[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill-nvidia-tao-train-sparse4d":3,"mdc--7ribeg-key":34,"related-org-nvidia-tao-train-sparse4d":2226,"related-repo-nvidia-tao-train-sparse4d":2386},{"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-sparse4d","train Sparse4D models for 3D perception","Sparse4D for multi-camera temporal 3D object detection and tracking. Uses sparse queries with deformable attention across camera views and time for end-to-end 3D perception, with an instance bank for temporal tracking. Use when training, evaluating, exporting, quantizing, or running inference for a TAO Sparse4D model. Trigger phrases include \"train Sparse4D\", \"multi-camera 3D detection\", \"temporal 3D tracker\", \"sparse query 3D perception\".",{"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},"3D","3d",{"name":9,"slug":8,"type":15},{"name":21,"slug":22,"type":15},"Computer Vision","computer-vision",2473,"https:\u002F\u002Fgithub.com\u002FNVIDIA\u002Fskills","2026-07-14T05:29:41.970999","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-sparse4d","---\nname: tao-train-sparse4d\ndescription: Sparse4D for multi-camera temporal 3D object detection and tracking. Uses sparse queries with deformable\n  attention across camera views and time for end-to-end 3D perception, with an instance bank for temporal tracking. Use when\n  training, evaluating, exporting, quantizing, or running inference for a TAO Sparse4D model. Trigger phrases include\n  \"train Sparse4D\", \"multi-camera 3D detection\", \"temporal 3D tracker\", \"sparse query 3D perception\".\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- temporal\n- 3d\n- detection\n- tracking\n---\n\n# Sparse4D\n\nSparse4D for multi-camera temporal 3D object detection and tracking. Uses sparse queries with deformable attention across camera views and time for end-to-end 3D perception. Includes instance bank for temporal tracking.\n\nUse a pretrained ResNet-101 backbone when one is available by setting\n`train.pretrained_model_path`. For local smoke validation, Sparse4D training\ncan run with an empty `train.pretrained_model_path`, but production runs should\nstill use a compatible PTM.\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:** sparse4d\n- **Formats:** ovpkl\n- **Monitoring metric:** val_mAP\n- Current TAO Sparse4D training emits this value in status\u002Flogs as\n  `img_bbox_NuScenes\u002FmAP` and `mAP`; AutoML metric\n  extractors should treat those emitted keys as aliases for `val_mAP`.\n  Multi-fidelity AutoML algorithms such as Hyperband, ASHA, and BOHB may\n  promote a checkpoint to a resume job that completes without emitting a fresh\n  `val_mAP` alias. In that case, compare AutoML's carried metric to the source\n  rung job that emitted `img_bbox_NuScenes\u002FmAP` or `mAP`, while still verifying\n  that the promoted job resumed from the explicit epoch\u002Fstep checkpoint,\n  produced a real checkpoint, and is usable for evaluate\u002Finference.\n\n### Per-Action Dataset Requirements\n\n| Action | Spec Key | Source | Files | List? |\n|---|---|---|---|---|\n| dataset_convert | aicity.root | id |  | No |\n| evaluate | dataset.data_root | eval_dataset | (from convert job, spec: aicity.split) | No |\n| evaluate | model.head.instance_bank.anchor | train_datasets | \u002Fresults\u002F{dataset_convert_job_id}\u002Fanchor_init.npy | No |\n| evaluate | dataset.train_dataset.ann_file | train_datasets | (from convert job, spec: aicity.split) | No |\n| evaluate | dataset.val_dataset.ann_file | eval_dataset | (from convert job, spec: aicity.split) | No |\n| evaluate | dataset.test_dataset.ann_file | inference_dataset | (from convert job, spec: aicity.split) | No |\n| export | model.head.instance_bank.anchor | train_datasets | \u002Fresults\u002F{dataset_convert_job_id}\u002Fanchor_init.npy | No |\n| inference | dataset.data_root | inference_dataset | (from convert job, spec: aicity.split) | No |\n| inference | model.head.instance_bank.anchor | train_datasets | \u002Fresults\u002F{dataset_convert_job_id}\u002Fanchor_init.npy | No |\n| inference | dataset.train_dataset.ann_file | train_datasets | (from convert job, spec: aicity.split) | No |\n| inference | dataset.val_dataset.ann_file | eval_dataset | (from convert job, spec: aicity.split) | No |\n| inference | dataset.test_dataset.ann_file | inference_dataset | (from convert job, spec: aicity.split) | No |\n| quantize | dataset.data_root | train_datasets | (from convert job, spec: aicity.split) | No |\n| quantize | model.head.instance_bank.anchor | train_datasets | \u002Fresults\u002F{dataset_convert_job_id}\u002Fanchor_init.npy | No |\n| quantize | dataset.train_dataset.ann_file | train_datasets | (from convert job, spec: aicity.split) | No |\n| quantize | dataset.val_dataset.ann_file | eval_dataset | (from convert job, spec: aicity.split) | No |\n| quantize | dataset.test_dataset.ann_file | inference_dataset | (from convert job, spec: aicity.split) | No |\n| quantize | dataset.quant_calibration_dataset.images_dir | train_datasets |  | No |\n| train | dataset.data_root | train_datasets | (from convert job, spec: aicity.split) | No |\n| train | model.head.instance_bank.anchor | train_datasets | \u002Fresults\u002F{dataset_convert_job_id}\u002Fanchor_init.npy | No |\n| train | dataset.train_dataset.ann_file | train_datasets | (from convert job, spec: aicity.split) | No |\n| train | dataset.val_dataset.ann_file | eval_dataset | (from convert job, spec: aicity.split) | No |\n| train | dataset.test_dataset.ann_file | inference_dataset | (from convert job, spec: aicity.split) | 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\"\nCONVERTED_SCENE = \"\u003Cscene-from-converter>\"  # e.g. \"subsetscene+bev-sensor-random-0\"\n```\n\n**train (mandatory data sources):**\n```python\nCONVERTED = \"s3:\u002F\u002Fbucket\u002Fresults\u002F\u003Cdataset_convert_job_id>\"\n{\n    \"train.num_epochs\": 30,\n    \"train.checkpoint_interval\": 10,\n    \"train.validation_interval\": 10,\n    \"train.num_gpus\": 1,\n    \"dataset.sequences.split_num\": 90,\n    \"dataset.train_dataset.sequences_split_num\": 90,\n    \"dataset.data_root\": f\"{S3_TRAIN}\u002Ftrain\",\n    \"model.head.instance_bank.anchor\": f\"{CONVERTED}\u002Fanchor_init.npy\",\n    \"dataset.train_dataset.ann_file\": f\"{CONVERTED}\u002Ftrain\u002F{CONVERTED_SCENE}_infos_train.pkl\",\n    \"dataset.val_dataset.ann_file\": f\"{CONVERTED}\u002Fval\u002F{CONVERTED_SCENE}_infos_val.pkl\",\n    \"dataset.test_dataset.ann_file\": f\"{CONVERTED}\u002Ftest\u002F{CONVERTED_SCENE}_infos_test.pkl\",\n}\n```\n\n**evaluate (mandatory data sources):**\n```python\nCONVERTED = \"s3:\u002F\u002Fbucket\u002Fresults\u002F\u003Cdataset_convert_job_id>\"\n{\n    \"dataset.data_root\": f\"{S3_EVAL}\u002Fval\",\n    \"model.head.instance_bank.anchor\": f\"{CONVERTED}\u002Fanchor_init.npy\",\n    \"dataset.train_dataset.ann_file\": f\"{CONVERTED}\u002Ftrain\u002F{CONVERTED_SCENE}_infos_train.pkl\",\n    \"dataset.val_dataset.ann_file\": f\"{CONVERTED}\u002Fval\u002F{CONVERTED_SCENE}_infos_val.pkl\",\n    \"dataset.test_dataset.ann_file\": f\"{CONVERTED}\u002Ftest\u002F{CONVERTED_SCENE}_infos_test.pkl\",\n}\n```\n\n**export (mandatory data sources):**\n```python\nCONVERTED = \"s3:\u002F\u002Fbucket\u002Fresults\u002F\u003Cdataset_convert_job_id>\"\n{\n    \"model.head.instance_bank.anchor\": f\"{CONVERTED}\u002Fanchor_init.npy\",\n}\n```\n\n**inference (mandatory data sources):**\n```python\nCONVERTED = \"s3:\u002F\u002Fbucket\u002Fresults\u002F\u003Cdataset_convert_job_id>\"\n{\n    \"dataset.data_root\": f\"{S3_EVAL}\u002Ftest\",\n    \"model.head.instance_bank.anchor\": f\"{CONVERTED}\u002Fanchor_init.npy\",\n    \"dataset.train_dataset.ann_file\": f\"{CONVERTED}\u002Ftrain\u002F{CONVERTED_SCENE}_infos_train.pkl\",\n    \"dataset.val_dataset.ann_file\": f\"{CONVERTED}\u002Fval\u002F{CONVERTED_SCENE}_infos_val.pkl\",\n    \"dataset.test_dataset.ann_file\": f\"{CONVERTED}\u002Ftest\u002F{CONVERTED_SCENE}_infos_test.pkl\",\n}\n```\n\n**quantize (mandatory data sources):**\n```python\nCONVERTED = \"s3:\u002F\u002Fbucket\u002Fresults\u002F\u003Cdataset_convert_job_id>\"\n{\n    \"dataset.data_root\": f\"{S3_TRAIN}\u002Ftrain\",\n    \"model.head.instance_bank.anchor\": f\"{CONVERTED}\u002Fanchor_init.npy\",\n    \"dataset.train_dataset.ann_file\": f\"{CONVERTED}\u002Ftrain\u002F{CONVERTED_SCENE}_infos_train.pkl\",\n    \"dataset.val_dataset.ann_file\": f\"{CONVERTED}\u002Fval\u002F{CONVERTED_SCENE}_infos_val.pkl\",\n    \"dataset.test_dataset.ann_file\": f\"{CONVERTED}\u002Ftest\u002F{CONVERTED_SCENE}_infos_test.pkl\",\n    \"dataset.quant_calibration_dataset.images_dir\": f\"{S3_TRAIN}\",\n}\n```\n\nSee `references\u002Flocal_docker_conversion.md` for local-docker conversion roots and mounts, H5 depth-path normalization, converted annotation filenames, smoke-run `max_num_cams`\u002Fanchor contracts for export compatibility, and converted-artifact verification before train\u002Fevaluate\u002Finference.\n\n## Eval Dataset\n\nOptional. Val\u002Ftest splits configured via dataset ann_file paths.\n\n## Important Parameters\n\n- **model.backbone**: Backbone. Default resnet_101.\n- **model.neck.out_channels**: FPN output channels. Default 256. num_outs=4.\n- **model.input_shape**: Input image shape [W, H]. Default [1408, 512].\n- **model.head.num_output**: Number of detection output queries. Default 300.\n- **model.head.num_decoder**: Number of decoder layers. Default 6.\n- **model.head.temporal**: Enable temporal reasoning. Default True.\n- **model.head.instance_bank.num_anchor**: Instance bank anchors. Default 900.\n- **model.head.instance_bank.num_temp_instances**: Temporal instance count. Default 600.\n- **model.depth_branch.loss_weight**: Depth supervision loss weight. Default 0.2.\n- **dataset.batch_size**: Per-GPU batch size. Default 2.\n- **dataset.num_frames**: Sequence length. Default 200.\n- **dataset.classes**: Detection classes. Default [person, gr1_t2, agility_digit, nova_carter]. num_ids=70 for tracking.\n- **train.optim.lr**: Learning rate. Default 5e-5. img_backbone lr_mult=0.2.\n- **train.lr_scheduler**: Cosine scheduler with linear warmup (500 iters, ratio 0.333).\n- **train.grad_clip.max_norm**: Gradient clipping. Default 25.\n- **train.precision**: Options: bf16, fp16, fp32. Default bf16.\n- **evaluate.metrics**: Eval metrics. Default [\"detection\"]. Optional tracking evaluation.\n- **evaluate.tracking.enabled**: Enable tracking evaluation. tracking_threshold=0.2.\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\n- Multi-GPU strategy: `ddp_find_unused_parameters_true` (no fsdp support)\n- `sync_batchnorm` is always enabled (True)\n- Iterations per epoch computed as: `num_frames * num_bev_groups \u002F (num_nodes * num_gpus * batch_size)`\n- **Scaling:** When increasing GPUs, effective batch size grows and iterations-per-epoch shrinks proportionally\n\n**Multi-node env vars** (set by orchestrator): `WORLD_SIZE`, `NODE_RANK`, `MASTER_ADDR`, `MASTER_PORT`, `NUM_GPU_PER_NODE`.\n\n## Hardware\n\nMinimum 2 GPU(s), recommended 8 GPU(s). 40GB+ (A100 recommended) VRAM per GPU. Multi-camera temporal model is memory intensive. bf16 required for practical training. Multi-GPU strongly recommended. Instance bank requires substantial memory for temporal reasoning.\n\n## Error Patterns\n\n**dataset_convert required**: Must run dataset_convert first to produce annotation pickles and anchor_init.npy.\n\n**dataset_convert container\u002Fcommand**: Sparse4D conversion is an AICity to\nOVPKL annotations conversion. Launch `dataset_convert` with the action-level\n`tao_toolkit.data_services` image and `annotations convert -e {config_path}`;\ndo not use the PyTorch `sparse4d` CLI for conversion. Train\u002Fevaluate\u002Fexport\u002F\ninference still use the model-level PyTorch image.\n\n**Stable raw-data path**: The AICity to OVPKL converter writes image paths into\nthe generated pickle files. Keep `aicity.root` at `\u002Fdata\u002Faicity_root` during\nconversion, then point `dataset.data_root` at the split folder, for example\n`\u002Fdata\u002Faicity_root\u002Ftrain` for training or `\u002Fdata\u002Faicity_root\u002Fval` for\nevaluation. This preserves the converter's absolute RGB paths and relative\ndepth paths.\n\n**H5 depth tuple mismatch**: If training fails with an H5 path error where the\ntrainer tries to open a camera directory such as\n`\u002Fdata\u002Faicity_root\u002Ftrain\u002F\u003Cscene>\u002FCamera`, run\n`models\u002Fsparse4d\u002Fscripts\u002Fnormalize_depth_paths.py --data-root \u003Chost-aicity-root>\u002Ftrain \u003Cconverted-ann-dir>`\nafter `dataset_convert` and before train\u002Fevaluate\u002Finference. The helper rewrites\nconverted `depth_map_path` tuples to point at\n`\u003Cscene>\u002Fdepth_maps\u002F\u003Ccamera>.h5` with the H5 dataset key basename.\n\n**Missing anchor file**: Set model.head.instance_bank.anchor to the anchor_init.npy path from dataset_convert results.\n\n**Temporal OOM**: Reduce dataset.num_frames or dataset.batch_size if running out of memory during temporal training.\n\n**Quantize image compatibility**: The model-skill wiring should pass\n`quantize.model_path` through the parent-model resolver, and checkpoint handoff\nshould select the exact epoch\u002Fstep checkpoint just like evaluate, inference,\nexport, and resume. TorchAO checkpoint quantization passes in the\n`validation-fixes-20260525` PyT image and writes\n`quantized_model_torchao.pth`. Older 7.0.0-rc PyT images may fail inside the\nSparse4D quantize entrypoint or lack ONNX quantization dependencies; do not\nremove or skip the advertised `quantize` action if that occurs. Report the\ncontainer\u002Fimage failure and keep the exact checkpoint path visible.\n\n## Spec Param \u002F Parent Model Inference\n\nSee `references\u002Fspec_param_inference.md` for the model-specific inference mappings from TAO Core `sparse4d.config.json` (the per-action spec-field to inference-function table) and the `parent_model`\u002F`parent_job_id` checkpoint-resolution rules that generated runners apply with SDK helpers before `create_job()`.\n",{"data":35,"body":45},{"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",[42,18,43,44],"temporal","detection","tracking",{"type":46,"children":47},"root",[48,57,63,84,91,184,190,299,333,339,424,431,1018,1024,1044,1084,1092,1222,1230,1294,1302,1337,1345,1409,1417,1488,1509,1515,1520,1526,1736,1742,1759,1853,1901,1946,1952,1957,1963,1973,2013,2061,2110,2120,2130,2171,2177,2220],{"type":49,"tag":50,"props":51,"children":53},"element","h1",{"id":52},"sparse4d",[54],{"type":55,"value":56},"text","Sparse4D",{"type":49,"tag":58,"props":59,"children":60},"p",{},[61],{"type":55,"value":62},"Sparse4D for multi-camera temporal 3D object detection and tracking. Uses sparse queries with deformable attention across camera views and time for end-to-end 3D perception. Includes instance bank for temporal tracking.",{"type":49,"tag":58,"props":64,"children":65},{},[66,68,75,77,82],{"type":55,"value":67},"Use a pretrained ResNet-101 backbone when one is available by setting\n",{"type":49,"tag":69,"props":70,"children":72},"code",{"className":71},[],[73],{"type":55,"value":74},"train.pretrained_model_path",{"type":55,"value":76},". For local smoke validation, Sparse4D training\ncan run with an empty ",{"type":49,"tag":69,"props":78,"children":80},{"className":79},[],[81],{"type":55,"value":74},{"type":55,"value":83},", but production runs should\nstill use a compatible PTM.",{"type":49,"tag":85,"props":86,"children":88},"h2",{"id":87},"dataclass-schemas",[89],{"type":55,"value":90},"Dataclass Schemas",{"type":49,"tag":58,"props":92,"children":93},{},[94,96,102,104,110,112,118,120,126,128,134,136,142,144,150,152,158,160,166,168,174,176,182],{"type":55,"value":95},"Generated TAO Core schemas are packaged in ",{"type":49,"tag":69,"props":97,"children":99},{"className":98},[],[100],{"type":55,"value":101},"schemas\u002F\u003Caction>.schema.json",{"type":55,"value":103},", with ",{"type":49,"tag":69,"props":105,"children":107},{"className":106},[],[108],{"type":55,"value":109},"schemas\u002Fmanifest.json",{"type":55,"value":111}," listing available actions. Each generated schema also emits ",{"type":49,"tag":69,"props":113,"children":115},{"className":114},[],[116],{"type":55,"value":117},"references\u002Fspec_template_\u003Caction>.yaml",{"type":55,"value":119}," from the schema top-level ",{"type":49,"tag":69,"props":121,"children":123},{"className":122},[],[124],{"type":55,"value":125},"default",{"type":55,"value":127}," field. AutoML enablement is declared at the model layer in ",{"type":49,"tag":69,"props":129,"children":131},{"className":130},[],[132],{"type":55,"value":133},"references\u002Fskill_info.yaml",{"type":55,"value":135}," via ",{"type":49,"tag":69,"props":137,"children":139},{"className":138},[],[140],{"type":55,"value":141},"automl_enabled",{"type":55,"value":143},". Runnable AutoML still requires ",{"type":49,"tag":69,"props":145,"children":147},{"className":146},[],[148],{"type":55,"value":149},"schemas\u002Ftrain.schema.json",{"type":55,"value":151}," and ",{"type":49,"tag":69,"props":153,"children":155},{"className":154},[],[156],{"type":55,"value":157},"references\u002Fspec_template_train.yaml",{"type":55,"value":159}," to exist and parse. Use the packaged train schema for ",{"type":49,"tag":69,"props":161,"children":163},{"className":162},[],[164],{"type":55,"value":165},"automl_default_parameters",{"type":55,"value":167},", ",{"type":49,"tag":69,"props":169,"children":171},{"className":170},[],[172],{"type":55,"value":173},"automl_disabled_parameters",{"type":55,"value":175},", defaults, min\u002Fmax bounds, enums, option weights, math conditions, dependencies, and popular parameters. Do not expect ",{"type":49,"tag":69,"props":177,"children":179},{"className":178},[],[180],{"type":55,"value":181},"~\u002Ftao-core",{"type":55,"value":183}," at runtime; maintainers regenerate schemas\u002Ftemplates before packaging the skill bank.",{"type":49,"tag":85,"props":185,"children":187},{"id":186},"train-action-policy",[188],{"type":55,"value":189},"Train Action Policy",{"type":49,"tag":58,"props":191,"children":192},{},[193,195,200,202,208,210,216,218,224,226,232,234,240,242,247,248,254,256,261,262,267,269,275,277,283,285,290,292,297],{"type":55,"value":194},"This model is AutoML-enabled at the model layer. Before handling any train-stage request, read ",{"type":49,"tag":69,"props":196,"children":198},{"className":197},[],[199],{"type":55,"value":133},{"type":55,"value":201}," and resolve the run override from either an explicit ",{"type":49,"tag":69,"props":203,"children":205},{"className":204},[],[206],{"type":55,"value":207},"automl_policy",{"type":55,"value":209}," value or the user's workflow request. Use ",{"type":49,"tag":69,"props":211,"children":213},{"className":212},[],[214],{"type":55,"value":215},"automl_policy: on",{"type":55,"value":217}," by default and only expose ",{"type":49,"tag":69,"props":219,"children":221},{"className":220},[],[222],{"type":55,"value":223},"on",{"type":55,"value":225}," \u002F ",{"type":49,"tag":69,"props":227,"children":229},{"className":228},[],[230],{"type":55,"value":231},"off",{"type":55,"value":233}," in new launch prompts. Treat phrases like \"turn off AutoML\", \"disable AutoML\", \"no HPO\", or \"plain training\" as ",{"type":49,"tag":69,"props":235,"children":237},{"className":236},[],[238],{"type":55,"value":239},"automl_policy: off",{"type":55,"value":241}," for this run only. When ",{"type":49,"tag":69,"props":243,"children":245},{"className":244},[],[246],{"type":55,"value":215},{"type":55,"value":167},{"type":49,"tag":69,"props":249,"children":251},{"className":250},[],[252],{"type":55,"value":253},"automl_enabled: true",{"type":55,"value":255},", and both ",{"type":49,"tag":69,"props":257,"children":259},{"className":258},[],[260],{"type":55,"value":149},{"type":55,"value":151},{"type":49,"tag":69,"props":263,"children":265},{"className":264},[],[266],{"type":55,"value":157},{"type":55,"value":268}," are packaged, route the train action through ",{"type":49,"tag":69,"props":270,"children":272},{"className":271},[],[273],{"type":55,"value":274},"tao-skill-bank:tao-run-automl",{"type":55,"value":276}," by default with this model's ",{"type":49,"tag":69,"props":278,"children":280},{"className":279},[],[281],{"type":55,"value":282},"skill_dir",{"type":55,"value":284},". Preserve workflow\u002Fapplication overrides for datasets, specs, output directories, GPU\u002Fplatform settings, parent checkpoints, and ",{"type":49,"tag":69,"props":286,"children":288},{"className":287},[],[289],{"type":55,"value":207},{"type":55,"value":291},". Use direct model training only when ",{"type":49,"tag":69,"props":293,"children":295},{"className":294},[],[296],{"type":55,"value":239},{"type":55,"value":298}," 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":49,"tag":58,"props":300,"children":301},{},[302,304,310,311,317,318,324,326,331],{"type":55,"value":303},"Non-train actions such as ",{"type":49,"tag":69,"props":305,"children":307},{"className":306},[],[308],{"type":55,"value":309},"evaluate",{"type":55,"value":167},{"type":49,"tag":69,"props":312,"children":314},{"className":313},[],[315],{"type":55,"value":316},"inference",{"type":55,"value":167},{"type":49,"tag":69,"props":319,"children":321},{"className":320},[],[322],{"type":55,"value":323},"export",{"type":55,"value":325},", and deploy flows stay in this model skill. The per-run ",{"type":49,"tag":69,"props":327,"children":329},{"className":328},[],[330],{"type":55,"value":207},{"type":55,"value":332}," override does not change model metadata.",{"type":49,"tag":85,"props":334,"children":336},{"id":335},"training-requirements",[337],{"type":55,"value":338},"Training Requirements",{"type":49,"tag":340,"props":341,"children":342},"ul",{},[343,355,365,375],{"type":49,"tag":344,"props":345,"children":346},"li",{},[347,353],{"type":49,"tag":348,"props":349,"children":350},"strong",{},[351],{"type":55,"value":352},"Dataset type:",{"type":55,"value":354}," sparse4d",{"type":49,"tag":344,"props":356,"children":357},{},[358,363],{"type":49,"tag":348,"props":359,"children":360},{},[361],{"type":55,"value":362},"Formats:",{"type":55,"value":364}," ovpkl",{"type":49,"tag":344,"props":366,"children":367},{},[368,373],{"type":49,"tag":348,"props":369,"children":370},{},[371],{"type":55,"value":372},"Monitoring metric:",{"type":55,"value":374}," val_mAP",{"type":49,"tag":344,"props":376,"children":377},{},[378,380,386,387,393,395,401,403,408,410,415,417,422],{"type":55,"value":379},"Current TAO Sparse4D training emits this value in status\u002Flogs as\n",{"type":49,"tag":69,"props":381,"children":383},{"className":382},[],[384],{"type":55,"value":385},"img_bbox_NuScenes\u002FmAP",{"type":55,"value":151},{"type":49,"tag":69,"props":388,"children":390},{"className":389},[],[391],{"type":55,"value":392},"mAP",{"type":55,"value":394},"; AutoML metric\nextractors should treat those emitted keys as aliases for ",{"type":49,"tag":69,"props":396,"children":398},{"className":397},[],[399],{"type":55,"value":400},"val_mAP",{"type":55,"value":402},".\nMulti-fidelity AutoML algorithms such as Hyperband, ASHA, and BOHB may\npromote a checkpoint to a resume job that completes without emitting a fresh\n",{"type":49,"tag":69,"props":404,"children":406},{"className":405},[],[407],{"type":55,"value":400},{"type":55,"value":409}," alias. In that case, compare AutoML's carried metric to the source\nrung job that emitted ",{"type":49,"tag":69,"props":411,"children":413},{"className":412},[],[414],{"type":55,"value":385},{"type":55,"value":416}," or ",{"type":49,"tag":69,"props":418,"children":420},{"className":419},[],[421],{"type":55,"value":392},{"type":55,"value":423},", while still verifying\nthat the promoted job resumed from the explicit epoch\u002Fstep checkpoint,\nproduced a real checkpoint, and is usable for evaluate\u002Finference.",{"type":49,"tag":425,"props":426,"children":428},"h3",{"id":427},"per-action-dataset-requirements",[429],{"type":55,"value":430},"Per-Action Dataset Requirements",{"type":49,"tag":432,"props":433,"children":434},"table",{},[435,469],{"type":49,"tag":436,"props":437,"children":438},"thead",{},[439],{"type":49,"tag":440,"props":441,"children":442},"tr",{},[443,449,454,459,464],{"type":49,"tag":444,"props":445,"children":446},"th",{},[447],{"type":55,"value":448},"Action",{"type":49,"tag":444,"props":450,"children":451},{},[452],{"type":55,"value":453},"Spec Key",{"type":49,"tag":444,"props":455,"children":456},{},[457],{"type":55,"value":458},"Source",{"type":49,"tag":444,"props":460,"children":461},{},[462],{"type":55,"value":463},"Files",{"type":49,"tag":444,"props":465,"children":466},{},[467],{"type":55,"value":468},"List?",{"type":49,"tag":470,"props":471,"children":472},"tbody",{},[473,500,526,552,576,600,625,648,671,694,717,740,763,787,810,833,856,879,902,926,949,972,995],{"type":49,"tag":440,"props":474,"children":475},{},[476,482,487,492,495],{"type":49,"tag":477,"props":478,"children":479},"td",{},[480],{"type":55,"value":481},"dataset_convert",{"type":49,"tag":477,"props":483,"children":484},{},[485],{"type":55,"value":486},"aicity.root",{"type":49,"tag":477,"props":488,"children":489},{},[490],{"type":55,"value":491},"id",{"type":49,"tag":477,"props":493,"children":494},{},[],{"type":49,"tag":477,"props":496,"children":497},{},[498],{"type":55,"value":499},"No",{"type":49,"tag":440,"props":501,"children":502},{},[503,507,512,517,522],{"type":49,"tag":477,"props":504,"children":505},{},[506],{"type":55,"value":309},{"type":49,"tag":477,"props":508,"children":509},{},[510],{"type":55,"value":511},"dataset.data_root",{"type":49,"tag":477,"props":513,"children":514},{},[515],{"type":55,"value":516},"eval_dataset",{"type":49,"tag":477,"props":518,"children":519},{},[520],{"type":55,"value":521},"(from convert job, spec: aicity.split)",{"type":49,"tag":477,"props":523,"children":524},{},[525],{"type":55,"value":499},{"type":49,"tag":440,"props":527,"children":528},{},[529,533,538,543,548],{"type":49,"tag":477,"props":530,"children":531},{},[532],{"type":55,"value":309},{"type":49,"tag":477,"props":534,"children":535},{},[536],{"type":55,"value":537},"model.head.instance_bank.anchor",{"type":49,"tag":477,"props":539,"children":540},{},[541],{"type":55,"value":542},"train_datasets",{"type":49,"tag":477,"props":544,"children":545},{},[546],{"type":55,"value":547},"\u002Fresults\u002F{dataset_convert_job_id}\u002Fanchor_init.npy",{"type":49,"tag":477,"props":549,"children":550},{},[551],{"type":55,"value":499},{"type":49,"tag":440,"props":553,"children":554},{},[555,559,564,568,572],{"type":49,"tag":477,"props":556,"children":557},{},[558],{"type":55,"value":309},{"type":49,"tag":477,"props":560,"children":561},{},[562],{"type":55,"value":563},"dataset.train_dataset.ann_file",{"type":49,"tag":477,"props":565,"children":566},{},[567],{"type":55,"value":542},{"type":49,"tag":477,"props":569,"children":570},{},[571],{"type":55,"value":521},{"type":49,"tag":477,"props":573,"children":574},{},[575],{"type":55,"value":499},{"type":49,"tag":440,"props":577,"children":578},{},[579,583,588,592,596],{"type":49,"tag":477,"props":580,"children":581},{},[582],{"type":55,"value":309},{"type":49,"tag":477,"props":584,"children":585},{},[586],{"type":55,"value":587},"dataset.val_dataset.ann_file",{"type":49,"tag":477,"props":589,"children":590},{},[591],{"type":55,"value":516},{"type":49,"tag":477,"props":593,"children":594},{},[595],{"type":55,"value":521},{"type":49,"tag":477,"props":597,"children":598},{},[599],{"type":55,"value":499},{"type":49,"tag":440,"props":601,"children":602},{},[603,607,612,617,621],{"type":49,"tag":477,"props":604,"children":605},{},[606],{"type":55,"value":309},{"type":49,"tag":477,"props":608,"children":609},{},[610],{"type":55,"value":611},"dataset.test_dataset.ann_file",{"type":49,"tag":477,"props":613,"children":614},{},[615],{"type":55,"value":616},"inference_dataset",{"type":49,"tag":477,"props":618,"children":619},{},[620],{"type":55,"value":521},{"type":49,"tag":477,"props":622,"children":623},{},[624],{"type":55,"value":499},{"type":49,"tag":440,"props":626,"children":627},{},[628,632,636,640,644],{"type":49,"tag":477,"props":629,"children":630},{},[631],{"type":55,"value":323},{"type":49,"tag":477,"props":633,"children":634},{},[635],{"type":55,"value":537},{"type":49,"tag":477,"props":637,"children":638},{},[639],{"type":55,"value":542},{"type":49,"tag":477,"props":641,"children":642},{},[643],{"type":55,"value":547},{"type":49,"tag":477,"props":645,"children":646},{},[647],{"type":55,"value":499},{"type":49,"tag":440,"props":649,"children":650},{},[651,655,659,663,667],{"type":49,"tag":477,"props":652,"children":653},{},[654],{"type":55,"value":316},{"type":49,"tag":477,"props":656,"children":657},{},[658],{"type":55,"value":511},{"type":49,"tag":477,"props":660,"children":661},{},[662],{"type":55,"value":616},{"type":49,"tag":477,"props":664,"children":665},{},[666],{"type":55,"value":521},{"type":49,"tag":477,"props":668,"children":669},{},[670],{"type":55,"value":499},{"type":49,"tag":440,"props":672,"children":673},{},[674,678,682,686,690],{"type":49,"tag":477,"props":675,"children":676},{},[677],{"type":55,"value":316},{"type":49,"tag":477,"props":679,"children":680},{},[681],{"type":55,"value":537},{"type":49,"tag":477,"props":683,"children":684},{},[685],{"type":55,"value":542},{"type":49,"tag":477,"props":687,"children":688},{},[689],{"type":55,"value":547},{"type":49,"tag":477,"props":691,"children":692},{},[693],{"type":55,"value":499},{"type":49,"tag":440,"props":695,"children":696},{},[697,701,705,709,713],{"type":49,"tag":477,"props":698,"children":699},{},[700],{"type":55,"value":316},{"type":49,"tag":477,"props":702,"children":703},{},[704],{"type":55,"value":563},{"type":49,"tag":477,"props":706,"children":707},{},[708],{"type":55,"value":542},{"type":49,"tag":477,"props":710,"children":711},{},[712],{"type":55,"value":521},{"type":49,"tag":477,"props":714,"children":715},{},[716],{"type":55,"value":499},{"type":49,"tag":440,"props":718,"children":719},{},[720,724,728,732,736],{"type":49,"tag":477,"props":721,"children":722},{},[723],{"type":55,"value":316},{"type":49,"tag":477,"props":725,"children":726},{},[727],{"type":55,"value":587},{"type":49,"tag":477,"props":729,"children":730},{},[731],{"type":55,"value":516},{"type":49,"tag":477,"props":733,"children":734},{},[735],{"type":55,"value":521},{"type":49,"tag":477,"props":737,"children":738},{},[739],{"type":55,"value":499},{"type":49,"tag":440,"props":741,"children":742},{},[743,747,751,755,759],{"type":49,"tag":477,"props":744,"children":745},{},[746],{"type":55,"value":316},{"type":49,"tag":477,"props":748,"children":749},{},[750],{"type":55,"value":611},{"type":49,"tag":477,"props":752,"children":753},{},[754],{"type":55,"value":616},{"type":49,"tag":477,"props":756,"children":757},{},[758],{"type":55,"value":521},{"type":49,"tag":477,"props":760,"children":761},{},[762],{"type":55,"value":499},{"type":49,"tag":440,"props":764,"children":765},{},[766,771,775,779,783],{"type":49,"tag":477,"props":767,"children":768},{},[769],{"type":55,"value":770},"quantize",{"type":49,"tag":477,"props":772,"children":773},{},[774],{"type":55,"value":511},{"type":49,"tag":477,"props":776,"children":777},{},[778],{"type":55,"value":542},{"type":49,"tag":477,"props":780,"children":781},{},[782],{"type":55,"value":521},{"type":49,"tag":477,"props":784,"children":785},{},[786],{"type":55,"value":499},{"type":49,"tag":440,"props":788,"children":789},{},[790,794,798,802,806],{"type":49,"tag":477,"props":791,"children":792},{},[793],{"type":55,"value":770},{"type":49,"tag":477,"props":795,"children":796},{},[797],{"type":55,"value":537},{"type":49,"tag":477,"props":799,"children":800},{},[801],{"type":55,"value":542},{"type":49,"tag":477,"props":803,"children":804},{},[805],{"type":55,"value":547},{"type":49,"tag":477,"props":807,"children":808},{},[809],{"type":55,"value":499},{"type":49,"tag":440,"props":811,"children":812},{},[813,817,821,825,829],{"type":49,"tag":477,"props":814,"children":815},{},[816],{"type":55,"value":770},{"type":49,"tag":477,"props":818,"children":819},{},[820],{"type":55,"value":563},{"type":49,"tag":477,"props":822,"children":823},{},[824],{"type":55,"value":542},{"type":49,"tag":477,"props":826,"children":827},{},[828],{"type":55,"value":521},{"type":49,"tag":477,"props":830,"children":831},{},[832],{"type":55,"value":499},{"type":49,"tag":440,"props":834,"children":835},{},[836,840,844,848,852],{"type":49,"tag":477,"props":837,"children":838},{},[839],{"type":55,"value":770},{"type":49,"tag":477,"props":841,"children":842},{},[843],{"type":55,"value":587},{"type":49,"tag":477,"props":845,"children":846},{},[847],{"type":55,"value":516},{"type":49,"tag":477,"props":849,"children":850},{},[851],{"type":55,"value":521},{"type":49,"tag":477,"props":853,"children":854},{},[855],{"type":55,"value":499},{"type":49,"tag":440,"props":857,"children":858},{},[859,863,867,871,875],{"type":49,"tag":477,"props":860,"children":861},{},[862],{"type":55,"value":770},{"type":49,"tag":477,"props":864,"children":865},{},[866],{"type":55,"value":611},{"type":49,"tag":477,"props":868,"children":869},{},[870],{"type":55,"value":616},{"type":49,"tag":477,"props":872,"children":873},{},[874],{"type":55,"value":521},{"type":49,"tag":477,"props":876,"children":877},{},[878],{"type":55,"value":499},{"type":49,"tag":440,"props":880,"children":881},{},[882,886,891,895,898],{"type":49,"tag":477,"props":883,"children":884},{},[885],{"type":55,"value":770},{"type":49,"tag":477,"props":887,"children":888},{},[889],{"type":55,"value":890},"dataset.quant_calibration_dataset.images_dir",{"type":49,"tag":477,"props":892,"children":893},{},[894],{"type":55,"value":542},{"type":49,"tag":477,"props":896,"children":897},{},[],{"type":49,"tag":477,"props":899,"children":900},{},[901],{"type":55,"value":499},{"type":49,"tag":440,"props":903,"children":904},{},[905,910,914,918,922],{"type":49,"tag":477,"props":906,"children":907},{},[908],{"type":55,"value":909},"train",{"type":49,"tag":477,"props":911,"children":912},{},[913],{"type":55,"value":511},{"type":49,"tag":477,"props":915,"children":916},{},[917],{"type":55,"value":542},{"type":49,"tag":477,"props":919,"children":920},{},[921],{"type":55,"value":521},{"type":49,"tag":477,"props":923,"children":924},{},[925],{"type":55,"value":499},{"type":49,"tag":440,"props":927,"children":928},{},[929,933,937,941,945],{"type":49,"tag":477,"props":930,"children":931},{},[932],{"type":55,"value":909},{"type":49,"tag":477,"props":934,"children":935},{},[936],{"type":55,"value":537},{"type":49,"tag":477,"props":938,"children":939},{},[940],{"type":55,"value":542},{"type":49,"tag":477,"props":942,"children":943},{},[944],{"type":55,"value":547},{"type":49,"tag":477,"props":946,"children":947},{},[948],{"type":55,"value":499},{"type":49,"tag":440,"props":950,"children":951},{},[952,956,960,964,968],{"type":49,"tag":477,"props":953,"children":954},{},[955],{"type":55,"value":909},{"type":49,"tag":477,"props":957,"children":958},{},[959],{"type":55,"value":563},{"type":49,"tag":477,"props":961,"children":962},{},[963],{"type":55,"value":542},{"type":49,"tag":477,"props":965,"children":966},{},[967],{"type":55,"value":521},{"type":49,"tag":477,"props":969,"children":970},{},[971],{"type":55,"value":499},{"type":49,"tag":440,"props":973,"children":974},{},[975,979,983,987,991],{"type":49,"tag":477,"props":976,"children":977},{},[978],{"type":55,"value":909},{"type":49,"tag":477,"props":980,"children":981},{},[982],{"type":55,"value":587},{"type":49,"tag":477,"props":984,"children":985},{},[986],{"type":55,"value":516},{"type":49,"tag":477,"props":988,"children":989},{},[990],{"type":55,"value":521},{"type":49,"tag":477,"props":992,"children":993},{},[994],{"type":55,"value":499},{"type":49,"tag":440,"props":996,"children":997},{},[998,1002,1006,1010,1014],{"type":49,"tag":477,"props":999,"children":1000},{},[1001],{"type":55,"value":909},{"type":49,"tag":477,"props":1003,"children":1004},{},[1005],{"type":55,"value":611},{"type":49,"tag":477,"props":1007,"children":1008},{},[1009],{"type":55,"value":616},{"type":49,"tag":477,"props":1011,"children":1012},{},[1013],{"type":55,"value":521},{"type":49,"tag":477,"props":1015,"children":1016},{},[1017],{"type":55,"value":499},{"type":49,"tag":425,"props":1019,"children":1021},{"id":1020},"typical-spec-overrides",[1022],{"type":55,"value":1023},"Typical Spec Overrides",{"type":49,"tag":58,"props":1025,"children":1026},{},[1027,1029,1034,1036,1042],{"type":55,"value":1028},"Data source overrides are ",{"type":49,"tag":348,"props":1030,"children":1031},{},[1032],{"type":55,"value":1033},"mandatory for every action",{"type":55,"value":1035}," — the agent MUST construct data source paths from the Per-Action Dataset Requirements table above and include them in ",{"type":49,"tag":69,"props":1037,"children":1039},{"className":1038},[],[1040],{"type":55,"value":1041},"spec_overrides",{"type":55,"value":1043},".",{"type":49,"tag":1045,"props":1046,"children":1051},"pre",{"className":1047,"code":1048,"language":1049,"meta":1050,"style":1050},"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\"\nCONVERTED_SCENE = \"\u003Cscene-from-converter>\"  # e.g. \"subsetscene+bev-sensor-random-0\"\n","python","",[1052],{"type":49,"tag":69,"props":1053,"children":1054},{"__ignoreMap":1050},[1055,1066,1075],{"type":49,"tag":1056,"props":1057,"children":1060},"span",{"class":1058,"line":1059},"line",1,[1061],{"type":49,"tag":1056,"props":1062,"children":1063},{},[1064],{"type":55,"value":1065},"S3_TRAIN = \"s3:\u002F\u002Fbucket\u002Fdata\u002Ftrain\"\n",{"type":49,"tag":1056,"props":1067,"children":1069},{"class":1058,"line":1068},2,[1070],{"type":49,"tag":1056,"props":1071,"children":1072},{},[1073],{"type":55,"value":1074},"S3_EVAL = \"s3:\u002F\u002Fbucket\u002Fdata\u002Feval\"\n",{"type":49,"tag":1056,"props":1076,"children":1078},{"class":1058,"line":1077},3,[1079],{"type":49,"tag":1056,"props":1080,"children":1081},{},[1082],{"type":55,"value":1083},"CONVERTED_SCENE = \"\u003Cscene-from-converter>\"  # e.g. \"subsetscene+bev-sensor-random-0\"\n",{"type":49,"tag":58,"props":1085,"children":1086},{},[1087],{"type":49,"tag":348,"props":1088,"children":1089},{},[1090],{"type":55,"value":1091},"train (mandatory data sources):",{"type":49,"tag":1045,"props":1093,"children":1095},{"className":1047,"code":1094,"language":1049,"meta":1050,"style":1050},"CONVERTED = \"s3:\u002F\u002Fbucket\u002Fresults\u002F\u003Cdataset_convert_job_id>\"\n{\n    \"train.num_epochs\": 30,\n    \"train.checkpoint_interval\": 10,\n    \"train.validation_interval\": 10,\n    \"train.num_gpus\": 1,\n    \"dataset.sequences.split_num\": 90,\n    \"dataset.train_dataset.sequences_split_num\": 90,\n    \"dataset.data_root\": f\"{S3_TRAIN}\u002Ftrain\",\n    \"model.head.instance_bank.anchor\": f\"{CONVERTED}\u002Fanchor_init.npy\",\n    \"dataset.train_dataset.ann_file\": f\"{CONVERTED}\u002Ftrain\u002F{CONVERTED_SCENE}_infos_train.pkl\",\n    \"dataset.val_dataset.ann_file\": f\"{CONVERTED}\u002Fval\u002F{CONVERTED_SCENE}_infos_val.pkl\",\n    \"dataset.test_dataset.ann_file\": f\"{CONVERTED}\u002Ftest\u002F{CONVERTED_SCENE}_infos_test.pkl\",\n}\n",[1096],{"type":49,"tag":69,"props":1097,"children":1098},{"__ignoreMap":1050},[1099,1107,1115,1123,1132,1141,1150,1159,1168,1177,1186,1195,1204,1213],{"type":49,"tag":1056,"props":1100,"children":1101},{"class":1058,"line":1059},[1102],{"type":49,"tag":1056,"props":1103,"children":1104},{},[1105],{"type":55,"value":1106},"CONVERTED = \"s3:\u002F\u002Fbucket\u002Fresults\u002F\u003Cdataset_convert_job_id>\"\n",{"type":49,"tag":1056,"props":1108,"children":1109},{"class":1058,"line":1068},[1110],{"type":49,"tag":1056,"props":1111,"children":1112},{},[1113],{"type":55,"value":1114},"{\n",{"type":49,"tag":1056,"props":1116,"children":1117},{"class":1058,"line":1077},[1118],{"type":49,"tag":1056,"props":1119,"children":1120},{},[1121],{"type":55,"value":1122},"    \"train.num_epochs\": 30,\n",{"type":49,"tag":1056,"props":1124,"children":1126},{"class":1058,"line":1125},4,[1127],{"type":49,"tag":1056,"props":1128,"children":1129},{},[1130],{"type":55,"value":1131},"    \"train.checkpoint_interval\": 10,\n",{"type":49,"tag":1056,"props":1133,"children":1135},{"class":1058,"line":1134},5,[1136],{"type":49,"tag":1056,"props":1137,"children":1138},{},[1139],{"type":55,"value":1140},"    \"train.validation_interval\": 10,\n",{"type":49,"tag":1056,"props":1142,"children":1144},{"class":1058,"line":1143},6,[1145],{"type":49,"tag":1056,"props":1146,"children":1147},{},[1148],{"type":55,"value":1149},"    \"train.num_gpus\": 1,\n",{"type":49,"tag":1056,"props":1151,"children":1153},{"class":1058,"line":1152},7,[1154],{"type":49,"tag":1056,"props":1155,"children":1156},{},[1157],{"type":55,"value":1158},"    \"dataset.sequences.split_num\": 90,\n",{"type":49,"tag":1056,"props":1160,"children":1162},{"class":1058,"line":1161},8,[1163],{"type":49,"tag":1056,"props":1164,"children":1165},{},[1166],{"type":55,"value":1167},"    \"dataset.train_dataset.sequences_split_num\": 90,\n",{"type":49,"tag":1056,"props":1169,"children":1171},{"class":1058,"line":1170},9,[1172],{"type":49,"tag":1056,"props":1173,"children":1174},{},[1175],{"type":55,"value":1176},"    \"dataset.data_root\": f\"{S3_TRAIN}\u002Ftrain\",\n",{"type":49,"tag":1056,"props":1178,"children":1180},{"class":1058,"line":1179},10,[1181],{"type":49,"tag":1056,"props":1182,"children":1183},{},[1184],{"type":55,"value":1185},"    \"model.head.instance_bank.anchor\": f\"{CONVERTED}\u002Fanchor_init.npy\",\n",{"type":49,"tag":1056,"props":1187,"children":1189},{"class":1058,"line":1188},11,[1190],{"type":49,"tag":1056,"props":1191,"children":1192},{},[1193],{"type":55,"value":1194},"    \"dataset.train_dataset.ann_file\": f\"{CONVERTED}\u002Ftrain\u002F{CONVERTED_SCENE}_infos_train.pkl\",\n",{"type":49,"tag":1056,"props":1196,"children":1198},{"class":1058,"line":1197},12,[1199],{"type":49,"tag":1056,"props":1200,"children":1201},{},[1202],{"type":55,"value":1203},"    \"dataset.val_dataset.ann_file\": f\"{CONVERTED}\u002Fval\u002F{CONVERTED_SCENE}_infos_val.pkl\",\n",{"type":49,"tag":1056,"props":1205,"children":1207},{"class":1058,"line":1206},13,[1208],{"type":49,"tag":1056,"props":1209,"children":1210},{},[1211],{"type":55,"value":1212},"    \"dataset.test_dataset.ann_file\": f\"{CONVERTED}\u002Ftest\u002F{CONVERTED_SCENE}_infos_test.pkl\",\n",{"type":49,"tag":1056,"props":1214,"children":1216},{"class":1058,"line":1215},14,[1217],{"type":49,"tag":1056,"props":1218,"children":1219},{},[1220],{"type":55,"value":1221},"}\n",{"type":49,"tag":58,"props":1223,"children":1224},{},[1225],{"type":49,"tag":348,"props":1226,"children":1227},{},[1228],{"type":55,"value":1229},"evaluate (mandatory data sources):",{"type":49,"tag":1045,"props":1231,"children":1233},{"className":1047,"code":1232,"language":1049,"meta":1050,"style":1050},"CONVERTED = \"s3:\u002F\u002Fbucket\u002Fresults\u002F\u003Cdataset_convert_job_id>\"\n{\n    \"dataset.data_root\": f\"{S3_EVAL}\u002Fval\",\n    \"model.head.instance_bank.anchor\": f\"{CONVERTED}\u002Fanchor_init.npy\",\n    \"dataset.train_dataset.ann_file\": f\"{CONVERTED}\u002Ftrain\u002F{CONVERTED_SCENE}_infos_train.pkl\",\n    \"dataset.val_dataset.ann_file\": f\"{CONVERTED}\u002Fval\u002F{CONVERTED_SCENE}_infos_val.pkl\",\n    \"dataset.test_dataset.ann_file\": f\"{CONVERTED}\u002Ftest\u002F{CONVERTED_SCENE}_infos_test.pkl\",\n}\n",[1234],{"type":49,"tag":69,"props":1235,"children":1236},{"__ignoreMap":1050},[1237,1244,1251,1259,1266,1273,1280,1287],{"type":49,"tag":1056,"props":1238,"children":1239},{"class":1058,"line":1059},[1240],{"type":49,"tag":1056,"props":1241,"children":1242},{},[1243],{"type":55,"value":1106},{"type":49,"tag":1056,"props":1245,"children":1246},{"class":1058,"line":1068},[1247],{"type":49,"tag":1056,"props":1248,"children":1249},{},[1250],{"type":55,"value":1114},{"type":49,"tag":1056,"props":1252,"children":1253},{"class":1058,"line":1077},[1254],{"type":49,"tag":1056,"props":1255,"children":1256},{},[1257],{"type":55,"value":1258},"    \"dataset.data_root\": f\"{S3_EVAL}\u002Fval\",\n",{"type":49,"tag":1056,"props":1260,"children":1261},{"class":1058,"line":1125},[1262],{"type":49,"tag":1056,"props":1263,"children":1264},{},[1265],{"type":55,"value":1185},{"type":49,"tag":1056,"props":1267,"children":1268},{"class":1058,"line":1134},[1269],{"type":49,"tag":1056,"props":1270,"children":1271},{},[1272],{"type":55,"value":1194},{"type":49,"tag":1056,"props":1274,"children":1275},{"class":1058,"line":1143},[1276],{"type":49,"tag":1056,"props":1277,"children":1278},{},[1279],{"type":55,"value":1203},{"type":49,"tag":1056,"props":1281,"children":1282},{"class":1058,"line":1152},[1283],{"type":49,"tag":1056,"props":1284,"children":1285},{},[1286],{"type":55,"value":1212},{"type":49,"tag":1056,"props":1288,"children":1289},{"class":1058,"line":1161},[1290],{"type":49,"tag":1056,"props":1291,"children":1292},{},[1293],{"type":55,"value":1221},{"type":49,"tag":58,"props":1295,"children":1296},{},[1297],{"type":49,"tag":348,"props":1298,"children":1299},{},[1300],{"type":55,"value":1301},"export (mandatory data sources):",{"type":49,"tag":1045,"props":1303,"children":1305},{"className":1047,"code":1304,"language":1049,"meta":1050,"style":1050},"CONVERTED = \"s3:\u002F\u002Fbucket\u002Fresults\u002F\u003Cdataset_convert_job_id>\"\n{\n    \"model.head.instance_bank.anchor\": f\"{CONVERTED}\u002Fanchor_init.npy\",\n}\n",[1306],{"type":49,"tag":69,"props":1307,"children":1308},{"__ignoreMap":1050},[1309,1316,1323,1330],{"type":49,"tag":1056,"props":1310,"children":1311},{"class":1058,"line":1059},[1312],{"type":49,"tag":1056,"props":1313,"children":1314},{},[1315],{"type":55,"value":1106},{"type":49,"tag":1056,"props":1317,"children":1318},{"class":1058,"line":1068},[1319],{"type":49,"tag":1056,"props":1320,"children":1321},{},[1322],{"type":55,"value":1114},{"type":49,"tag":1056,"props":1324,"children":1325},{"class":1058,"line":1077},[1326],{"type":49,"tag":1056,"props":1327,"children":1328},{},[1329],{"type":55,"value":1185},{"type":49,"tag":1056,"props":1331,"children":1332},{"class":1058,"line":1125},[1333],{"type":49,"tag":1056,"props":1334,"children":1335},{},[1336],{"type":55,"value":1221},{"type":49,"tag":58,"props":1338,"children":1339},{},[1340],{"type":49,"tag":348,"props":1341,"children":1342},{},[1343],{"type":55,"value":1344},"inference (mandatory data sources):",{"type":49,"tag":1045,"props":1346,"children":1348},{"className":1047,"code":1347,"language":1049,"meta":1050,"style":1050},"CONVERTED = \"s3:\u002F\u002Fbucket\u002Fresults\u002F\u003Cdataset_convert_job_id>\"\n{\n    \"dataset.data_root\": f\"{S3_EVAL}\u002Ftest\",\n    \"model.head.instance_bank.anchor\": f\"{CONVERTED}\u002Fanchor_init.npy\",\n    \"dataset.train_dataset.ann_file\": f\"{CONVERTED}\u002Ftrain\u002F{CONVERTED_SCENE}_infos_train.pkl\",\n    \"dataset.val_dataset.ann_file\": f\"{CONVERTED}\u002Fval\u002F{CONVERTED_SCENE}_infos_val.pkl\",\n    \"dataset.test_dataset.ann_file\": f\"{CONVERTED}\u002Ftest\u002F{CONVERTED_SCENE}_infos_test.pkl\",\n}\n",[1349],{"type":49,"tag":69,"props":1350,"children":1351},{"__ignoreMap":1050},[1352,1359,1366,1374,1381,1388,1395,1402],{"type":49,"tag":1056,"props":1353,"children":1354},{"class":1058,"line":1059},[1355],{"type":49,"tag":1056,"props":1356,"children":1357},{},[1358],{"type":55,"value":1106},{"type":49,"tag":1056,"props":1360,"children":1361},{"class":1058,"line":1068},[1362],{"type":49,"tag":1056,"props":1363,"children":1364},{},[1365],{"type":55,"value":1114},{"type":49,"tag":1056,"props":1367,"children":1368},{"class":1058,"line":1077},[1369],{"type":49,"tag":1056,"props":1370,"children":1371},{},[1372],{"type":55,"value":1373},"    \"dataset.data_root\": f\"{S3_EVAL}\u002Ftest\",\n",{"type":49,"tag":1056,"props":1375,"children":1376},{"class":1058,"line":1125},[1377],{"type":49,"tag":1056,"props":1378,"children":1379},{},[1380],{"type":55,"value":1185},{"type":49,"tag":1056,"props":1382,"children":1383},{"class":1058,"line":1134},[1384],{"type":49,"tag":1056,"props":1385,"children":1386},{},[1387],{"type":55,"value":1194},{"type":49,"tag":1056,"props":1389,"children":1390},{"class":1058,"line":1143},[1391],{"type":49,"tag":1056,"props":1392,"children":1393},{},[1394],{"type":55,"value":1203},{"type":49,"tag":1056,"props":1396,"children":1397},{"class":1058,"line":1152},[1398],{"type":49,"tag":1056,"props":1399,"children":1400},{},[1401],{"type":55,"value":1212},{"type":49,"tag":1056,"props":1403,"children":1404},{"class":1058,"line":1161},[1405],{"type":49,"tag":1056,"props":1406,"children":1407},{},[1408],{"type":55,"value":1221},{"type":49,"tag":58,"props":1410,"children":1411},{},[1412],{"type":49,"tag":348,"props":1413,"children":1414},{},[1415],{"type":55,"value":1416},"quantize (mandatory data sources):",{"type":49,"tag":1045,"props":1418,"children":1420},{"className":1047,"code":1419,"language":1049,"meta":1050,"style":1050},"CONVERTED = \"s3:\u002F\u002Fbucket\u002Fresults\u002F\u003Cdataset_convert_job_id>\"\n{\n    \"dataset.data_root\": f\"{S3_TRAIN}\u002Ftrain\",\n    \"model.head.instance_bank.anchor\": f\"{CONVERTED}\u002Fanchor_init.npy\",\n    \"dataset.train_dataset.ann_file\": f\"{CONVERTED}\u002Ftrain\u002F{CONVERTED_SCENE}_infos_train.pkl\",\n    \"dataset.val_dataset.ann_file\": f\"{CONVERTED}\u002Fval\u002F{CONVERTED_SCENE}_infos_val.pkl\",\n    \"dataset.test_dataset.ann_file\": f\"{CONVERTED}\u002Ftest\u002F{CONVERTED_SCENE}_infos_test.pkl\",\n    \"dataset.quant_calibration_dataset.images_dir\": f\"{S3_TRAIN}\",\n}\n",[1421],{"type":49,"tag":69,"props":1422,"children":1423},{"__ignoreMap":1050},[1424,1431,1438,1445,1452,1459,1466,1473,1481],{"type":49,"tag":1056,"props":1425,"children":1426},{"class":1058,"line":1059},[1427],{"type":49,"tag":1056,"props":1428,"children":1429},{},[1430],{"type":55,"value":1106},{"type":49,"tag":1056,"props":1432,"children":1433},{"class":1058,"line":1068},[1434],{"type":49,"tag":1056,"props":1435,"children":1436},{},[1437],{"type":55,"value":1114},{"type":49,"tag":1056,"props":1439,"children":1440},{"class":1058,"line":1077},[1441],{"type":49,"tag":1056,"props":1442,"children":1443},{},[1444],{"type":55,"value":1176},{"type":49,"tag":1056,"props":1446,"children":1447},{"class":1058,"line":1125},[1448],{"type":49,"tag":1056,"props":1449,"children":1450},{},[1451],{"type":55,"value":1185},{"type":49,"tag":1056,"props":1453,"children":1454},{"class":1058,"line":1134},[1455],{"type":49,"tag":1056,"props":1456,"children":1457},{},[1458],{"type":55,"value":1194},{"type":49,"tag":1056,"props":1460,"children":1461},{"class":1058,"line":1143},[1462],{"type":49,"tag":1056,"props":1463,"children":1464},{},[1465],{"type":55,"value":1203},{"type":49,"tag":1056,"props":1467,"children":1468},{"class":1058,"line":1152},[1469],{"type":49,"tag":1056,"props":1470,"children":1471},{},[1472],{"type":55,"value":1212},{"type":49,"tag":1056,"props":1474,"children":1475},{"class":1058,"line":1161},[1476],{"type":49,"tag":1056,"props":1477,"children":1478},{},[1479],{"type":55,"value":1480},"    \"dataset.quant_calibration_dataset.images_dir\": f\"{S3_TRAIN}\",\n",{"type":49,"tag":1056,"props":1482,"children":1483},{"class":1058,"line":1170},[1484],{"type":49,"tag":1056,"props":1485,"children":1486},{},[1487],{"type":55,"value":1221},{"type":49,"tag":58,"props":1489,"children":1490},{},[1491,1493,1499,1501,1507],{"type":55,"value":1492},"See ",{"type":49,"tag":69,"props":1494,"children":1496},{"className":1495},[],[1497],{"type":55,"value":1498},"references\u002Flocal_docker_conversion.md",{"type":55,"value":1500}," for local-docker conversion roots and mounts, H5 depth-path normalization, converted annotation filenames, smoke-run ",{"type":49,"tag":69,"props":1502,"children":1504},{"className":1503},[],[1505],{"type":55,"value":1506},"max_num_cams",{"type":55,"value":1508},"\u002Fanchor contracts for export compatibility, and converted-artifact verification before train\u002Fevaluate\u002Finference.",{"type":49,"tag":85,"props":1510,"children":1512},{"id":1511},"eval-dataset",[1513],{"type":55,"value":1514},"Eval Dataset",{"type":49,"tag":58,"props":1516,"children":1517},{},[1518],{"type":55,"value":1519},"Optional. Val\u002Ftest splits configured via dataset ann_file paths.",{"type":49,"tag":85,"props":1521,"children":1523},{"id":1522},"important-parameters",[1524],{"type":55,"value":1525},"Important Parameters",{"type":49,"tag":340,"props":1527,"children":1528},{},[1529,1539,1549,1572,1582,1592,1602,1612,1622,1632,1642,1652,1669,1679,1689,1699,1709,1726],{"type":49,"tag":344,"props":1530,"children":1531},{},[1532,1537],{"type":49,"tag":348,"props":1533,"children":1534},{},[1535],{"type":55,"value":1536},"model.backbone",{"type":55,"value":1538},": Backbone. Default resnet_101.",{"type":49,"tag":344,"props":1540,"children":1541},{},[1542,1547],{"type":49,"tag":348,"props":1543,"children":1544},{},[1545],{"type":55,"value":1546},"model.neck.out_channels",{"type":55,"value":1548},": FPN output channels. Default 256. num_outs=4.",{"type":49,"tag":344,"props":1550,"children":1551},{},[1552,1557,1559,1564,1566,1571],{"type":49,"tag":348,"props":1553,"children":1554},{},[1555],{"type":55,"value":1556},"model.input_shape",{"type":55,"value":1558},": Input image shape ",{"type":49,"tag":1056,"props":1560,"children":1561},{},[1562],{"type":55,"value":1563},"W, H",{"type":55,"value":1565},". Default ",{"type":49,"tag":1056,"props":1567,"children":1568},{},[1569],{"type":55,"value":1570},"1408, 512",{"type":55,"value":1043},{"type":49,"tag":344,"props":1573,"children":1574},{},[1575,1580],{"type":49,"tag":348,"props":1576,"children":1577},{},[1578],{"type":55,"value":1579},"model.head.num_output",{"type":55,"value":1581},": Number of detection output queries. Default 300.",{"type":49,"tag":344,"props":1583,"children":1584},{},[1585,1590],{"type":49,"tag":348,"props":1586,"children":1587},{},[1588],{"type":55,"value":1589},"model.head.num_decoder",{"type":55,"value":1591},": Number of decoder layers. Default 6.",{"type":49,"tag":344,"props":1593,"children":1594},{},[1595,1600],{"type":49,"tag":348,"props":1596,"children":1597},{},[1598],{"type":55,"value":1599},"model.head.temporal",{"type":55,"value":1601},": Enable temporal reasoning. Default True.",{"type":49,"tag":344,"props":1603,"children":1604},{},[1605,1610],{"type":49,"tag":348,"props":1606,"children":1607},{},[1608],{"type":55,"value":1609},"model.head.instance_bank.num_anchor",{"type":55,"value":1611},": Instance bank anchors. Default 900.",{"type":49,"tag":344,"props":1613,"children":1614},{},[1615,1620],{"type":49,"tag":348,"props":1616,"children":1617},{},[1618],{"type":55,"value":1619},"model.head.instance_bank.num_temp_instances",{"type":55,"value":1621},": Temporal instance count. Default 600.",{"type":49,"tag":344,"props":1623,"children":1624},{},[1625,1630],{"type":49,"tag":348,"props":1626,"children":1627},{},[1628],{"type":55,"value":1629},"model.depth_branch.loss_weight",{"type":55,"value":1631},": Depth supervision loss weight. Default 0.2.",{"type":49,"tag":344,"props":1633,"children":1634},{},[1635,1640],{"type":49,"tag":348,"props":1636,"children":1637},{},[1638],{"type":55,"value":1639},"dataset.batch_size",{"type":55,"value":1641},": Per-GPU batch size. Default 2.",{"type":49,"tag":344,"props":1643,"children":1644},{},[1645,1650],{"type":49,"tag":348,"props":1646,"children":1647},{},[1648],{"type":55,"value":1649},"dataset.num_frames",{"type":55,"value":1651},": Sequence length. Default 200.",{"type":49,"tag":344,"props":1653,"children":1654},{},[1655,1660,1662,1667],{"type":49,"tag":348,"props":1656,"children":1657},{},[1658],{"type":55,"value":1659},"dataset.classes",{"type":55,"value":1661},": Detection classes. Default ",{"type":49,"tag":1056,"props":1663,"children":1664},{},[1665],{"type":55,"value":1666},"person, gr1_t2, agility_digit, nova_carter",{"type":55,"value":1668},". num_ids=70 for tracking.",{"type":49,"tag":344,"props":1670,"children":1671},{},[1672,1677],{"type":49,"tag":348,"props":1673,"children":1674},{},[1675],{"type":55,"value":1676},"train.optim.lr",{"type":55,"value":1678},": Learning rate. Default 5e-5. img_backbone lr_mult=0.2.",{"type":49,"tag":344,"props":1680,"children":1681},{},[1682,1687],{"type":49,"tag":348,"props":1683,"children":1684},{},[1685],{"type":55,"value":1686},"train.lr_scheduler",{"type":55,"value":1688},": Cosine scheduler with linear warmup (500 iters, ratio 0.333).",{"type":49,"tag":344,"props":1690,"children":1691},{},[1692,1697],{"type":49,"tag":348,"props":1693,"children":1694},{},[1695],{"type":55,"value":1696},"train.grad_clip.max_norm",{"type":55,"value":1698},": Gradient clipping. Default 25.",{"type":49,"tag":344,"props":1700,"children":1701},{},[1702,1707],{"type":49,"tag":348,"props":1703,"children":1704},{},[1705],{"type":55,"value":1706},"train.precision",{"type":55,"value":1708},": Options: bf16, fp16, fp32. Default bf16.",{"type":49,"tag":344,"props":1710,"children":1711},{},[1712,1717,1719,1724],{"type":49,"tag":348,"props":1713,"children":1714},{},[1715],{"type":55,"value":1716},"evaluate.metrics",{"type":55,"value":1718},": Eval metrics. Default ",{"type":49,"tag":1056,"props":1720,"children":1721},{},[1722],{"type":55,"value":1723},"\"detection\"",{"type":55,"value":1725},". Optional tracking evaluation.",{"type":49,"tag":344,"props":1727,"children":1728},{},[1729,1734],{"type":49,"tag":348,"props":1730,"children":1731},{},[1732],{"type":55,"value":1733},"evaluate.tracking.enabled",{"type":55,"value":1735},": Enable tracking evaluation. tracking_threshold=0.2.",{"type":49,"tag":85,"props":1737,"children":1739},{"id":1738},"multi-gpu-multi-node",[1740],{"type":55,"value":1741},"Multi-GPU \u002F Multi-Node",{"type":49,"tag":58,"props":1743,"children":1744},{},[1745,1750,1752,1757],{"type":49,"tag":348,"props":1746,"children":1747},{},[1748],{"type":55,"value":1749},"Launch method:",{"type":55,"value":1751}," Lightning-managed (single ",{"type":49,"tag":69,"props":1753,"children":1755},{"className":1754},[],[1756],{"type":55,"value":1049},{"type":55,"value":1758}," process, Lightning spawns workers).",{"type":49,"tag":432,"props":1760,"children":1761},{},[1762,1782],{"type":49,"tag":436,"props":1763,"children":1764},{},[1765],{"type":49,"tag":440,"props":1766,"children":1767},{},[1768,1772,1777],{"type":49,"tag":444,"props":1769,"children":1770},{},[1771],{"type":55,"value":453},{"type":49,"tag":444,"props":1773,"children":1774},{},[1775],{"type":55,"value":1776},"Description",{"type":49,"tag":444,"props":1778,"children":1779},{},[1780],{"type":55,"value":1781},"Default",{"type":49,"tag":470,"props":1783,"children":1784},{},[1785,1807,1832],{"type":49,"tag":440,"props":1786,"children":1787},{},[1788,1797,1802],{"type":49,"tag":477,"props":1789,"children":1790},{},[1791],{"type":49,"tag":69,"props":1792,"children":1794},{"className":1793},[],[1795],{"type":55,"value":1796},"train.num_gpus",{"type":49,"tag":477,"props":1798,"children":1799},{},[1800],{"type":55,"value":1801},"Number of GPUs",{"type":49,"tag":477,"props":1803,"children":1804},{},[1805],{"type":55,"value":1806},"1",{"type":49,"tag":440,"props":1808,"children":1809},{},[1810,1819,1824],{"type":49,"tag":477,"props":1811,"children":1812},{},[1813],{"type":49,"tag":69,"props":1814,"children":1816},{"className":1815},[],[1817],{"type":55,"value":1818},"train.gpu_ids",{"type":49,"tag":477,"props":1820,"children":1821},{},[1822],{"type":55,"value":1823},"GPU device indices",{"type":49,"tag":477,"props":1825,"children":1826},{},[1827],{"type":49,"tag":1056,"props":1828,"children":1829},{},[1830],{"type":55,"value":1831},"0",{"type":49,"tag":440,"props":1833,"children":1834},{},[1835,1844,1849],{"type":49,"tag":477,"props":1836,"children":1837},{},[1838],{"type":49,"tag":69,"props":1839,"children":1841},{"className":1840},[],[1842],{"type":55,"value":1843},"train.num_nodes",{"type":49,"tag":477,"props":1845,"children":1846},{},[1847],{"type":55,"value":1848},"Number of nodes",{"type":49,"tag":477,"props":1850,"children":1851},{},[1852],{"type":55,"value":1806},{"type":49,"tag":340,"props":1854,"children":1855},{},[1856,1869,1880,1891],{"type":49,"tag":344,"props":1857,"children":1858},{},[1859,1861,1867],{"type":55,"value":1860},"Multi-GPU strategy: ",{"type":49,"tag":69,"props":1862,"children":1864},{"className":1863},[],[1865],{"type":55,"value":1866},"ddp_find_unused_parameters_true",{"type":55,"value":1868}," (no fsdp support)",{"type":49,"tag":344,"props":1870,"children":1871},{},[1872,1878],{"type":49,"tag":69,"props":1873,"children":1875},{"className":1874},[],[1876],{"type":55,"value":1877},"sync_batchnorm",{"type":55,"value":1879}," is always enabled (True)",{"type":49,"tag":344,"props":1881,"children":1882},{},[1883,1885],{"type":55,"value":1884},"Iterations per epoch computed as: ",{"type":49,"tag":69,"props":1886,"children":1888},{"className":1887},[],[1889],{"type":55,"value":1890},"num_frames * num_bev_groups \u002F (num_nodes * num_gpus * batch_size)",{"type":49,"tag":344,"props":1892,"children":1893},{},[1894,1899],{"type":49,"tag":348,"props":1895,"children":1896},{},[1897],{"type":55,"value":1898},"Scaling:",{"type":55,"value":1900}," When increasing GPUs, effective batch size grows and iterations-per-epoch shrinks proportionally",{"type":49,"tag":58,"props":1902,"children":1903},{},[1904,1909,1911,1917,1918,1924,1925,1931,1932,1938,1939,1945],{"type":49,"tag":348,"props":1905,"children":1906},{},[1907],{"type":55,"value":1908},"Multi-node env vars",{"type":55,"value":1910}," (set by orchestrator): ",{"type":49,"tag":69,"props":1912,"children":1914},{"className":1913},[],[1915],{"type":55,"value":1916},"WORLD_SIZE",{"type":55,"value":167},{"type":49,"tag":69,"props":1919,"children":1921},{"className":1920},[],[1922],{"type":55,"value":1923},"NODE_RANK",{"type":55,"value":167},{"type":49,"tag":69,"props":1926,"children":1928},{"className":1927},[],[1929],{"type":55,"value":1930},"MASTER_ADDR",{"type":55,"value":167},{"type":49,"tag":69,"props":1933,"children":1935},{"className":1934},[],[1936],{"type":55,"value":1937},"MASTER_PORT",{"type":55,"value":167},{"type":49,"tag":69,"props":1940,"children":1942},{"className":1941},[],[1943],{"type":55,"value":1944},"NUM_GPU_PER_NODE",{"type":55,"value":1043},{"type":49,"tag":85,"props":1947,"children":1949},{"id":1948},"hardware",[1950],{"type":55,"value":1951},"Hardware",{"type":49,"tag":58,"props":1953,"children":1954},{},[1955],{"type":55,"value":1956},"Minimum 2 GPU(s), recommended 8 GPU(s). 40GB+ (A100 recommended) VRAM per GPU. Multi-camera temporal model is memory intensive. bf16 required for practical training. Multi-GPU strongly recommended. Instance bank requires substantial memory for temporal reasoning.",{"type":49,"tag":85,"props":1958,"children":1960},{"id":1959},"error-patterns",[1961],{"type":55,"value":1962},"Error Patterns",{"type":49,"tag":58,"props":1964,"children":1965},{},[1966,1971],{"type":49,"tag":348,"props":1967,"children":1968},{},[1969],{"type":55,"value":1970},"dataset_convert required",{"type":55,"value":1972},": Must run dataset_convert first to produce annotation pickles and anchor_init.npy.",{"type":49,"tag":58,"props":1974,"children":1975},{},[1976,1981,1983,1988,1990,1996,1998,2004,2006,2011],{"type":49,"tag":348,"props":1977,"children":1978},{},[1979],{"type":55,"value":1980},"dataset_convert container\u002Fcommand",{"type":55,"value":1982},": Sparse4D conversion is an AICity to\nOVPKL annotations conversion. Launch ",{"type":49,"tag":69,"props":1984,"children":1986},{"className":1985},[],[1987],{"type":55,"value":481},{"type":55,"value":1989}," with the action-level\n",{"type":49,"tag":69,"props":1991,"children":1993},{"className":1992},[],[1994],{"type":55,"value":1995},"tao_toolkit.data_services",{"type":55,"value":1997}," image and ",{"type":49,"tag":69,"props":1999,"children":2001},{"className":2000},[],[2002],{"type":55,"value":2003},"annotations convert -e {config_path}",{"type":55,"value":2005},";\ndo not use the PyTorch ",{"type":49,"tag":69,"props":2007,"children":2009},{"className":2008},[],[2010],{"type":55,"value":52},{"type":55,"value":2012}," CLI for conversion. Train\u002Fevaluate\u002Fexport\u002F\ninference still use the model-level PyTorch image.",{"type":49,"tag":58,"props":2014,"children":2015},{},[2016,2021,2023,2028,2030,2036,2038,2043,2045,2051,2053,2059],{"type":49,"tag":348,"props":2017,"children":2018},{},[2019],{"type":55,"value":2020},"Stable raw-data path",{"type":55,"value":2022},": The AICity to OVPKL converter writes image paths into\nthe generated pickle files. Keep ",{"type":49,"tag":69,"props":2024,"children":2026},{"className":2025},[],[2027],{"type":55,"value":486},{"type":55,"value":2029}," at ",{"type":49,"tag":69,"props":2031,"children":2033},{"className":2032},[],[2034],{"type":55,"value":2035},"\u002Fdata\u002Faicity_root",{"type":55,"value":2037}," during\nconversion, then point ",{"type":49,"tag":69,"props":2039,"children":2041},{"className":2040},[],[2042],{"type":55,"value":511},{"type":55,"value":2044}," at the split folder, for example\n",{"type":49,"tag":69,"props":2046,"children":2048},{"className":2047},[],[2049],{"type":55,"value":2050},"\u002Fdata\u002Faicity_root\u002Ftrain",{"type":55,"value":2052}," for training or ",{"type":49,"tag":69,"props":2054,"children":2056},{"className":2055},[],[2057],{"type":55,"value":2058},"\u002Fdata\u002Faicity_root\u002Fval",{"type":55,"value":2060}," for\nevaluation. This preserves the converter's absolute RGB paths and relative\ndepth paths.",{"type":49,"tag":58,"props":2062,"children":2063},{},[2064,2069,2071,2077,2079,2085,2087,2092,2094,2100,2102,2108],{"type":49,"tag":348,"props":2065,"children":2066},{},[2067],{"type":55,"value":2068},"H5 depth tuple mismatch",{"type":55,"value":2070},": If training fails with an H5 path error where the\ntrainer tries to open a camera directory such as\n",{"type":49,"tag":69,"props":2072,"children":2074},{"className":2073},[],[2075],{"type":55,"value":2076},"\u002Fdata\u002Faicity_root\u002Ftrain\u002F\u003Cscene>\u002FCamera",{"type":55,"value":2078},", run\n",{"type":49,"tag":69,"props":2080,"children":2082},{"className":2081},[],[2083],{"type":55,"value":2084},"models\u002Fsparse4d\u002Fscripts\u002Fnormalize_depth_paths.py --data-root \u003Chost-aicity-root>\u002Ftrain \u003Cconverted-ann-dir>",{"type":55,"value":2086},"\nafter ",{"type":49,"tag":69,"props":2088,"children":2090},{"className":2089},[],[2091],{"type":55,"value":481},{"type":55,"value":2093}," and before train\u002Fevaluate\u002Finference. The helper rewrites\nconverted ",{"type":49,"tag":69,"props":2095,"children":2097},{"className":2096},[],[2098],{"type":55,"value":2099},"depth_map_path",{"type":55,"value":2101}," tuples to point at\n",{"type":49,"tag":69,"props":2103,"children":2105},{"className":2104},[],[2106],{"type":55,"value":2107},"\u003Cscene>\u002Fdepth_maps\u002F\u003Ccamera>.h5",{"type":55,"value":2109}," with the H5 dataset key basename.",{"type":49,"tag":58,"props":2111,"children":2112},{},[2113,2118],{"type":49,"tag":348,"props":2114,"children":2115},{},[2116],{"type":55,"value":2117},"Missing anchor file",{"type":55,"value":2119},": Set model.head.instance_bank.anchor to the anchor_init.npy path from dataset_convert results.",{"type":49,"tag":58,"props":2121,"children":2122},{},[2123,2128],{"type":49,"tag":348,"props":2124,"children":2125},{},[2126],{"type":55,"value":2127},"Temporal OOM",{"type":55,"value":2129},": Reduce dataset.num_frames or dataset.batch_size if running out of memory during temporal training.",{"type":49,"tag":58,"props":2131,"children":2132},{},[2133,2138,2140,2146,2148,2154,2156,2162,2164,2169],{"type":49,"tag":348,"props":2134,"children":2135},{},[2136],{"type":55,"value":2137},"Quantize image compatibility",{"type":55,"value":2139},": The model-skill wiring should pass\n",{"type":49,"tag":69,"props":2141,"children":2143},{"className":2142},[],[2144],{"type":55,"value":2145},"quantize.model_path",{"type":55,"value":2147}," through the parent-model resolver, and checkpoint handoff\nshould select the exact epoch\u002Fstep checkpoint just like evaluate, inference,\nexport, and resume. TorchAO checkpoint quantization passes in the\n",{"type":49,"tag":69,"props":2149,"children":2151},{"className":2150},[],[2152],{"type":55,"value":2153},"validation-fixes-20260525",{"type":55,"value":2155}," PyT image and writes\n",{"type":49,"tag":69,"props":2157,"children":2159},{"className":2158},[],[2160],{"type":55,"value":2161},"quantized_model_torchao.pth",{"type":55,"value":2163},". Older 7.0.0-rc PyT images may fail inside the\nSparse4D quantize entrypoint or lack ONNX quantization dependencies; do not\nremove or skip the advertised ",{"type":49,"tag":69,"props":2165,"children":2167},{"className":2166},[],[2168],{"type":55,"value":770},{"type":55,"value":2170}," action if that occurs. Report the\ncontainer\u002Fimage failure and keep the exact checkpoint path visible.",{"type":49,"tag":85,"props":2172,"children":2174},{"id":2173},"spec-param-parent-model-inference",[2175],{"type":55,"value":2176},"Spec Param \u002F Parent Model Inference",{"type":49,"tag":58,"props":2178,"children":2179},{},[2180,2181,2187,2189,2195,2197,2203,2205,2211,2213,2219],{"type":55,"value":1492},{"type":49,"tag":69,"props":2182,"children":2184},{"className":2183},[],[2185],{"type":55,"value":2186},"references\u002Fspec_param_inference.md",{"type":55,"value":2188}," for the model-specific inference mappings from TAO Core ",{"type":49,"tag":69,"props":2190,"children":2192},{"className":2191},[],[2193],{"type":55,"value":2194},"sparse4d.config.json",{"type":55,"value":2196}," (the per-action spec-field to inference-function table) and the ",{"type":49,"tag":69,"props":2198,"children":2200},{"className":2199},[],[2201],{"type":55,"value":2202},"parent_model",{"type":55,"value":2204},"\u002F",{"type":49,"tag":69,"props":2206,"children":2208},{"className":2207},[],[2209],{"type":55,"value":2210},"parent_job_id",{"type":55,"value":2212}," checkpoint-resolution rules that generated runners apply with SDK helpers before ",{"type":49,"tag":69,"props":2214,"children":2216},{"className":2215},[],[2217],{"type":55,"value":2218},"create_job()",{"type":55,"value":1043},{"type":49,"tag":2221,"props":2222,"children":2223},"style",{},[2224],{"type":55,"value":2225},"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":2227,"total":2385},[2228,2246,2263,2274,2286,2300,2313,2327,2340,2351,2365,2374],{"slug":2229,"name":2229,"fn":2230,"description":2231,"org":2232,"tags":2233,"stars":2243,"repoUrl":2244,"updatedAt":2245},"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},[2234,2237,2240],{"name":2235,"slug":2236,"type":15},"Documentation","documentation",{"name":2238,"slug":2239,"type":15},"MCP","mcp",{"name":2241,"slug":2242,"type":15},"Search","search",21777,"https:\u002F\u002Fgithub.com\u002FNVIDIA\u002FNemoClaw","2026-07-20T06:00:01.461044",{"slug":2247,"name":2247,"fn":2248,"description":2249,"org":2250,"tags":2251,"stars":2260,"repoUrl":2261,"updatedAt":2262},"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},[2252,2255,2258],{"name":2253,"slug":2254,"type":15},"Containers","containers",{"name":2256,"slug":2257,"type":15},"Deployment","deployment",{"name":2259,"slug":1049,"type":15},"Python",17049,"https:\u002F\u002Fgithub.com\u002FNVIDIA\u002FMegatron-LM","2026-07-27T06:06:11.249662",{"slug":2264,"name":2264,"fn":2265,"description":2266,"org":2267,"tags":2268,"stars":2260,"repoUrl":2261,"updatedAt":2273},"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},[2269,2272],{"name":2270,"slug":2271,"type":15},"CI\u002FCD","ci-cd",{"name":2256,"slug":2257,"type":15},"2026-07-14T05:25:59.97109",{"slug":2275,"name":2275,"fn":2276,"description":2277,"org":2278,"tags":2279,"stars":2260,"repoUrl":2261,"updatedAt":2285},"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},[2280,2281,2282],{"name":2270,"slug":2271,"type":15},{"name":2256,"slug":2257,"type":15},{"name":2283,"slug":2284,"type":15},"GitHub","github","2026-07-27T06:06:12.278222",{"slug":2287,"name":2287,"fn":2288,"description":2289,"org":2290,"tags":2291,"stars":2260,"repoUrl":2261,"updatedAt":2299},"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},[2292,2295,2296],{"name":2293,"slug":2294,"type":15},"Debugging","debugging",{"name":2283,"slug":2284,"type":15},{"name":2297,"slug":2298,"type":15},"Triage","triage","2026-07-14T05:25:57.442089",{"slug":2301,"name":2301,"fn":2302,"description":2303,"org":2304,"tags":2305,"stars":2260,"repoUrl":2261,"updatedAt":2312},"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},[2306,2309],{"name":2307,"slug":2308,"type":15},"Best Practices","best-practices",{"name":2310,"slug":2311,"type":15},"Code Analysis","code-analysis","2026-07-14T05:25:56.18433",{"slug":2314,"name":2314,"fn":2315,"description":2316,"org":2317,"tags":2318,"stars":2260,"repoUrl":2261,"updatedAt":2326},"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},[2319,2322,2325],{"name":2320,"slug":2321,"type":15},"Machine Learning","machine-learning",{"name":2323,"slug":2324,"type":15},"Migration","migration",{"name":9,"slug":8,"type":15},"2026-07-17T06:07:11.777011",{"slug":2328,"name":2328,"fn":2329,"description":2330,"org":2331,"tags":2332,"stars":2260,"repoUrl":2261,"updatedAt":2339},"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},[2333,2336],{"name":2334,"slug":2335,"type":15},"QA","qa",{"name":2337,"slug":2338,"type":15},"Testing","testing","2026-07-14T05:25:53.673039",{"slug":2341,"name":2341,"fn":2342,"description":2343,"org":2344,"tags":2345,"stars":2260,"repoUrl":2261,"updatedAt":2350},"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},[2346,2347],{"name":2256,"slug":2257,"type":15},{"name":2348,"slug":2349,"type":15},"Infrastructure","infrastructure","2026-07-14T05:25:49.362534",{"slug":2352,"name":2352,"fn":2353,"description":2354,"org":2355,"tags":2356,"stars":2260,"repoUrl":2261,"updatedAt":2364},"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},[2357,2360,2361],{"name":2358,"slug":2359,"type":15},"Code Review","code-review",{"name":2283,"slug":2284,"type":15},{"name":2362,"slug":2363,"type":15},"Pull Requests","pull-requests","2026-07-14T05:26:01.226578",{"slug":2366,"name":2366,"fn":2367,"description":2368,"org":2369,"tags":2370,"stars":2260,"repoUrl":2261,"updatedAt":2373},"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},[2371,2372],{"name":2334,"slug":2335,"type":15},{"name":2337,"slug":2338,"type":15},"2026-07-14T05:25:54.928983",{"slug":2375,"name":2375,"fn":2376,"description":2377,"org":2378,"tags":2379,"stars":2260,"repoUrl":2261,"updatedAt":2384},"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},[2380,2383],{"name":2381,"slug":2382,"type":15},"Automation","automation",{"name":2270,"slug":2271,"type":15},"2026-07-30T05:29:03.275638",496,{"items":2387,"total":2483},[2388,2405,2415,2429,2439,2454,2469],{"slug":2389,"name":2389,"fn":2390,"description":2391,"org":2392,"tags":2393,"stars":23,"repoUrl":24,"updatedAt":2404},"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},[2394,2397,2400,2401],{"name":2395,"slug":2396,"type":15},"Data Analysis","data-analysis",{"name":2398,"slug":2399,"type":15},"Data Engineering","data-engineering",{"name":9,"slug":8,"type":15},{"name":2402,"slug":2403,"type":15},"Performance","performance","2026-07-14T05:28:43.176466",{"slug":2406,"name":2406,"fn":2407,"description":2408,"org":2409,"tags":2410,"stars":23,"repoUrl":24,"updatedAt":2414},"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},[2411,2412,2413],{"name":2256,"slug":2257,"type":15},{"name":2348,"slug":2349,"type":15},{"name":9,"slug":8,"type":15},"2026-07-14T05:29:06.667109",{"slug":2416,"name":2416,"fn":2417,"description":2418,"org":2419,"tags":2420,"stars":23,"repoUrl":24,"updatedAt":2428},"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},[2421,2424,2425],{"name":2422,"slug":2423,"type":15},"Agents","agents",{"name":9,"slug":8,"type":15},{"name":2426,"slug":2427,"type":15},"Research","research","2026-07-14T05:28:06.816956",{"slug":2430,"name":2430,"fn":2431,"description":2432,"org":2433,"tags":2434,"stars":23,"repoUrl":24,"updatedAt":2438},"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},[2435,2436,2437],{"name":2395,"slug":2396,"type":15},{"name":9,"slug":8,"type":15},{"name":2337,"slug":2338,"type":15},"2026-07-17T05:29:03.913266",{"slug":2440,"name":2440,"fn":2441,"description":2442,"org":2443,"tags":2444,"stars":23,"repoUrl":24,"updatedAt":2453},"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},[2445,2446,2449,2450],{"name":2381,"slug":2382,"type":15},{"name":2447,"slug":2448,"type":15},"Imaging","imaging",{"name":9,"slug":8,"type":15},{"name":2451,"slug":2452,"type":15},"Video","video","2026-07-17T05:28:53.905004",{"slug":2455,"name":2455,"fn":2456,"description":2457,"org":2458,"tags":2459,"stars":23,"repoUrl":24,"updatedAt":2468},"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},[2460,2461,2464,2465],{"name":2256,"slug":2257,"type":15},{"name":2462,"slug":2463,"type":15},"Docker","docker",{"name":9,"slug":8,"type":15},{"name":2466,"slug":2467,"type":15},"Operations","operations","2026-07-17T05:28:56.913999",{"slug":2470,"name":2470,"fn":2471,"description":2472,"org":2473,"tags":2474,"stars":23,"repoUrl":24,"updatedAt":2482},"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},[2475,2476,2479],{"name":9,"slug":8,"type":15},{"name":2477,"slug":2478,"type":15},"Quantum Computing","quantum-computing",{"name":2480,"slug":2481,"type":15},"Simulation","simulation","2026-07-14T05:26:58.898253",305]