,"value":513},{"type":48,"tag":426,"props":568,"children":569},{},[570,574,579,583,587],{"type":48,"tag":463,"props":571,"children":572},{},[573],{"type":54,"value":547},{"type":48,"tag":463,"props":575,"children":576},{},[577],{"type":54,"value":578},"dataset.val_data_sources",{"type":48,"tag":463,"props":580,"children":581},{},[582],{"type":54,"value":476},{"type":48,"tag":463,"props":584,"children":585},{},[586],{"type":54,"value":481},{"type":48,"tag":463,"props":588,"children":589},{},[590],{"type":54,"value":486},{"type":48,"tag":426,"props":592,"children":593},{},[594,598,603,608,612],{"type":48,"tag":463,"props":595,"children":596},{},[597],{"type":54,"value":547},{"type":48,"tag":463,"props":599,"children":600},{},[601],{"type":54,"value":602},"dataset.quant_calibration_data_sources",{"type":48,"tag":463,"props":604,"children":605},{},[606],{"type":54,"value":607},"calibration\u002Feval dataset",{"type":48,"tag":463,"props":609,"children":610},{},[611],{"type":54,"value":481},{"type":48,"tag":463,"props":613,"children":614},{},[615],{"type":54,"value":486},{"type":48,"tag":426,"props":617,"children":618},{},[619,624,628,632,636],{"type":48,"tag":463,"props":620,"children":621},{},[622],{"type":54,"value":623},"train",{"type":48,"tag":463,"props":625,"children":626},{},[627],{"type":54,"value":552},{"type":48,"tag":463,"props":629,"children":630},{},[631],{"type":54,"value":557},{"type":48,"tag":463,"props":633,"children":634},{},[635],{"type":54,"value":562},{"type":48,"tag":463,"props":637,"children":638},{},[639],{"type":54,"value":513},{"type":48,"tag":426,"props":641,"children":642},{},[643,647,651,655,659],{"type":48,"tag":463,"props":644,"children":645},{},[646],{"type":54,"value":623},{"type":48,"tag":463,"props":648,"children":649},{},[650],{"type":54,"value":578},{"type":48,"tag":463,"props":652,"children":653},{},[654],{"type":54,"value":476},{"type":48,"tag":463,"props":656,"children":657},{},[658],{"type":54,"value":481},{"type":48,"tag":463,"props":660,"children":661},{},[662],{"type":54,"value":486},{"type":48,"tag":57,"props":664,"children":665},{},[666,668,673,675,681,683,689],{"type":54,"value":667},"The runner may source image archives as ",{"type":48,"tag":73,"props":669,"children":671},{"className":670},[],[672],{"type":54,"value":508},{"type":54,"value":674},", but direct local\nDocker TAO CLI specs must point ",{"type":48,"tag":73,"props":676,"children":678},{"className":677},[],[679],{"type":54,"value":680},"image_dir",{"type":54,"value":682}," to an extracted image directory.\nSkill metadata marks these archive-backed image sources with\n",{"type":48,"tag":73,"props":684,"children":686},{"className":685},[],[687],{"type":54,"value":688},"runtime: extracted_folder",{"type":54,"value":690}," so a fresh runner can unpack the archive before\nlaunching TAO.",{"type":48,"tag":411,"props":692,"children":694},{"id":693},"typical-spec-overrides",[695],{"type":54,"value":696},"Typical Spec Overrides",{"type":48,"tag":57,"props":698,"children":699},{},[700,702,707,709,715],{"type":54,"value":701},"Data source overrides are ",{"type":48,"tag":383,"props":703,"children":704},{},[705],{"type":54,"value":706},"mandatory for every action",{"type":54,"value":708}," — the agent MUST construct data source paths from the Per-Action Dataset Requirements table above and include them in ",{"type":48,"tag":73,"props":710,"children":712},{"className":711},[],[713],{"type":54,"value":714},"spec_overrides",{"type":54,"value":716},".",{"type":48,"tag":718,"props":719,"children":724},"pre",{"className":720,"code":721,"language":722,"meta":723,"style":723},"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","",[725],{"type":48,"tag":73,"props":726,"children":727},{"__ignoreMap":723},[728,739],{"type":48,"tag":729,"props":730,"children":733},"span",{"class":731,"line":732},"line",1,[734],{"type":48,"tag":729,"props":735,"children":736},{},[737],{"type":54,"value":738},"S3_TRAIN = \"s3:\u002F\u002Fbucket\u002Fdata\u002Ftrain\"\n",{"type":48,"tag":729,"props":740,"children":742},{"class":731,"line":741},2,[743],{"type":48,"tag":729,"props":744,"children":745},{},[746],{"type":54,"value":747},"S3_EVAL = \"s3:\u002F\u002Fbucket\u002Fdata\u002Feval\"\n",{"type":48,"tag":57,"props":749,"children":750},{},[751],{"type":48,"tag":383,"props":752,"children":753},{},[754],{"type":54,"value":755},"train (mandatory data sources):",{"type":48,"tag":718,"props":757,"children":759},{"className":720,"code":758,"language":722,"meta":723,"style":723},"{\n    \"train.num_epochs\": 10,\n    \"train.checkpoint_interval\": 10,\n    \"train.validation_interval\": 10,\n    \"train.num_gpus\": 1,\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",[760],{"type":48,"tag":73,"props":761,"children":762},{"__ignoreMap":723},[763,771,779,788,797,806,815,824],{"type":48,"tag":729,"props":764,"children":765},{"class":731,"line":732},[766],{"type":48,"tag":729,"props":767,"children":768},{},[769],{"type":54,"value":770},"{\n",{"type":48,"tag":729,"props":772,"children":773},{"class":731,"line":741},[774],{"type":48,"tag":729,"props":775,"children":776},{},[777],{"type":54,"value":778},"    \"train.num_epochs\": 10,\n",{"type":48,"tag":729,"props":780,"children":782},{"class":731,"line":781},3,[783],{"type":48,"tag":729,"props":784,"children":785},{},[786],{"type":54,"value":787},"    \"train.checkpoint_interval\": 10,\n",{"type":48,"tag":729,"props":789,"children":791},{"class":731,"line":790},4,[792],{"type":48,"tag":729,"props":793,"children":794},{},[795],{"type":54,"value":796},"    \"train.validation_interval\": 10,\n",{"type":48,"tag":729,"props":798,"children":800},{"class":731,"line":799},5,[801],{"type":48,"tag":729,"props":802,"children":803},{},[804],{"type":54,"value":805},"    \"train.num_gpus\": 1,\n",{"type":48,"tag":729,"props":807,"children":809},{"class":731,"line":808},6,[810],{"type":48,"tag":729,"props":811,"children":812},{},[813],{"type":54,"value":814},"    \"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":48,"tag":729,"props":816,"children":818},{"class":731,"line":817},7,[819],{"type":48,"tag":729,"props":820,"children":821},{},[822],{"type":54,"value":823},"    \"dataset.val_data_sources\": {\"image_dir\": f\"{S3_EVAL}\u002Fimages.tar.gz\", \"json_file\": f\"{S3_EVAL}\u002Fannotations.json\"},\n",{"type":48,"tag":729,"props":825,"children":827},{"class":731,"line":826},8,[828],{"type":48,"tag":729,"props":829,"children":830},{},[831],{"type":54,"value":832},"}\n",{"type":48,"tag":57,"props":834,"children":835},{},[836],{"type":48,"tag":383,"props":837,"children":838},{},[839,841,846],{"type":54,"value":840},"deploy\u002Fgen_trt_engine (use ",{"type":48,"tag":73,"props":842,"children":844},{"className":843},[],[845],{"type":54,"value":102},{"type":54,"value":847},"):",{"type":48,"tag":718,"props":849,"children":851},{"className":720,"code":850,"language":722,"meta":723,"style":723},"{\n    \"gen_trt_engine.onnx_file\": \"\u003Cexported_onnx_uri>\",\n    \"gen_trt_engine.trt_engine\": \"\u003Coutput_engine_path>\",\n    \"gen_trt_engine.tensorrt.data_type\": \"FP16\",\n}\n",[852],{"type":48,"tag":73,"props":853,"children":854},{"__ignoreMap":723},[855,862,870,878,886],{"type":48,"tag":729,"props":856,"children":857},{"class":731,"line":732},[858],{"type":48,"tag":729,"props":859,"children":860},{},[861],{"type":54,"value":770},{"type":48,"tag":729,"props":863,"children":864},{"class":731,"line":741},[865],{"type":48,"tag":729,"props":866,"children":867},{},[868],{"type":54,"value":869},"    \"gen_trt_engine.onnx_file\": \"\u003Cexported_onnx_uri>\",\n",{"type":48,"tag":729,"props":871,"children":872},{"class":731,"line":781},[873],{"type":48,"tag":729,"props":874,"children":875},{},[876],{"type":54,"value":877},"    \"gen_trt_engine.trt_engine\": \"\u003Coutput_engine_path>\",\n",{"type":48,"tag":729,"props":879,"children":880},{"class":731,"line":790},[881],{"type":48,"tag":729,"props":882,"children":883},{},[884],{"type":54,"value":885},"    \"gen_trt_engine.tensorrt.data_type\": \"FP16\",\n",{"type":48,"tag":729,"props":887,"children":888},{"class":731,"line":799},[889],{"type":48,"tag":729,"props":890,"children":891},{},[892],{"type":54,"value":832},{"type":48,"tag":57,"props":894,"children":895},{},[896],{"type":48,"tag":383,"props":897,"children":898},{},[899],{"type":54,"value":900},"inference (mandatory data sources):",{"type":48,"tag":718,"props":902,"children":904},{"className":720,"code":903,"language":722,"meta":723,"style":723},"{\n    \"inference.checkpoint\": \"\u003Cselected train\u002FAutoML checkpoint>\",\n    \"dataset.infer_data_sources.image_dir\": [f\"{S3_EVAL}\u002Fimages.tar.gz\"],\n    \"dataset.infer_data_sources.captions\": [\n        \"fire extinguisher\",\n        \"cone\",\n        \"cart\",\n        \"forklift\"\n    ],\n}\n",[905],{"type":48,"tag":73,"props":906,"children":907},{"__ignoreMap":723},[908,915,923,931,939,947,955,963,971,980],{"type":48,"tag":729,"props":909,"children":910},{"class":731,"line":732},[911],{"type":48,"tag":729,"props":912,"children":913},{},[914],{"type":54,"value":770},{"type":48,"tag":729,"props":916,"children":917},{"class":731,"line":741},[918],{"type":48,"tag":729,"props":919,"children":920},{},[921],{"type":54,"value":922},"    \"inference.checkpoint\": \"\u003Cselected train\u002FAutoML checkpoint>\",\n",{"type":48,"tag":729,"props":924,"children":925},{"class":731,"line":781},[926],{"type":48,"tag":729,"props":927,"children":928},{},[929],{"type":54,"value":930},"    \"dataset.infer_data_sources.image_dir\": [f\"{S3_EVAL}\u002Fimages.tar.gz\"],\n",{"type":48,"tag":729,"props":932,"children":933},{"class":731,"line":790},[934],{"type":48,"tag":729,"props":935,"children":936},{},[937],{"type":54,"value":938},"    \"dataset.infer_data_sources.captions\": [\n",{"type":48,"tag":729,"props":940,"children":941},{"class":731,"line":799},[942],{"type":48,"tag":729,"props":943,"children":944},{},[945],{"type":54,"value":946},"        \"fire extinguisher\",\n",{"type":48,"tag":729,"props":948,"children":949},{"class":731,"line":808},[950],{"type":48,"tag":729,"props":951,"children":952},{},[953],{"type":54,"value":954},"        \"cone\",\n",{"type":48,"tag":729,"props":956,"children":957},{"class":731,"line":817},[958],{"type":48,"tag":729,"props":959,"children":960},{},[961],{"type":54,"value":962},"        \"cart\",\n",{"type":48,"tag":729,"props":964,"children":965},{"class":731,"line":826},[966],{"type":48,"tag":729,"props":967,"children":968},{},[969],{"type":54,"value":970},"        \"forklift\"\n",{"type":48,"tag":729,"props":972,"children":974},{"class":731,"line":973},9,[975],{"type":48,"tag":729,"props":976,"children":977},{},[978],{"type":54,"value":979},"    ],\n",{"type":48,"tag":729,"props":981,"children":983},{"class":731,"line":982},10,[984],{"type":48,"tag":729,"props":985,"children":986},{},[987],{"type":54,"value":832},{"type":48,"tag":57,"props":989,"children":990},{},[991],{"type":48,"tag":383,"props":992,"children":993},{},[994],{"type":54,"value":995},"evaluate (mandatory data sources):",{"type":48,"tag":718,"props":997,"children":999},{"className":720,"code":998,"language":722,"meta":723,"style":723},"{\n    \"evaluate.checkpoint\": \"\u003Cselected train\u002FAutoML checkpoint>\",\n    \"dataset.test_data_sources\": {\"image_dir\": f\"{S3_EVAL}\u002Fimages.tar.gz\", \"json_file\": f\"{S3_EVAL}\u002Fannotations.json\"},\n}\n",[1000],{"type":48,"tag":73,"props":1001,"children":1002},{"__ignoreMap":723},[1003,1010,1018,1026],{"type":48,"tag":729,"props":1004,"children":1005},{"class":731,"line":732},[1006],{"type":48,"tag":729,"props":1007,"children":1008},{},[1009],{"type":54,"value":770},{"type":48,"tag":729,"props":1011,"children":1012},{"class":731,"line":741},[1013],{"type":48,"tag":729,"props":1014,"children":1015},{},[1016],{"type":54,"value":1017},"    \"evaluate.checkpoint\": \"\u003Cselected train\u002FAutoML checkpoint>\",\n",{"type":48,"tag":729,"props":1019,"children":1020},{"class":731,"line":781},[1021],{"type":48,"tag":729,"props":1022,"children":1023},{},[1024],{"type":54,"value":1025},"    \"dataset.test_data_sources\": {\"image_dir\": f\"{S3_EVAL}\u002Fimages.tar.gz\", \"json_file\": f\"{S3_EVAL}\u002Fannotations.json\"},\n",{"type":48,"tag":729,"props":1027,"children":1028},{"class":731,"line":790},[1029],{"type":48,"tag":729,"props":1030,"children":1031},{},[1032],{"type":54,"value":832},{"type":48,"tag":57,"props":1034,"children":1035},{},[1036],{"type":48,"tag":383,"props":1037,"children":1038},{},[1039],{"type":54,"value":1040},"quantize (mandatory data sources):",{"type":48,"tag":718,"props":1042,"children":1044},{"className":720,"code":1043,"language":722,"meta":723,"style":723},"{\n    \"quantize.model_path\": \"\u003Cselected train checkpoint or exported ONNX model>\",\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_EVAL}\u002Fimages.tar.gz\", \"json_file\": f\"{S3_EVAL}\u002Fannotations.json\"},\n}\n",[1045],{"type":48,"tag":73,"props":1046,"children":1047},{"__ignoreMap":723},[1048,1055,1063,1070,1077,1085],{"type":48,"tag":729,"props":1049,"children":1050},{"class":731,"line":732},[1051],{"type":48,"tag":729,"props":1052,"children":1053},{},[1054],{"type":54,"value":770},{"type":48,"tag":729,"props":1056,"children":1057},{"class":731,"line":741},[1058],{"type":48,"tag":729,"props":1059,"children":1060},{},[1061],{"type":54,"value":1062},"    \"quantize.model_path\": \"\u003Cselected train checkpoint or exported ONNX model>\",\n",{"type":48,"tag":729,"props":1064,"children":1065},{"class":731,"line":781},[1066],{"type":48,"tag":729,"props":1067,"children":1068},{},[1069],{"type":54,"value":814},{"type":48,"tag":729,"props":1071,"children":1072},{"class":731,"line":790},[1073],{"type":48,"tag":729,"props":1074,"children":1075},{},[1076],{"type":54,"value":823},{"type":48,"tag":729,"props":1078,"children":1079},{"class":731,"line":799},[1080],{"type":48,"tag":729,"props":1081,"children":1082},{},[1083],{"type":54,"value":1084},"    \"dataset.quant_calibration_data_sources\": {\"image_dir\": f\"{S3_EVAL}\u002Fimages.tar.gz\", \"json_file\": f\"{S3_EVAL}\u002Fannotations.json\"},\n",{"type":48,"tag":729,"props":1086,"children":1087},{"class":731,"line":808},[1088],{"type":48,"tag":729,"props":1089,"children":1090},{},[1091],{"type":54,"value":832},{"type":48,"tag":122,"props":1093,"children":1095},{"id":1094},"eval-dataset",[1096],{"type":54,"value":1097},"Eval Dataset",{"type":48,"tag":57,"props":1099,"children":1100},{},[1101],{"type":54,"value":1102},"Optional. Validation uses COCO-format annotations for mAP even though training can use ODVG format.",{"type":48,"tag":122,"props":1104,"children":1106},{"id":1105},"important-parameters",[1107],{"type":54,"value":1108},"Important Parameters",{"type":48,"tag":375,"props":1110,"children":1111},{},[1112,1122,1132,1142,1152,1162,1172,1190],{"type":48,"tag":379,"props":1113,"children":1114},{},[1115,1120],{"type":48,"tag":383,"props":1116,"children":1117},{},[1118],{"type":54,"value":1119},"model.backbone",{"type":54,"value":1121},": Default swin_tiny_224_1k. Also supports resnet_50 and other Swin variants. Swin generally performs better for grounding tasks.",{"type":48,"tag":379,"props":1123,"children":1124},{},[1125,1130],{"type":48,"tag":383,"props":1126,"children":1127},{},[1128],{"type":54,"value":1129},"model.text_encoder_type",{"type":54,"value":1131},": BERT model for text encoding. Default bert-base-uncased. max_text_len defaults to 256.",{"type":48,"tag":379,"props":1133,"children":1134},{},[1135,1140],{"type":48,"tag":383,"props":1136,"children":1137},{},[1138],{"type":54,"value":1139},"model.max_text_len",{"type":54,"value":1141},": Keep this aligned with the dataset label\u002Ftoken\nposition maps. Do not shrink it for smoke tests unless the corresponding\nlabel maps are regenerated with the same length; otherwise validation can\nfail with a matrix shape mismatch between token probabilities and position\nmaps.",{"type":48,"tag":379,"props":1143,"children":1144},{},[1145,1150],{"type":48,"tag":383,"props":1146,"children":1147},{},[1148],{"type":54,"value":1149},"train.optim.lr",{"type":54,"value":1151},": Learning rate. Default 2e-4. lr_backbone 2e-5. Supports bf16 precision in addition to fp16\u002Ffp32.",{"type":48,"tag":379,"props":1153,"children":1154},{},[1155,1160],{"type":48,"tag":383,"props":1156,"children":1157},{},[1158],{"type":54,"value":1159},"dataset.max_labels",{"type":54,"value":1161},": Maximum labels per image during training. Default 50. Increase for dense annotation datasets.",{"type":48,"tag":379,"props":1163,"children":1164},{},[1165,1170],{"type":48,"tag":383,"props":1166,"children":1167},{},[1168],{"type":54,"value":1169},"model.num_queries",{"type":54,"value":1171},": Object queries. Default 900 (higher than DINO's 300) due to open-vocabulary nature.",{"type":48,"tag":379,"props":1173,"children":1174},{},[1175,1180,1182,1188],{"type":48,"tag":383,"props":1176,"children":1177},{},[1178],{"type":54,"value":1179},"model.num_queries \u002F model.num_select",{"type":54,"value":1181},": Keep ",{"type":48,"tag":73,"props":1183,"children":1185},{"className":1184},[],[1186],{"type":54,"value":1187},"num_queries",{"type":54,"value":1189}," high enough\nfor the number of matched ODVG targets in a batch. Very small smoke values\nsuch as 20 can fail during Hungarian target indexing on dense images; use at\nleast 100 for minimal Grounding DINO smoke runs unless the dataset is known\nto have fewer objects per image.",{"type":48,"tag":379,"props":1191,"children":1192},{},[1193,1198,1200,1205],{"type":48,"tag":383,"props":1194,"children":1195},{},[1196],{"type":54,"value":1197},"train.optim.lr_steps",{"type":54,"value":1199},": MultiStep LR schedule. Default ",{"type":48,"tag":729,"props":1201,"children":1202},{},[1203],{"type":54,"value":1204},"10",{"type":54,"value":716},{"type":48,"tag":122,"props":1207,"children":1209},{"id":1208},"multi-gpu-multi-node",[1210],{"type":54,"value":1211},"Multi-GPU \u002F Multi-Node",{"type":48,"tag":57,"props":1213,"children":1214},{},[1215,1220,1222,1227],{"type":48,"tag":383,"props":1216,"children":1217},{},[1218],{"type":54,"value":1219},"Launch method:",{"type":54,"value":1221}," Lightning-managed (single ",{"type":48,"tag":73,"props":1223,"children":1225},{"className":1224},[],[1226],{"type":54,"value":722},{"type":54,"value":1228}," process, Lightning spawns workers).",{"type":48,"tag":418,"props":1230,"children":1231},{},[1232,1252],{"type":48,"tag":422,"props":1233,"children":1234},{},[1235],{"type":48,"tag":426,"props":1236,"children":1237},{},[1238,1242,1247],{"type":48,"tag":430,"props":1239,"children":1240},{},[1241],{"type":54,"value":439},{"type":48,"tag":430,"props":1243,"children":1244},{},[1245],{"type":54,"value":1246},"Description",{"type":48,"tag":430,"props":1248,"children":1249},{},[1250],{"type":54,"value":1251},"Default",{"type":48,"tag":456,"props":1253,"children":1254},{},[1255,1277,1302,1323],{"type":48,"tag":426,"props":1256,"children":1257},{},[1258,1267,1272],{"type":48,"tag":463,"props":1259,"children":1260},{},[1261],{"type":48,"tag":73,"props":1262,"children":1264},{"className":1263},[],[1265],{"type":54,"value":1266},"train.num_gpus",{"type":48,"tag":463,"props":1268,"children":1269},{},[1270],{"type":54,"value":1271},"Number of GPUs",{"type":48,"tag":463,"props":1273,"children":1274},{},[1275],{"type":54,"value":1276},"1",{"type":48,"tag":426,"props":1278,"children":1279},{},[1280,1289,1294],{"type":48,"tag":463,"props":1281,"children":1282},{},[1283],{"type":48,"tag":73,"props":1284,"children":1286},{"className":1285},[],[1287],{"type":54,"value":1288},"train.gpu_ids",{"type":48,"tag":463,"props":1290,"children":1291},{},[1292],{"type":54,"value":1293},"GPU device indices",{"type":48,"tag":463,"props":1295,"children":1296},{},[1297],{"type":48,"tag":729,"props":1298,"children":1299},{},[1300],{"type":54,"value":1301},"0",{"type":48,"tag":426,"props":1303,"children":1304},{},[1305,1314,1319],{"type":48,"tag":463,"props":1306,"children":1307},{},[1308],{"type":48,"tag":73,"props":1309,"children":1311},{"className":1310},[],[1312],{"type":54,"value":1313},"train.num_nodes",{"type":48,"tag":463,"props":1315,"children":1316},{},[1317],{"type":54,"value":1318},"Number of nodes",{"type":48,"tag":463,"props":1320,"children":1321},{},[1322],{"type":54,"value":1276},{"type":48,"tag":426,"props":1324,"children":1325},{},[1326,1335,1352],{"type":48,"tag":463,"props":1327,"children":1328},{},[1329],{"type":48,"tag":73,"props":1330,"children":1332},{"className":1331},[],[1333],{"type":54,"value":1334},"train.distributed_strategy",{"type":48,"tag":463,"props":1336,"children":1337},{},[1338,1344,1346],{"type":48,"tag":73,"props":1339,"children":1341},{"className":1340},[],[1342],{"type":54,"value":1343},"ddp",{"type":54,"value":1345}," or ",{"type":48,"tag":73,"props":1347,"children":1349},{"className":1348},[],[1350],{"type":54,"value":1351},"fsdp",{"type":48,"tag":463,"props":1353,"children":1354},{},[1355],{"type":48,"tag":73,"props":1356,"children":1358},{"className":1357},[],[1359],{"type":54,"value":1343},{"type":48,"tag":57,"props":1361,"children":1362},{},[1363,1365,1371,1372,1378,1379,1385,1386,1392],{"type":54,"value":1364},"Same DDP\u002FFSDP behavior as DINO. Multi-node requires ",{"type":48,"tag":73,"props":1366,"children":1368},{"className":1367},[],[1369],{"type":54,"value":1370},"WORLD_SIZE",{"type":54,"value":204},{"type":48,"tag":73,"props":1373,"children":1375},{"className":1374},[],[1376],{"type":54,"value":1377},"NODE_RANK",{"type":54,"value":204},{"type":48,"tag":73,"props":1380,"children":1382},{"className":1381},[],[1383],{"type":54,"value":1384},"MASTER_ADDR",{"type":54,"value":204},{"type":48,"tag":73,"props":1387,"children":1389},{"className":1388},[],[1390],{"type":54,"value":1391},"MASTER_PORT",{"type":54,"value":1393}," env vars set by orchestrator.",{"type":48,"tag":122,"props":1395,"children":1397},{"id":1396},"export-trt-defaults",[1398],{"type":54,"value":1399},"Export \u002F TRT Defaults",{"type":48,"tag":375,"props":1401,"children":1402},{},[1403,1415,1467,1477,1482],{"type":48,"tag":379,"props":1404,"children":1405},{},[1406,1408,1414],{"type":54,"value":1407},"Export input: 960x544 (larger than other OD models), opset 17. Keep\nGrounding-DINO export specs at the template export resolution for smoke tests;\nreducing export to very small image sizes such as 128x128 can trigger a\nPyTorch ONNX shape-inference assertion in the contrastive text head during\n",{"type":48,"tag":73,"props":1409,"children":1411},{"className":1410},[],[1412],{"type":54,"value":1413},"torch.onnx.export",{"type":54,"value":716},{"type":48,"tag":379,"props":1416,"children":1417},{},[1418,1420,1426,1428,1433,1434,1439,1441,1446,1447,1452,1454,1459,1461,1466],{"type":54,"value":1419},"The parent PyTorch ",{"type":48,"tag":73,"props":1421,"children":1423},{"className":1422},[],[1424],{"type":54,"value":1425},"grounding_dino",{"type":54,"value":1427}," CLI supports ",{"type":48,"tag":73,"props":1429,"children":1431},{"className":1430},[],[1432],{"type":54,"value":623},{"type":54,"value":204},{"type":48,"tag":73,"props":1435,"children":1437},{"className":1436},[],[1438],{"type":54,"value":86},{"type":54,"value":1440},",\n",{"type":48,"tag":73,"props":1442,"children":1444},{"className":1443},[],[1445],{"type":54,"value":94},{"type":54,"value":204},{"type":48,"tag":73,"props":1448,"children":1450},{"className":1449},[],[1451],{"type":54,"value":358},{"type":54,"value":1453},", and ",{"type":48,"tag":73,"props":1455,"children":1457},{"className":1456},[],[1458],{"type":54,"value":547},{"type":54,"value":1460},". Run TensorRT engine generation,\nTensorRT inference, and TensorRT evaluation through ",{"type":48,"tag":73,"props":1462,"children":1464},{"className":1463},[],[1465],{"type":54,"value":102},{"type":54,"value":716},{"type":48,"tag":379,"props":1468,"children":1469},{},[1470,1472],{"type":54,"value":1471},"TRT data types: FP32, FP16 only — ",{"type":48,"tag":383,"props":1473,"children":1474},{},[1475],{"type":54,"value":1476},"INT8 is NOT supported",{"type":48,"tag":379,"props":1478,"children":1479},{},[1480],{"type":54,"value":1481},"TRT workspace: 8192 MB (8x larger than other OD models)",{"type":48,"tag":379,"props":1483,"children":1484},{},[1485],{"type":54,"value":1486},"TRT max_batch_size: 4",{"type":48,"tag":122,"props":1488,"children":1490},{"id":1489},"hardware",[1491],{"type":54,"value":1492},"Hardware",{"type":48,"tag":57,"props":1494,"children":1495},{},[1496],{"type":54,"value":1497},"Minimum 1 GPU(s), recommended 4 GPU(s). 24GB+ (A100 recommended) VRAM per GPU. Grounding DINO is heavier than standard DINO due to the text encoder (BERT). 24GB+ GPU memory recommended. Reduce batch_size for 16GB GPUs.",{"type":48,"tag":122,"props":1499,"children":1501},{"id":1500},"error-patterns",[1502],{"type":54,"value":1503},"Error Patterns",{"type":48,"tag":57,"props":1505,"children":1506},{},[1507,1512],{"type":48,"tag":383,"props":1508,"children":1509},{},[1510],{"type":54,"value":1511},"CUDA out of memory",{"type":54,"value":1513},": Reduce batch_size (4 -> 2 -> 1). The BERT text encoder adds significant memory overhead on top of the vision backbone.",{"type":48,"tag":57,"props":1515,"children":1516},{},[1517,1522],{"type":48,"tag":383,"props":1518,"children":1519},{},[1520],{"type":54,"value":1521},"Val annotation category IDs",{"type":54,"value":1523},": Validation annotations should have category IDs starting from 0 for correct loss computation. Use annotation format conversion if needed.",{"type":48,"tag":57,"props":1525,"children":1526},{},[1527,1532],{"type":48,"tag":383,"props":1528,"children":1529},{},[1530],{"type":54,"value":1531},"Text encoder loading error",{"type":54,"value":1533},": Ensure the container has access to download bert-base-uncased weights or provide a local path.",{"type":48,"tag":57,"props":1535,"children":1536},{},[1537,1542,1544,1550,1552,1558,1560,1566],{"type":48,"tag":383,"props":1538,"children":1539},{},[1540],{"type":54,"value":1541},"Quantize with a PyTorch checkpoint fails in TAO Toolkit 7.0.0-rc-226",{"type":54,"value":1543},":\nThe container's Grounding-DINO quantize script passes ",{"type":48,"tag":73,"props":1545,"children":1547},{"className":1546},[],[1548],{"type":54,"value":1549},"cap_lists=None",{"type":54,"value":1551}," when\nloading a checkpoint, which fails in ",{"type":48,"tag":73,"props":1553,"children":1555},{"className":1554},[],[1556],{"type":54,"value":1557},"post_process.py",{"type":54,"value":1559},". ONNX quantization uses\nthe exported ONNX artifact and COCO calibration data, but the default rc-226\nPyTorch image also lacks the ",{"type":48,"tag":73,"props":1561,"children":1563},{"className":1562},[],[1564],{"type":54,"value":1565},"modelopt.onnx.quantization",{"type":54,"value":1567}," module. Treat this as\nan image\u002FSDK blocker, not a checkpoint resolver issue.",{"type":48,"tag":57,"props":1569,"children":1570},{},[1571,1581,1583,1588,1590,1595],{"type":48,"tag":383,"props":1572,"children":1573},{},[1574,1576],{"type":54,"value":1575},"mat1 and mat2 shapes cannot be multiplied in ",{"type":48,"tag":73,"props":1577,"children":1579},{"className":1578},[],[1580],{"type":54,"value":1557},{"type":54,"value":1582},": The text\ntoken length and label position maps are inconsistent, commonly because\n",{"type":48,"tag":73,"props":1584,"children":1586},{"className":1585},[],[1587],{"type":54,"value":1139},{"type":54,"value":1589}," was overridden below the default 256 while the dataset\nlabel maps still use 256-length position maps. Restore ",{"type":48,"tag":73,"props":1591,"children":1593},{"className":1592},[],[1594],{"type":54,"value":1139},{"type":54,"value":1596}," or\nregenerate the label maps with the same length.",{"type":48,"tag":57,"props":1598,"children":1599},{},[1600,1611,1613,1618,1620,1625,1627,1633],{"type":48,"tag":383,"props":1601,"children":1602},{},[1603,1605],{"type":54,"value":1604},"index is out of bounds for dimension 0 in ",{"type":48,"tag":73,"props":1606,"children":1608},{"className":1607},[],[1609],{"type":54,"value":1610},"criterion.py",{"type":54,"value":1612},": ",{"type":48,"tag":73,"props":1614,"children":1616},{"className":1615},[],[1617],{"type":54,"value":1169},{"type":54,"value":1619},"\nis too small for the matched ODVG targets in the current batch. Increase\n",{"type":48,"tag":73,"props":1621,"children":1623},{"className":1622},[],[1624],{"type":54,"value":1169},{"type":54,"value":1626}," and keep ",{"type":48,"tag":73,"props":1628,"children":1630},{"className":1629},[],[1631],{"type":54,"value":1632},"model.num_select",{"type":54,"value":1634}," compatible with it.",{"type":48,"tag":57,"props":1636,"children":1637},{},[1638,1649,1651,1656,1658,1663],{"type":48,"tag":383,"props":1639,"children":1640},{},[1641,1643],{"type":54,"value":1642},"NotADirectoryError with ",{"type":48,"tag":73,"props":1644,"children":1646},{"className":1645},[],[1647],{"type":54,"value":1648},"images.tar.gz\u002F\u003Cimage>.jpg",{"type":54,"value":1650},": The direct TAO CLI is\ntrying to traverse an archive path as a directory. Extract the archive and set\nthe relevant ",{"type":48,"tag":73,"props":1652,"children":1654},{"className":1653},[],[1655],{"type":54,"value":680},{"type":54,"value":1657}," field to the extracted image folder; archive-backed\nskill data sources use ",{"type":48,"tag":73,"props":1659,"children":1661},{"className":1660},[],[1662],{"type":54,"value":688},{"type":54,"value":1664}," for this reason.",{"type":48,"tag":122,"props":1666,"children":1668},{"id":1667},"spec-param-parent-model-inference",[1669],{"type":54,"value":1670},"Spec Param \u002F Parent Model Inference",{"type":48,"tag":57,"props":1672,"children":1673},{},[1674,1676,1682,1684,1690,1692,1698],{"type":54,"value":1675},"Model-specific inference mappings belong in this MD file, not in ",{"type":48,"tag":73,"props":1677,"children":1679},{"className":1678},[],[1680],{"type":54,"value":1681},"config.json",{"type":54,"value":1683},". Generated runners should read this section and apply the mappings with SDK helpers before ",{"type":48,"tag":73,"props":1685,"children":1687},{"className":1686},[],[1688],{"type":54,"value":1689},"create_job()",{"type":54,"value":1691},". This mirrors the old microservices ",{"type":48,"tag":73,"props":1693,"children":1695},{"className":1694},[],[1696],{"type":54,"value":1697},"infer_params.py",{"type":54,"value":1699}," flow.",{"type":48,"tag":57,"props":1701,"children":1702},{},[1703,1705,1711],{"type":54,"value":1704},"Inference mappings from TAO Core ",{"type":48,"tag":73,"props":1706,"children":1708},{"className":1707},[],[1709],{"type":54,"value":1710},"grounding_dino.config.json",{"type":54,"value":1712},":",{"type":48,"tag":418,"props":1714,"children":1715},{},[1716,1741],{"type":48,"tag":422,"props":1717,"children":1718},{},[1719],{"type":48,"tag":426,"props":1720,"children":1721},{},[1722,1726,1731,1736],{"type":48,"tag":430,"props":1723,"children":1724},{},[1725],{"type":54,"value":434},{"type":48,"tag":430,"props":1727,"children":1728},{},[1729],{"type":54,"value":1730},"Spec Field",{"type":48,"tag":430,"props":1732,"children":1733},{},[1734],{"type":54,"value":1735},"Inference Function",{"type":48,"tag":430,"props":1737,"children":1738},{},[1739],{"type":54,"value":1740},"Meaning",{"type":48,"tag":456,"props":1742,"children":1743},{},[1744,1774,1804,1832,1862,1889,1917,1947,1974,2001,2029,2057,2084,2111,2139,2166,2193,2223,2250,2278],{"type":48,"tag":426,"props":1745,"children":1746},{},[1747,1751,1760,1769],{"type":48,"tag":463,"props":1748,"children":1749},{},[1750],{"type":54,"value":86},{"type":48,"tag":463,"props":1752,"children":1753},{},[1754],{"type":48,"tag":73,"props":1755,"children":1757},{"className":1756},[],[1758],{"type":54,"value":1759},"encryption_key",{"type":48,"tag":463,"props":1761,"children":1762},{},[1763],{"type":48,"tag":73,"props":1764,"children":1766},{"className":1765},[],[1767],{"type":54,"value":1768},"key",{"type":48,"tag":463,"props":1770,"children":1771},{},[1772],{"type":54,"value":1773},"encryption key",{"type":48,"tag":426,"props":1775,"children":1776},{},[1777,1781,1790,1799],{"type":48,"tag":463,"props":1778,"children":1779},{},[1780],{"type":54,"value":86},{"type":48,"tag":463,"props":1782,"children":1783},{},[1784],{"type":48,"tag":73,"props":1785,"children":1787},{"className":1786},[],[1788],{"type":54,"value":1789},"evaluate.checkpoint",{"type":48,"tag":463,"props":1791,"children":1792},{},[1793],{"type":48,"tag":73,"props":1794,"children":1796},{"className":1795},[],[1797],{"type":54,"value":1798},"parent_model",{"type":48,"tag":463,"props":1800,"children":1801},{},[1802],{"type":54,"value":1803},"model file inferred from the parent job results folder",{"type":48,"tag":426,"props":1805,"children":1806},{},[1807,1811,1820,1828],{"type":48,"tag":463,"props":1808,"children":1809},{},[1810],{"type":54,"value":86},{"type":48,"tag":463,"props":1812,"children":1813},{},[1814],{"type":48,"tag":73,"props":1815,"children":1817},{"className":1816},[],[1818],{"type":54,"value":1819},"evaluate.trt_engine",{"type":48,"tag":463,"props":1821,"children":1822},{},[1823],{"type":48,"tag":73,"props":1824,"children":1826},{"className":1825},[],[1827],{"type":54,"value":1798},{"type":48,"tag":463,"props":1829,"children":1830},{},[1831],{"type":54,"value":1803},{"type":48,"tag":426,"props":1833,"children":1834},{},[1835,1839,1848,1857],{"type":48,"tag":463,"props":1836,"children":1837},{},[1838],{"type":54,"value":86},{"type":48,"tag":463,"props":1840,"children":1841},{},[1842],{"type":48,"tag":73,"props":1843,"children":1845},{"className":1844},[],[1846],{"type":54,"value":1847},"results_dir",{"type":48,"tag":463,"props":1849,"children":1850},{},[1851],{"type":48,"tag":73,"props":1852,"children":1854},{"className":1853},[],[1855],{"type":54,"value":1856},"output_dir",{"type":48,"tag":463,"props":1858,"children":1859},{},[1860],{"type":54,"value":1861},"current job results directory",{"type":48,"tag":426,"props":1863,"children":1864},{},[1865,1869,1877,1885],{"type":48,"tag":463,"props":1866,"children":1867},{},[1868],{"type":54,"value":358},{"type":48,"tag":463,"props":1870,"children":1871},{},[1872],{"type":48,"tag":73,"props":1873,"children":1875},{"className":1874},[],[1876],{"type":54,"value":1759},{"type":48,"tag":463,"props":1878,"children":1879},{},[1880],{"type":48,"tag":73,"props":1881,"children":1883},{"className":1882},[],[1884],{"type":54,"value":1768},{"type":48,"tag":463,"props":1886,"children":1887},{},[1888],{"type":54,"value":1773},{"type":48,"tag":426,"props":1890,"children":1891},{},[1892,1896,1905,1913],{"type":48,"tag":463,"props":1893,"children":1894},{},[1895],{"type":54,"value":358},{"type":48,"tag":463,"props":1897,"children":1898},{},[1899],{"type":48,"tag":73,"props":1900,"children":1902},{"className":1901},[],[1903],{"type":54,"value":1904},"export.checkpoint",{"type":48,"tag":463,"props":1906,"children":1907},{},[1908],{"type":48,"tag":73,"props":1909,"children":1911},{"className":1910},[],[1912],{"type":54,"value":1798},{"type":48,"tag":463,"props":1914,"children":1915},{},[1916],{"type":54,"value":1803},{"type":48,"tag":426,"props":1918,"children":1919},{},[1920,1924,1933,1942],{"type":48,"tag":463,"props":1921,"children":1922},{},[1923],{"type":54,"value":358},{"type":48,"tag":463,"props":1925,"children":1926},{},[1927],{"type":48,"tag":73,"props":1928,"children":1930},{"className":1929},[],[1931],{"type":54,"value":1932},"export.onnx_file",{"type":48,"tag":463,"props":1934,"children":1935},{},[1936],{"type":48,"tag":73,"props":1937,"children":1939},{"className":1938},[],[1940],{"type":54,"value":1941},"create_onnx_file",{"type":48,"tag":463,"props":1943,"children":1944},{},[1945],{"type":54,"value":1946},"output ONNX path",{"type":48,"tag":426,"props":1948,"children":1949},{},[1950,1954,1962,1970],{"type":48,"tag":463,"props":1951,"children":1952},{},[1953],{"type":54,"value":358},{"type":48,"tag":463,"props":1955,"children":1956},{},[1957],{"type":48,"tag":73,"props":1958,"children":1960},{"className":1959},[],[1961],{"type":54,"value":1847},{"type":48,"tag":463,"props":1963,"children":1964},{},[1965],{"type":48,"tag":73,"props":1966,"children":1968},{"className":1967},[],[1969],{"type":54,"value":1856},{"type":48,"tag":463,"props":1971,"children":1972},{},[1973],{"type":54,"value":1861},{"type":48,"tag":426,"props":1975,"children":1976},{},[1977,1981,1989,1997],{"type":48,"tag":463,"props":1978,"children":1979},{},[1980],{"type":54,"value":94},{"type":48,"tag":463,"props":1982,"children":1983},{},[1984],{"type":48,"tag":73,"props":1985,"children":1987},{"className":1986},[],[1988],{"type":54,"value":1759},{"type":48,"tag":463,"props":1990,"children":1991},{},[1992],{"type":48,"tag":73,"props":1993,"children":1995},{"className":1994},[],[1996],{"type":54,"value":1768},{"type":48,"tag":463,"props":1998,"children":1999},{},[2000],{"type":54,"value":1773},{"type":48,"tag":426,"props":2002,"children":2003},{},[2004,2008,2017,2025],{"type":48,"tag":463,"props":2005,"children":2006},{},[2007],{"type":54,"value":94},{"type":48,"tag":463,"props":2009,"children":2010},{},[2011],{"type":48,"tag":73,"props":2012,"children":2014},{"className":2013},[],[2015],{"type":54,"value":2016},"inference.checkpoint",{"type":48,"tag":463,"props":2018,"children":2019},{},[2020],{"type":48,"tag":73,"props":2021,"children":2023},{"className":2022},[],[2024],{"type":54,"value":1798},{"type":48,"tag":463,"props":2026,"children":2027},{},[2028],{"type":54,"value":1803},{"type":48,"tag":426,"props":2030,"children":2031},{},[2032,2036,2045,2053],{"type":48,"tag":463,"props":2033,"children":2034},{},[2035],{"type":54,"value":94},{"type":48,"tag":463,"props":2037,"children":2038},{},[2039],{"type":48,"tag":73,"props":2040,"children":2042},{"className":2041},[],[2043],{"type":54,"value":2044},"inference.trt_engine",{"type":48,"tag":463,"props":2046,"children":2047},{},[2048],{"type":48,"tag":73,"props":2049,"children":2051},{"className":2050},[],[2052],{"type":54,"value":1798},{"type":48,"tag":463,"props":2054,"children":2055},{},[2056],{"type":54,"value":1803},{"type":48,"tag":426,"props":2058,"children":2059},{},[2060,2064,2072,2080],{"type":48,"tag":463,"props":2061,"children":2062},{},[2063],{"type":54,"value":94},{"type":48,"tag":463,"props":2065,"children":2066},{},[2067],{"type":48,"tag":73,"props":2068,"children":2070},{"className":2069},[],[2071],{"type":54,"value":1847},{"type":48,"tag":463,"props":2073,"children":2074},{},[2075],{"type":48,"tag":73,"props":2076,"children":2078},{"className":2077},[],[2079],{"type":54,"value":1856},{"type":48,"tag":463,"props":2081,"children":2082},{},[2083],{"type":54,"value":1861},{"type":48,"tag":426,"props":2085,"children":2086},{},[2087,2091,2099,2107],{"type":48,"tag":463,"props":2088,"children":2089},{},[2090],{"type":54,"value":547},{"type":48,"tag":463,"props":2092,"children":2093},{},[2094],{"type":48,"tag":73,"props":2095,"children":2097},{"className":2096},[],[2098],{"type":54,"value":1759},{"type":48,"tag":463,"props":2100,"children":2101},{},[2102],{"type":48,"tag":73,"props":2103,"children":2105},{"className":2104},[],[2106],{"type":54,"value":1768},{"type":48,"tag":463,"props":2108,"children":2109},{},[2110],{"type":54,"value":1773},{"type":48,"tag":426,"props":2112,"children":2113},{},[2114,2118,2127,2135],{"type":48,"tag":463,"props":2115,"children":2116},{},[2117],{"type":54,"value":547},{"type":48,"tag":463,"props":2119,"children":2120},{},[2121],{"type":48,"tag":73,"props":2122,"children":2124},{"className":2123},[],[2125],{"type":54,"value":2126},"quantize.model_path",{"type":48,"tag":463,"props":2128,"children":2129},{},[2130],{"type":48,"tag":73,"props":2131,"children":2133},{"className":2132},[],[2134],{"type":54,"value":1798},{"type":48,"tag":463,"props":2136,"children":2137},{},[2138],{"type":54,"value":1803},{"type":48,"tag":426,"props":2140,"children":2141},{},[2142,2146,2154,2162],{"type":48,"tag":463,"props":2143,"children":2144},{},[2145],{"type":54,"value":547},{"type":48,"tag":463,"props":2147,"children":2148},{},[2149],{"type":48,"tag":73,"props":2150,"children":2152},{"className":2151},[],[2153],{"type":54,"value":1847},{"type":48,"tag":463,"props":2155,"children":2156},{},[2157],{"type":48,"tag":73,"props":2158,"children":2160},{"className":2159},[],[2161],{"type":54,"value":1856},{"type":48,"tag":463,"props":2163,"children":2164},{},[2165],{"type":54,"value":1861},{"type":48,"tag":426,"props":2167,"children":2168},{},[2169,2173,2181,2189],{"type":48,"tag":463,"props":2170,"children":2171},{},[2172],{"type":54,"value":623},{"type":48,"tag":463,"props":2174,"children":2175},{},[2176],{"type":48,"tag":73,"props":2177,"children":2179},{"className":2178},[],[2180],{"type":54,"value":1759},{"type":48,"tag":463,"props":2182,"children":2183},{},[2184],{"type":48,"tag":73,"props":2185,"children":2187},{"className":2186},[],[2188],{"type":54,"value":1768},{"type":48,"tag":463,"props":2190,"children":2191},{},[2192],{"type":54,"value":1773},{"type":48,"tag":426,"props":2194,"children":2195},{},[2196,2200,2209,2218],{"type":48,"tag":463,"props":2197,"children":2198},{},[2199],{"type":54,"value":623},{"type":48,"tag":463,"props":2201,"children":2202},{},[2203],{"type":48,"tag":73,"props":2204,"children":2206},{"className":2205},[],[2207],{"type":54,"value":2208},"model.pretrained_backbone_path",{"type":48,"tag":463,"props":2210,"children":2211},{},[2212],{"type":48,"tag":73,"props":2213,"children":2215},{"className":2214},[],[2216],{"type":54,"value":2217},"ptm_if_no_resume_model",{"type":48,"tag":463,"props":2219,"children":2220},{},[2221],{"type":54,"value":2222},"PTM when no resume checkpoint exists",{"type":48,"tag":426,"props":2224,"children":2225},{},[2226,2230,2238,2246],{"type":48,"tag":463,"props":2227,"children":2228},{},[2229],{"type":54,"value":623},{"type":48,"tag":463,"props":2231,"children":2232},{},[2233],{"type":48,"tag":73,"props":2234,"children":2236},{"className":2235},[],[2237],{"type":54,"value":1847},{"type":48,"tag":463,"props":2239,"children":2240},{},[2241],{"type":48,"tag":73,"props":2242,"children":2244},{"className":2243},[],[2245],{"type":54,"value":1856},{"type":48,"tag":463,"props":2247,"children":2248},{},[2249],{"type":54,"value":1861},{"type":48,"tag":426,"props":2251,"children":2252},{},[2253,2257,2266,2274],{"type":48,"tag":463,"props":2254,"children":2255},{},[2256],{"type":54,"value":623},{"type":48,"tag":463,"props":2258,"children":2259},{},[2260],{"type":48,"tag":73,"props":2261,"children":2263},{"className":2262},[],[2264],{"type":54,"value":2265},"train.pretrained_model_path",{"type":48,"tag":463,"props":2267,"children":2268},{},[2269],{"type":48,"tag":73,"props":2270,"children":2272},{"className":2271},[],[2273],{"type":54,"value":2217},{"type":48,"tag":463,"props":2275,"children":2276},{},[2277],{"type":54,"value":2222},{"type":48,"tag":426,"props":2279,"children":2280},{},[2281,2285,2294,2303],{"type":48,"tag":463,"props":2282,"children":2283},{},[2284],{"type":54,"value":623},{"type":48,"tag":463,"props":2286,"children":2287},{},[2288],{"type":48,"tag":73,"props":2289,"children":2291},{"className":2290},[],[2292],{"type":54,"value":2293},"train.resume_training_checkpoint_path",{"type":48,"tag":463,"props":2295,"children":2296},{},[2297],{"type":48,"tag":73,"props":2298,"children":2300},{"className":2299},[],[2301],{"type":54,"value":2302},"resume_model",{"type":48,"tag":463,"props":2304,"children":2305},{},[2306],{"type":54,"value":2307},"model file inferred from the current job results folder",{"type":48,"tag":57,"props":2309,"children":2310},{},[2311,2313,2318,2319,2325,2327,2333,2335,2340],{"type":54,"value":2312},"For ",{"type":48,"tag":73,"props":2314,"children":2316},{"className":2315},[],[2317],{"type":54,"value":1798},{"type":54,"value":1345},{"type":48,"tag":73,"props":2320,"children":2322},{"className":2321},[],[2323],{"type":54,"value":2324},"parent_model_folder",{"type":54,"value":2326},", pass the upstream train\u002Fexport\u002FAutoML child job id as ",{"type":48,"tag":73,"props":2328,"children":2330},{"className":2329},[],[2331],{"type":54,"value":2332},"parent_job_id",{"type":54,"value":2334},". The SDK lists the parent result folder, filters checkpoint artifacts, and returns the selected model file or folder. Do not add these mappings back to ",{"type":48,"tag":73,"props":2336,"children":2338},{"className":2337},[],[2339],{"type":54,"value":1681},{"type":54,"value":2341}," and do not patch generated runner scripts to guess checkpoint paths.",{"type":48,"tag":57,"props":2343,"children":2344},{},[2345,2347,2353,2355,2361,2363,2368,2369,2374,2375,2380,2381,2387,2388,2393],{"type":54,"value":2346},"When selecting a Grounding-DINO checkpoint outside the SDK resolver, match the\nintended epoch\u002Fstep artifact exactly, for example\n",{"type":48,"tag":73,"props":2348,"children":2350},{"className":2349},[],[2351],{"type":54,"value":2352},"model_epoch_000_step_00046.pth",{"type":54,"value":2354},". The ",{"type":48,"tag":73,"props":2356,"children":2358},{"className":2357},[],[2359],{"type":54,"value":2360},"gdino_model_latest.pth",{"type":54,"value":2362}," symlink is valid\nonly when latest is explicitly requested. Carry structural model settings such\nas ",{"type":48,"tag":73,"props":2364,"children":2366},{"className":2365},[],[2367],{"type":54,"value":1119},{"type":54,"value":204},{"type":48,"tag":73,"props":2370,"children":2372},{"className":2371},[],[2373],{"type":54,"value":1169},{"type":54,"value":204},{"type":48,"tag":73,"props":2376,"children":2378},{"className":2377},[],[2379],{"type":54,"value":1632},{"type":54,"value":1440},{"type":48,"tag":73,"props":2382,"children":2384},{"className":2383},[],[2385],{"type":54,"value":2386},"model.num_feature_levels",{"type":54,"value":204},{"type":48,"tag":73,"props":2389,"children":2391},{"className":2390},[],[2392],{"type":54,"value":1139},{"type":54,"value":2394},", and export input resolution\nforward into evaluate, inference, export, and deploy specs so checkpoint and\nengine shapes match.",{"type":48,"tag":122,"props":2396,"children":2398},{"id":2397},"deployment",[2399],{"type":54,"value":2400},"Deployment",{"type":48,"tag":375,"props":2402,"children":2403},{},[2404],{"type":48,"tag":379,"props":2405,"children":2406},{},[2407],{"type":48,"tag":2408,"props":2409,"children":2410},"a",{"href":102},[2411],{"type":54,"value":2412},"tao-deploy-grounding-dino",{"type":48,"tag":2414,"props":2415,"children":2416},"style",{},[2417],{"type":54,"value":2418},"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":2420,"total":2522},[2421,2438,2450,2464,2476,2493,2508],{"slug":2422,"name":2422,"fn":2423,"description":2424,"org":2425,"tags":2426,"stars":23,"repoUrl":24,"updatedAt":2437},"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},[2427,2430,2433,2434],{"name":2428,"slug":2429,"type":15},"Data Analysis","data-analysis",{"name":2431,"slug":2432,"type":15},"Data Engineering","data-engineering",{"name":9,"slug":8,"type":15},{"name":2435,"slug":2436,"type":15},"Performance","performance","2026-07-14T05:28:43.176466",{"slug":2439,"name":2439,"fn":2440,"description":2441,"org":2442,"tags":2443,"stars":23,"repoUrl":24,"updatedAt":2449},"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},[2444,2445,2448],{"name":2400,"slug":2397,"type":15},{"name":2446,"slug":2447,"type":15},"Infrastructure","infrastructure",{"name":9,"slug":8,"type":15},"2026-07-14T05:29:06.667109",{"slug":2451,"name":2451,"fn":2452,"description":2453,"org":2454,"tags":2455,"stars":23,"repoUrl":24,"updatedAt":2463},"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},[2456,2459,2460],{"name":2457,"slug":2458,"type":15},"Agents","agents",{"name":9,"slug":8,"type":15},{"name":2461,"slug":2462,"type":15},"Research","research","2026-07-14T05:28:06.816956",{"slug":2465,"name":2465,"fn":2466,"description":2467,"org":2468,"tags":2469,"stars":23,"repoUrl":24,"updatedAt":2475},"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},[2470,2471,2472],{"name":2428,"slug":2429,"type":15},{"name":9,"slug":8,"type":15},{"name":2473,"slug":2474,"type":15},"Testing","testing","2026-07-17T05:29:03.913266",{"slug":2477,"name":2477,"fn":2478,"description":2479,"org":2480,"tags":2481,"stars":23,"repoUrl":24,"updatedAt":2492},"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},[2482,2485,2488,2489],{"name":2483,"slug":2484,"type":15},"Automation","automation",{"name":2486,"slug":2487,"type":15},"Imaging","imaging",{"name":9,"slug":8,"type":15},{"name":2490,"slug":2491,"type":15},"Video","video","2026-07-17T05:28:53.905004",{"slug":2494,"name":2494,"fn":2495,"description":2496,"org":2497,"tags":2498,"stars":23,"repoUrl":24,"updatedAt":2507},"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},[2499,2500,2503,2504],{"name":2400,"slug":2397,"type":15},{"name":2501,"slug":2502,"type":15},"Docker","docker",{"name":9,"slug":8,"type":15},{"name":2505,"slug":2506,"type":15},"Operations","operations","2026-07-17T05:28:56.913999",{"slug":2509,"name":2509,"fn":2510,"description":2511,"org":2512,"tags":2513,"stars":23,"repoUrl":24,"updatedAt":2521},"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},[2514,2515,2518],{"name":9,"slug":8,"type":15},{"name":2516,"slug":2517,"type":15},"Quantum Computing","quantum-computing",{"name":2519,"slug":2520,"type":15},"Simulation","simulation","2026-07-14T05:26:58.898253",305,{"items":2524,"total":2672},[2525,2543,2558,2569,2581,2595,2608,2620,2631,2640,2654,2663],{"slug":2526,"name":2526,"fn":2527,"description":2528,"org":2529,"tags":2530,"stars":2540,"repoUrl":2541,"updatedAt":2542},"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},[2531,2534,2537],{"name":2532,"slug":2533,"type":15},"Documentation","documentation",{"name":2535,"slug":2536,"type":15},"MCP","mcp",{"name":2538,"slug":2539,"type":15},"Search","search",21777,"https:\u002F\u002Fgithub.com\u002FNVIDIA\u002FNemoClaw","2026-07-20T06:00:01.461044",{"slug":2544,"name":2544,"fn":2545,"description":2546,"org":2547,"tags":2548,"stars":2555,"repoUrl":2556,"updatedAt":2557},"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},[2549,2552,2553],{"name":2550,"slug":2551,"type":15},"Containers","containers",{"name":2400,"slug":2397,"type":15},{"name":2554,"slug":722,"type":15},"Python",17049,"https:\u002F\u002Fgithub.com\u002FNVIDIA\u002FMegatron-LM","2026-07-27T06:06:11.249662",{"slug":2559,"name":2559,"fn":2560,"description":2561,"org":2562,"tags":2563,"stars":2555,"repoUrl":2556,"updatedAt":2568},"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},[2564,2567],{"name":2565,"slug":2566,"type":15},"CI\u002FCD","ci-cd",{"name":2400,"slug":2397,"type":15},"2026-07-14T05:25:59.97109",{"slug":2570,"name":2570,"fn":2571,"description":2572,"org":2573,"tags":2574,"stars":2555,"repoUrl":2556,"updatedAt":2580},"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},[2575,2576,2577],{"name":2565,"slug":2566,"type":15},{"name":2400,"slug":2397,"type":15},{"name":2578,"slug":2579,"type":15},"GitHub","github","2026-07-27T06:06:12.278222",{"slug":2582,"name":2582,"fn":2583,"description":2584,"org":2585,"tags":2586,"stars":2555,"repoUrl":2556,"updatedAt":2594},"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},[2587,2590,2591],{"name":2588,"slug":2589,"type":15},"Debugging","debugging",{"name":2578,"slug":2579,"type":15},{"name":2592,"slug":2593,"type":15},"Triage","triage","2026-07-14T05:25:57.442089",{"slug":2596,"name":2596,"fn":2597,"description":2598,"org":2599,"tags":2600,"stars":2555,"repoUrl":2556,"updatedAt":2607},"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},[2601,2604],{"name":2602,"slug":2603,"type":15},"Best Practices","best-practices",{"name":2605,"slug":2606,"type":15},"Code Analysis","code-analysis","2026-07-14T05:25:56.18433",{"slug":2609,"name":2609,"fn":2610,"description":2611,"org":2612,"tags":2613,"stars":2555,"repoUrl":2556,"updatedAt":2619},"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},[2614,2615,2618],{"name":17,"slug":18,"type":15},{"name":2616,"slug":2617,"type":15},"Migration","migration",{"name":9,"slug":8,"type":15},"2026-07-17T06:07:11.777011",{"slug":2621,"name":2621,"fn":2622,"description":2623,"org":2624,"tags":2625,"stars":2555,"repoUrl":2556,"updatedAt":2630},"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},[2626,2629],{"name":2627,"slug":2628,"type":15},"QA","qa",{"name":2473,"slug":2474,"type":15},"2026-07-14T05:25:53.673039",{"slug":2632,"name":2632,"fn":2633,"description":2634,"org":2635,"tags":2636,"stars":2555,"repoUrl":2556,"updatedAt":2639},"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},[2637,2638],{"name":2400,"slug":2397,"type":15},{"name":2446,"slug":2447,"type":15},"2026-07-14T05:25:49.362534",{"slug":2641,"name":2641,"fn":2642,"description":2643,"org":2644,"tags":2645,"stars":2555,"repoUrl":2556,"updatedAt":2653},"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},[2646,2649,2650],{"name":2647,"slug":2648,"type":15},"Code Review","code-review",{"name":2578,"slug":2579,"type":15},{"name":2651,"slug":2652,"type":15},"Pull Requests","pull-requests","2026-07-14T05:26:01.226578",{"slug":2655,"name":2655,"fn":2656,"description":2657,"org":2658,"tags":2659,"stars":2555,"repoUrl":2556,"updatedAt":2662},"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},[2660,2661],{"name":2627,"slug":2628,"type":15},{"name":2473,"slug":2474,"type":15},"2026-07-14T05:25:54.928983",{"slug":2664,"name":2664,"fn":2665,"description":2666,"org":2667,"tags":2668,"stars":2555,"repoUrl":2556,"updatedAt":2671},"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},[2669,2670],{"name":2483,"slug":2484,"type":15},{"name":2565,"slug":2566,"type":15},"2026-07-30T05:29:03.275638",496]