[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill-nvidia-tao-train-segformer":3,"mdc-st6glj-key":34,"related-org-nvidia-tao-train-segformer":2643,"related-repo-nvidia-tao-train-segformer":2799},{"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-segformer","train SegFormer models for semantic segmentation","SegFormer for semantic segmentation. Lightweight transformer-based architecture with hierarchical feature extraction, efficient for real-time segmentation tasks. Use when training, evaluating, exporting, quantizing, or running inference for a TAO SegFormer model. Trigger phrases include \"train SegFormer\", \"semantic segmentation\", \"lightweight transformer segmenter\", \"real-time semantic segmentation\".",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},"nvidia","NVIDIA","https:\u002F\u002Fpexgzepcugksgbtrxkhf.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Forg-logos\u002Fnvidia.png",[12,16,19,20],{"name":13,"slug":14,"type":15},"Deep Learning","deep-learning","tag",{"name":17,"slug":18,"type":15},"Machine Learning","machine-learning",{"name":9,"slug":8,"type":15},{"name":21,"slug":22,"type":15},"Computer Vision","computer-vision",2473,"https:\u002F\u002Fgithub.com\u002FNVIDIA\u002Fskills","2026-07-14T05:26:39.021372","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-segformer","---\nname: tao-train-segformer\ndescription: SegFormer for semantic segmentation. Lightweight transformer-based architecture with hierarchical feature\n  extraction, efficient for real-time segmentation tasks. Use when training, evaluating, exporting, quantizing, or running\n  inference for a TAO SegFormer model. Trigger phrases include \"train SegFormer\", \"semantic segmentation\", \"lightweight\n  transformer segmenter\", \"real-time semantic segmentation\".\nlicense: Apache-2.0\ncompatibility: Requires docker + nvidia-container-toolkit.\nmetadata:\n  version: \"0.1.0\"\n  author: NVIDIA Corporation\nallowed-tools: Read Bash\ntags:\n- segmentation\n---\n\n# SegFormer\n\nSegFormer for semantic segmentation. Lightweight transformer-based architecture with hierarchical feature extraction. Efficient for real-time segmentation tasks.\n\nSet model.backbone.pretrained_backbone_path for backbone weights.\n\nFor TAO Deploy TensorRT actions (`gen_trt_engine`, TensorRT `evaluate`, and TensorRT `inference`), read `references\u002Ftao-deploy-segformer.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## Supported Actions\n\nThe packaged SegFormer PyT CLI supports `train`, `evaluate`, `export`, `inference`, `quantize`, and `default_specs`. This model skill exposes `train`, `evaluate`, `export`, `inference`, and `quantize`; resume\u002Fretrain is performed through `train` with `train.resume_training_checkpoint_path`.\n\nThe parent PyT CLI does not expose `gen_trt_engine`. Use `models\u002Fsegformer\u002Fdeploy` for TensorRT engine generation, TensorRT evaluation, and TensorRT inference.\n\n## Training Requirements\n\n- **Dataset type:** segmentation\n- **Formats:** unet\n- **Monitoring metric:** val_miou, maximize\n\n### Per-Action Dataset Requirements\n\n| Action | Spec Key | Source | Files | List? |\n|---|---|---|---|---|\n| evaluate | dataset.segment.root_dir | eval_dataset | extracted root containing `images\u002F\u003Csplit>` and `masks\u002F\u003Csplit>` | No |\n| export | dataset.segment.root_dir | train_datasets | extracted root containing `images\u002F\u003Csplit>` and `masks\u002F\u003Csplit>` | No |\n| inference | dataset.segment.root_dir | inference_dataset | extracted root containing `images\u002F\u003Csplit>` and `masks\u002F\u003Csplit>` | No |\n| quantize | dataset.segment.root_dir | train_datasets | extracted root containing `images\u002F\u003Csplit>` and `masks\u002F\u003Csplit>` | No |\n| quantize | dataset.segment.quant_calibration_dataset.images_dir | calibration_dataset | extracted image directory | No |\n| train | dataset.segment.root_dir | train_datasets | extracted root containing `images\u002F\u003Csplit>` and `masks\u002F\u003Csplit>` | 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\nSEG_TRAIN_ROOT = \"\u002Fdata\u002Fsegformer\u002Ftrain\"\nSEG_EVAL_ROOT = \"\u002Fdata\u002Fsegformer\u002Feval\"\nSEG_INFER_ROOT = \"\u002Fdata\u002Fsegformer\u002Finfer\"\nCAL_IMAGES = f\"{SEG_TRAIN_ROOT}\u002Fimages\u002Ftrain\"\n```\n\n**train (mandatory data sources):**\n```python\n{\n    \"train.num_gpus\": 1,\n    \"train.num_epochs\": 10,\n    \"train.checkpoint_interval\": 10,\n    \"train.validation_interval\": 10,\n    \"dataset.segment.batch_size\": 4,\n    \"dataset.segment.root_dir\": SEG_TRAIN_ROOT,\n}\n```\n\n**evaluate (mandatory data sources):**\n```python\n{\n    \"evaluate.batch_size\": 4,\n    \"dataset.segment.root_dir\": SEG_EVAL_ROOT,\n    \"evaluate.checkpoint\": CHECKPOINT,\n}\n```\n\n**inference (mandatory data sources):**\n```python\n{\n    \"dataset.segment.batch_size\": 1,\n    \"dataset.segment.root_dir\": SEG_INFER_ROOT,\n    \"inference.checkpoint\": CHECKPOINT,\n}\n```\n\n**export (mandatory data sources):**\n```python\n{\n    \"dataset.segment.root_dir\": SEG_TRAIN_ROOT,\n    \"export.checkpoint\": CHECKPOINT,\n    \"export.input_height\": 256,\n    \"export.input_width\": 256,\n    \"export.onnx_file\": ONNX_FILE,\n}\n```\n\n**quantize (mandatory data sources):**\n```python\n{\n    \"dataset.segment.root_dir\": SEG_TRAIN_ROOT,\n    \"dataset.segment.quant_calibration_dataset.images_dir\": CAL_IMAGES,\n    \"quantize.model_path\": CHECKPOINT,\n}\n```\n\nIf the source dataset is delivered as separate `images\u002F*.tar.gz` and\n`masks\u002F*.tar.gz` archives, extract them before launch so `root_dir` contains\ndirectories such as `images\u002Ftrain`, `images\u002Fval`, `images\u002Ftest`, `masks\u002Ftrain`,\nand `masks\u002Fval`. Do not point `dataset.segment.root_dir` at an archive staging\nfolder that still contains only tarballs.\n## Eval Dataset\n\nOptional. Validation data is typically part of the root_dir structure.\n\n## Important Parameters\n\n- **dataset.segment.num_classes**: Number of segmentation classes. Default 2 (binary). Must match the number of classes in your mask annotations.\n- **model.backbone.type**: Default fan_small_12_p4_hybrid. Supported includes FAN variants, SegFormer MIT variants, and others.\n- **dataset.segment.root_dir**: Root directory of the segmentation dataset.\n- **dataset.segment.img_size**: Input image size. Default 256. Increase for finer segmentation at the cost of memory.\n- **train.optim.lr**: Learning rate. Default 6e-5.\n- **model.freeze_backbone**: Whether to freeze the backbone during training. Useful for fine-tuning with limited data.\n- **dataset.segment.batch_size**: Per-GPU batch size. Default 8.\n- **dataset.segment.label_transform**: Use the string `\"None\"` when no label\n  transform is desired. Do not set this to JSON\u002FYAML null; strict schema merge\n  treats the field as a string enum.\n- **dataset.segment.palette**: For grayscale masks, use one integer per RGB\n  entry, for example `rgb: [85]`. Preserve the dataset's actual label ids and\n  class names rather than normalizing them unless the user explicitly asks for a\n  conversion.\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.sync_batchnorm` | Sync BN across GPUs | configurable |\n| `train.use_distributed_sampler` | Use distributed sampler | configurable |\n\n- Multi-GPU strategy: `ddp_find_unused_parameters_true`\n- No fsdp support\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 1 GPU(s), recommended 2 GPU(s). 16GB+ (V100 or A100) VRAM per GPU. SegFormer is relatively lightweight. Default img_size=256 is memory-friendly. Increase img_size for higher resolution at the cost of memory and speed.\n\n## Error Patterns\n\n**CUDA out of memory**: Reduce batch_size or img_size. SegFormer memory scales quadratically with image size.\n\n**num_classes mismatch**: Ensure dataset.segment.num_classes matches the actual number of classes in your mask annotations.\n\n**TensorBoard unsupported for segmentation training**: Keep `train.tensorboard.enabled: false`. The SegFormer training entrypoint asserts that TensorBoard visualization is not supported for segmentation, so do not enable TensorBoard just to extract AutoML metrics; use log parsing or a post-train evaluator instead.\n\n**AutoML metric extraction**: SegFormer train status files report `val_miou` alongside `val_loss`, `val_acc`, and other validation KPIs. Default AutoML train launches must optimize `val_miou` with `direction: maximize`; do not optimize `val_loss` for default model invocations.\n\nFor AutoML or long segmentation sweeps, read `val_miou` from\n`results_dir\u002Ftrain\u002Fstatus.json` first. If the wrapper reports a terminal\nfailure but the structured status file reached the configured training budget\nand contains finite `val_miou`, report the recovered metric with the wrapper\nfailure noted instead of discarding the measurement.\n\nFor high-resolution custom segmentation targets, keep dataset paths as per-run\ninputs. Do not add customer\u002Fuser-specific roots to this reusable skill. When the\nuser asks for a fixed full-budget search, remember that bracket algorithms\n(`asha`, `bohb`, `dehb`, `hyperband`, `hyperband_es`, `pbt`) may intentionally\nlower `train.num_epochs` for some recommendations; use Bayesian\u002FBFBO or lock the\nbudget if every recommendation must run the full epoch count.\n\n**Checkpoint handoff**: For evaluate\u002Fexport\u002Finference\u002Fquantize\u002Fresume, use the checkpoint resolver on the best AutoML child job's `results_dir\u002Ftrain\u002F` folder and select the action-appropriate `model_epoch_*.pth` checkpoint, such as `model_epoch_000_step_00010.pth`. SegFormer may also write `segformer_model_latest.pth`, but that should only be used when a caller explicitly requests latest. Preserve `dataset.segment.num_classes`, `dataset.segment.img_size`, and `dataset.segment.root_dir` overrides for downstream actions.\n\n**Resume\u002Fretrain checkpoint**: Resume uses `train.resume_training_checkpoint_path`.\nPass the exact resolved checkpoint from the previous train output, not a guessed\n`model.pth` path. A resumed one-epoch run should produce the next checkpoint in\nthe new results directory, for example `model_epoch_001_step_00020.pth`.\n\n**Export \u002F TensorRT shape alignment**: Keep `export.input_height` and\n`export.input_width` aligned with `dataset.segment.img_size` unless the trained\nmodel and deploy specs have been validated at another resolution. The packaged\nfresh-install path is validated at `256x256`, matching the default SegFormer\ndataset and deploy templates.\n\n**Parent `segformer gen_trt_engine` rejected by the PyT CLI**: In the validated 7.0.0 PyT container, `segformer gen_trt_engine` is not a valid parent-model subtask. Use the SegFormer deploy workflow (`references\u002Ftao-deploy-segformer.md`) for TensorRT engine generation, TensorRT evaluation, and TensorRT inference.\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 `segformer.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.backbone.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\n## Deployment\n\n- [tao-deploy-segformer](references\u002Ftao-deploy-segformer.md)\n",{"data":35,"body":43},{"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],"segmentation",{"type":44,"children":45},"root",[46,55,61,66,120,127,220,226,335,367,373,465,485,491,527,534,781,787,806,855,863,938,946,991,999,1044,1052,1112,1120,1164,1237,1243,1248,1254,1362,1368,1385,1522,1541,1586,1592,1597,1603,1613,1623,1641,1695,1722,1778,1839,1871,1910,1942,1948,1977,1990,2584,2619,2625,2637],{"type":47,"tag":48,"props":49,"children":51},"element","h1",{"id":50},"segformer",[52],{"type":53,"value":54},"text","SegFormer",{"type":47,"tag":56,"props":57,"children":58},"p",{},[59],{"type":53,"value":60},"SegFormer for semantic segmentation. Lightweight transformer-based architecture with hierarchical feature extraction. Efficient for real-time segmentation tasks.",{"type":47,"tag":56,"props":62,"children":63},{},[64],{"type":53,"value":65},"Set model.backbone.pretrained_backbone_path for backbone weights.",{"type":47,"tag":56,"props":67,"children":68},{},[69,71,78,80,86,88,94,96,102,104,110,112,118],{"type":53,"value":70},"For TAO Deploy TensorRT actions (",{"type":47,"tag":72,"props":73,"children":75},"code",{"className":74},[],[76],{"type":53,"value":77},"gen_trt_engine",{"type":53,"value":79},", TensorRT ",{"type":47,"tag":72,"props":81,"children":83},{"className":82},[],[84],{"type":53,"value":85},"evaluate",{"type":53,"value":87},", and TensorRT ",{"type":47,"tag":72,"props":89,"children":91},{"className":90},[],[92],{"type":53,"value":93},"inference",{"type":53,"value":95},"), read ",{"type":47,"tag":72,"props":97,"children":99},{"className":98},[],[100],{"type":53,"value":101},"references\u002Ftao-deploy-segformer.md",{"type":53,"value":103}," first. Deploy spec templates live in this skill's ",{"type":47,"tag":72,"props":105,"children":107},{"className":106},[],[108],{"type":53,"value":109},"references\u002F",{"type":53,"value":111}," folder with the ",{"type":47,"tag":72,"props":113,"children":115},{"className":114},[],[116],{"type":53,"value":117},"spec_template_deploy_*.yaml",{"type":53,"value":119}," prefix.",{"type":47,"tag":121,"props":122,"children":124},"h2",{"id":123},"dataclass-schemas",[125],{"type":53,"value":126},"Dataclass Schemas",{"type":47,"tag":56,"props":128,"children":129},{},[130,132,138,140,146,148,154,156,162,164,170,172,178,180,186,188,194,196,202,204,210,212,218],{"type":53,"value":131},"Generated TAO Core schemas are packaged in ",{"type":47,"tag":72,"props":133,"children":135},{"className":134},[],[136],{"type":53,"value":137},"schemas\u002F\u003Caction>.schema.json",{"type":53,"value":139},", with ",{"type":47,"tag":72,"props":141,"children":143},{"className":142},[],[144],{"type":53,"value":145},"schemas\u002Fmanifest.json",{"type":53,"value":147}," listing available actions. Each generated schema also emits ",{"type":47,"tag":72,"props":149,"children":151},{"className":150},[],[152],{"type":53,"value":153},"references\u002Fspec_template_\u003Caction>.yaml",{"type":53,"value":155}," from the schema top-level ",{"type":47,"tag":72,"props":157,"children":159},{"className":158},[],[160],{"type":53,"value":161},"default",{"type":53,"value":163}," field. AutoML enablement is declared at the model layer in ",{"type":47,"tag":72,"props":165,"children":167},{"className":166},[],[168],{"type":53,"value":169},"references\u002Fskill_info.yaml",{"type":53,"value":171}," via ",{"type":47,"tag":72,"props":173,"children":175},{"className":174},[],[176],{"type":53,"value":177},"automl_enabled",{"type":53,"value":179},". Runnable AutoML still requires ",{"type":47,"tag":72,"props":181,"children":183},{"className":182},[],[184],{"type":53,"value":185},"schemas\u002Ftrain.schema.json",{"type":53,"value":187}," and ",{"type":47,"tag":72,"props":189,"children":191},{"className":190},[],[192],{"type":53,"value":193},"references\u002Fspec_template_train.yaml",{"type":53,"value":195}," to exist and parse. Use the packaged train schema for ",{"type":47,"tag":72,"props":197,"children":199},{"className":198},[],[200],{"type":53,"value":201},"automl_default_parameters",{"type":53,"value":203},", ",{"type":47,"tag":72,"props":205,"children":207},{"className":206},[],[208],{"type":53,"value":209},"automl_disabled_parameters",{"type":53,"value":211},", defaults, min\u002Fmax bounds, enums, option weights, math conditions, dependencies, and popular parameters. Do not expect ",{"type":47,"tag":72,"props":213,"children":215},{"className":214},[],[216],{"type":53,"value":217},"~\u002Ftao-core",{"type":53,"value":219}," at runtime; maintainers regenerate schemas\u002Ftemplates before packaging the skill bank.",{"type":47,"tag":121,"props":221,"children":223},{"id":222},"train-action-policy",[224],{"type":53,"value":225},"Train Action Policy",{"type":47,"tag":56,"props":227,"children":228},{},[229,231,236,238,244,246,252,254,260,262,268,270,276,278,283,284,290,292,297,298,303,305,311,313,319,321,326,328,333],{"type":53,"value":230},"This model is AutoML-enabled at the model layer. Before handling any train-stage request, read ",{"type":47,"tag":72,"props":232,"children":234},{"className":233},[],[235],{"type":53,"value":169},{"type":53,"value":237}," and resolve the run override from either an explicit ",{"type":47,"tag":72,"props":239,"children":241},{"className":240},[],[242],{"type":53,"value":243},"automl_policy",{"type":53,"value":245}," value or the user's workflow request. Use ",{"type":47,"tag":72,"props":247,"children":249},{"className":248},[],[250],{"type":53,"value":251},"automl_policy: on",{"type":53,"value":253}," by default and only expose ",{"type":47,"tag":72,"props":255,"children":257},{"className":256},[],[258],{"type":53,"value":259},"on",{"type":53,"value":261}," \u002F ",{"type":47,"tag":72,"props":263,"children":265},{"className":264},[],[266],{"type":53,"value":267},"off",{"type":53,"value":269}," in new launch prompts. Treat phrases like \"turn off AutoML\", \"disable AutoML\", \"no HPO\", or \"plain training\" as ",{"type":47,"tag":72,"props":271,"children":273},{"className":272},[],[274],{"type":53,"value":275},"automl_policy: off",{"type":53,"value":277}," for this run only. When ",{"type":47,"tag":72,"props":279,"children":281},{"className":280},[],[282],{"type":53,"value":251},{"type":53,"value":203},{"type":47,"tag":72,"props":285,"children":287},{"className":286},[],[288],{"type":53,"value":289},"automl_enabled: true",{"type":53,"value":291},", and both ",{"type":47,"tag":72,"props":293,"children":295},{"className":294},[],[296],{"type":53,"value":185},{"type":53,"value":187},{"type":47,"tag":72,"props":299,"children":301},{"className":300},[],[302],{"type":53,"value":193},{"type":53,"value":304}," are packaged, route the train action through ",{"type":47,"tag":72,"props":306,"children":308},{"className":307},[],[309],{"type":53,"value":310},"tao-skill-bank:tao-run-automl",{"type":53,"value":312}," by default with this model's ",{"type":47,"tag":72,"props":314,"children":316},{"className":315},[],[317],{"type":53,"value":318},"skill_dir",{"type":53,"value":320},". Preserve workflow\u002Fapplication overrides for datasets, specs, output directories, GPU\u002Fplatform settings, parent checkpoints, and ",{"type":47,"tag":72,"props":322,"children":324},{"className":323},[],[325],{"type":53,"value":243},{"type":53,"value":327},". Use direct model training only when ",{"type":47,"tag":72,"props":329,"children":331},{"className":330},[],[332],{"type":53,"value":275},{"type":53,"value":334}," 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":47,"tag":56,"props":336,"children":337},{},[338,340,345,346,351,352,358,360,365],{"type":53,"value":339},"Non-train actions such as ",{"type":47,"tag":72,"props":341,"children":343},{"className":342},[],[344],{"type":53,"value":85},{"type":53,"value":203},{"type":47,"tag":72,"props":347,"children":349},{"className":348},[],[350],{"type":53,"value":93},{"type":53,"value":203},{"type":47,"tag":72,"props":353,"children":355},{"className":354},[],[356],{"type":53,"value":357},"export",{"type":53,"value":359},", and deploy flows stay in this model skill. The per-run ",{"type":47,"tag":72,"props":361,"children":363},{"className":362},[],[364],{"type":53,"value":243},{"type":53,"value":366}," override does not change model metadata.",{"type":47,"tag":121,"props":368,"children":370},{"id":369},"supported-actions",[371],{"type":53,"value":372},"Supported Actions",{"type":47,"tag":56,"props":374,"children":375},{},[376,378,384,385,390,391,396,397,402,403,409,411,417,419,424,425,430,431,436,437,442,443,448,450,455,457,463],{"type":53,"value":377},"The packaged SegFormer PyT CLI supports ",{"type":47,"tag":72,"props":379,"children":381},{"className":380},[],[382],{"type":53,"value":383},"train",{"type":53,"value":203},{"type":47,"tag":72,"props":386,"children":388},{"className":387},[],[389],{"type":53,"value":85},{"type":53,"value":203},{"type":47,"tag":72,"props":392,"children":394},{"className":393},[],[395],{"type":53,"value":357},{"type":53,"value":203},{"type":47,"tag":72,"props":398,"children":400},{"className":399},[],[401],{"type":53,"value":93},{"type":53,"value":203},{"type":47,"tag":72,"props":404,"children":406},{"className":405},[],[407],{"type":53,"value":408},"quantize",{"type":53,"value":410},", and ",{"type":47,"tag":72,"props":412,"children":414},{"className":413},[],[415],{"type":53,"value":416},"default_specs",{"type":53,"value":418},". This model skill exposes ",{"type":47,"tag":72,"props":420,"children":422},{"className":421},[],[423],{"type":53,"value":383},{"type":53,"value":203},{"type":47,"tag":72,"props":426,"children":428},{"className":427},[],[429],{"type":53,"value":85},{"type":53,"value":203},{"type":47,"tag":72,"props":432,"children":434},{"className":433},[],[435],{"type":53,"value":357},{"type":53,"value":203},{"type":47,"tag":72,"props":438,"children":440},{"className":439},[],[441],{"type":53,"value":93},{"type":53,"value":410},{"type":47,"tag":72,"props":444,"children":446},{"className":445},[],[447],{"type":53,"value":408},{"type":53,"value":449},"; resume\u002Fretrain is performed through ",{"type":47,"tag":72,"props":451,"children":453},{"className":452},[],[454],{"type":53,"value":383},{"type":53,"value":456}," with ",{"type":47,"tag":72,"props":458,"children":460},{"className":459},[],[461],{"type":53,"value":462},"train.resume_training_checkpoint_path",{"type":53,"value":464},".",{"type":47,"tag":56,"props":466,"children":467},{},[468,470,475,477,483],{"type":53,"value":469},"The parent PyT CLI does not expose ",{"type":47,"tag":72,"props":471,"children":473},{"className":472},[],[474],{"type":53,"value":77},{"type":53,"value":476},". Use ",{"type":47,"tag":72,"props":478,"children":480},{"className":479},[],[481],{"type":53,"value":482},"models\u002Fsegformer\u002Fdeploy",{"type":53,"value":484}," for TensorRT engine generation, TensorRT evaluation, and TensorRT inference.",{"type":47,"tag":121,"props":486,"children":488},{"id":487},"training-requirements",[489],{"type":53,"value":490},"Training Requirements",{"type":47,"tag":492,"props":493,"children":494},"ul",{},[495,507,517],{"type":47,"tag":496,"props":497,"children":498},"li",{},[499,505],{"type":47,"tag":500,"props":501,"children":502},"strong",{},[503],{"type":53,"value":504},"Dataset type:",{"type":53,"value":506}," segmentation",{"type":47,"tag":496,"props":508,"children":509},{},[510,515],{"type":47,"tag":500,"props":511,"children":512},{},[513],{"type":53,"value":514},"Formats:",{"type":53,"value":516}," unet",{"type":47,"tag":496,"props":518,"children":519},{},[520,525],{"type":47,"tag":500,"props":521,"children":522},{},[523],{"type":53,"value":524},"Monitoring metric:",{"type":53,"value":526}," val_miou, maximize",{"type":47,"tag":528,"props":529,"children":531},"h3",{"id":530},"per-action-dataset-requirements",[532],{"type":53,"value":533},"Per-Action Dataset Requirements",{"type":47,"tag":535,"props":536,"children":537},"table",{},[538,572],{"type":47,"tag":539,"props":540,"children":541},"thead",{},[542],{"type":47,"tag":543,"props":544,"children":545},"tr",{},[546,552,557,562,567],{"type":47,"tag":547,"props":548,"children":549},"th",{},[550],{"type":53,"value":551},"Action",{"type":47,"tag":547,"props":553,"children":554},{},[555],{"type":53,"value":556},"Spec Key",{"type":47,"tag":547,"props":558,"children":559},{},[560],{"type":53,"value":561},"Source",{"type":47,"tag":547,"props":563,"children":564},{},[565],{"type":53,"value":566},"Files",{"type":47,"tag":547,"props":568,"children":569},{},[570],{"type":53,"value":571},"List?",{"type":47,"tag":573,"props":574,"children":575},"tbody",{},[576,617,652,687,721,747],{"type":47,"tag":543,"props":577,"children":578},{},[579,584,589,594,612],{"type":47,"tag":580,"props":581,"children":582},"td",{},[583],{"type":53,"value":85},{"type":47,"tag":580,"props":585,"children":586},{},[587],{"type":53,"value":588},"dataset.segment.root_dir",{"type":47,"tag":580,"props":590,"children":591},{},[592],{"type":53,"value":593},"eval_dataset",{"type":47,"tag":580,"props":595,"children":596},{},[597,599,605,606],{"type":53,"value":598},"extracted root containing ",{"type":47,"tag":72,"props":600,"children":602},{"className":601},[],[603],{"type":53,"value":604},"images\u002F\u003Csplit>",{"type":53,"value":187},{"type":47,"tag":72,"props":607,"children":609},{"className":608},[],[610],{"type":53,"value":611},"masks\u002F\u003Csplit>",{"type":47,"tag":580,"props":613,"children":614},{},[615],{"type":53,"value":616},"No",{"type":47,"tag":543,"props":618,"children":619},{},[620,624,628,633,648],{"type":47,"tag":580,"props":621,"children":622},{},[623],{"type":53,"value":357},{"type":47,"tag":580,"props":625,"children":626},{},[627],{"type":53,"value":588},{"type":47,"tag":580,"props":629,"children":630},{},[631],{"type":53,"value":632},"train_datasets",{"type":47,"tag":580,"props":634,"children":635},{},[636,637,642,643],{"type":53,"value":598},{"type":47,"tag":72,"props":638,"children":640},{"className":639},[],[641],{"type":53,"value":604},{"type":53,"value":187},{"type":47,"tag":72,"props":644,"children":646},{"className":645},[],[647],{"type":53,"value":611},{"type":47,"tag":580,"props":649,"children":650},{},[651],{"type":53,"value":616},{"type":47,"tag":543,"props":653,"children":654},{},[655,659,663,668,683],{"type":47,"tag":580,"props":656,"children":657},{},[658],{"type":53,"value":93},{"type":47,"tag":580,"props":660,"children":661},{},[662],{"type":53,"value":588},{"type":47,"tag":580,"props":664,"children":665},{},[666],{"type":53,"value":667},"inference_dataset",{"type":47,"tag":580,"props":669,"children":670},{},[671,672,677,678],{"type":53,"value":598},{"type":47,"tag":72,"props":673,"children":675},{"className":674},[],[676],{"type":53,"value":604},{"type":53,"value":187},{"type":47,"tag":72,"props":679,"children":681},{"className":680},[],[682],{"type":53,"value":611},{"type":47,"tag":580,"props":684,"children":685},{},[686],{"type":53,"value":616},{"type":47,"tag":543,"props":688,"children":689},{},[690,694,698,702,717],{"type":47,"tag":580,"props":691,"children":692},{},[693],{"type":53,"value":408},{"type":47,"tag":580,"props":695,"children":696},{},[697],{"type":53,"value":588},{"type":47,"tag":580,"props":699,"children":700},{},[701],{"type":53,"value":632},{"type":47,"tag":580,"props":703,"children":704},{},[705,706,711,712],{"type":53,"value":598},{"type":47,"tag":72,"props":707,"children":709},{"className":708},[],[710],{"type":53,"value":604},{"type":53,"value":187},{"type":47,"tag":72,"props":713,"children":715},{"className":714},[],[716],{"type":53,"value":611},{"type":47,"tag":580,"props":718,"children":719},{},[720],{"type":53,"value":616},{"type":47,"tag":543,"props":722,"children":723},{},[724,728,733,738,743],{"type":47,"tag":580,"props":725,"children":726},{},[727],{"type":53,"value":408},{"type":47,"tag":580,"props":729,"children":730},{},[731],{"type":53,"value":732},"dataset.segment.quant_calibration_dataset.images_dir",{"type":47,"tag":580,"props":734,"children":735},{},[736],{"type":53,"value":737},"calibration_dataset",{"type":47,"tag":580,"props":739,"children":740},{},[741],{"type":53,"value":742},"extracted image directory",{"type":47,"tag":580,"props":744,"children":745},{},[746],{"type":53,"value":616},{"type":47,"tag":543,"props":748,"children":749},{},[750,754,758,762,777],{"type":47,"tag":580,"props":751,"children":752},{},[753],{"type":53,"value":383},{"type":47,"tag":580,"props":755,"children":756},{},[757],{"type":53,"value":588},{"type":47,"tag":580,"props":759,"children":760},{},[761],{"type":53,"value":632},{"type":47,"tag":580,"props":763,"children":764},{},[765,766,771,772],{"type":53,"value":598},{"type":47,"tag":72,"props":767,"children":769},{"className":768},[],[770],{"type":53,"value":604},{"type":53,"value":187},{"type":47,"tag":72,"props":773,"children":775},{"className":774},[],[776],{"type":53,"value":611},{"type":47,"tag":580,"props":778,"children":779},{},[780],{"type":53,"value":616},{"type":47,"tag":528,"props":782,"children":784},{"id":783},"typical-spec-overrides",[785],{"type":53,"value":786},"Typical Spec Overrides",{"type":47,"tag":56,"props":788,"children":789},{},[790,792,797,799,805],{"type":53,"value":791},"Data source overrides are ",{"type":47,"tag":500,"props":793,"children":794},{},[795],{"type":53,"value":796},"mandatory for every action",{"type":53,"value":798}," — the agent MUST construct data source paths from the Per-Action Dataset Requirements table above and include them in ",{"type":47,"tag":72,"props":800,"children":802},{"className":801},[],[803],{"type":53,"value":804},"spec_overrides",{"type":53,"value":464},{"type":47,"tag":807,"props":808,"children":813},"pre",{"className":809,"code":810,"language":811,"meta":812,"style":812},"language-python shiki shiki-themes material-theme-lighter material-theme material-theme-palenight","SEG_TRAIN_ROOT = \"\u002Fdata\u002Fsegformer\u002Ftrain\"\nSEG_EVAL_ROOT = \"\u002Fdata\u002Fsegformer\u002Feval\"\nSEG_INFER_ROOT = \"\u002Fdata\u002Fsegformer\u002Finfer\"\nCAL_IMAGES = f\"{SEG_TRAIN_ROOT}\u002Fimages\u002Ftrain\"\n","python","",[814],{"type":47,"tag":72,"props":815,"children":816},{"__ignoreMap":812},[817,828,837,846],{"type":47,"tag":818,"props":819,"children":822},"span",{"class":820,"line":821},"line",1,[823],{"type":47,"tag":818,"props":824,"children":825},{},[826],{"type":53,"value":827},"SEG_TRAIN_ROOT = \"\u002Fdata\u002Fsegformer\u002Ftrain\"\n",{"type":47,"tag":818,"props":829,"children":831},{"class":820,"line":830},2,[832],{"type":47,"tag":818,"props":833,"children":834},{},[835],{"type":53,"value":836},"SEG_EVAL_ROOT = \"\u002Fdata\u002Fsegformer\u002Feval\"\n",{"type":47,"tag":818,"props":838,"children":840},{"class":820,"line":839},3,[841],{"type":47,"tag":818,"props":842,"children":843},{},[844],{"type":53,"value":845},"SEG_INFER_ROOT = \"\u002Fdata\u002Fsegformer\u002Finfer\"\n",{"type":47,"tag":818,"props":847,"children":849},{"class":820,"line":848},4,[850],{"type":47,"tag":818,"props":851,"children":852},{},[853],{"type":53,"value":854},"CAL_IMAGES = f\"{SEG_TRAIN_ROOT}\u002Fimages\u002Ftrain\"\n",{"type":47,"tag":56,"props":856,"children":857},{},[858],{"type":47,"tag":500,"props":859,"children":860},{},[861],{"type":53,"value":862},"train (mandatory data sources):",{"type":47,"tag":807,"props":864,"children":866},{"className":809,"code":865,"language":811,"meta":812,"style":812},"{\n    \"train.num_gpus\": 1,\n    \"train.num_epochs\": 10,\n    \"train.checkpoint_interval\": 10,\n    \"train.validation_interval\": 10,\n    \"dataset.segment.batch_size\": 4,\n    \"dataset.segment.root_dir\": SEG_TRAIN_ROOT,\n}\n",[867],{"type":47,"tag":72,"props":868,"children":869},{"__ignoreMap":812},[870,878,886,894,902,911,920,929],{"type":47,"tag":818,"props":871,"children":872},{"class":820,"line":821},[873],{"type":47,"tag":818,"props":874,"children":875},{},[876],{"type":53,"value":877},"{\n",{"type":47,"tag":818,"props":879,"children":880},{"class":820,"line":830},[881],{"type":47,"tag":818,"props":882,"children":883},{},[884],{"type":53,"value":885},"    \"train.num_gpus\": 1,\n",{"type":47,"tag":818,"props":887,"children":888},{"class":820,"line":839},[889],{"type":47,"tag":818,"props":890,"children":891},{},[892],{"type":53,"value":893},"    \"train.num_epochs\": 10,\n",{"type":47,"tag":818,"props":895,"children":896},{"class":820,"line":848},[897],{"type":47,"tag":818,"props":898,"children":899},{},[900],{"type":53,"value":901},"    \"train.checkpoint_interval\": 10,\n",{"type":47,"tag":818,"props":903,"children":905},{"class":820,"line":904},5,[906],{"type":47,"tag":818,"props":907,"children":908},{},[909],{"type":53,"value":910},"    \"train.validation_interval\": 10,\n",{"type":47,"tag":818,"props":912,"children":914},{"class":820,"line":913},6,[915],{"type":47,"tag":818,"props":916,"children":917},{},[918],{"type":53,"value":919},"    \"dataset.segment.batch_size\": 4,\n",{"type":47,"tag":818,"props":921,"children":923},{"class":820,"line":922},7,[924],{"type":47,"tag":818,"props":925,"children":926},{},[927],{"type":53,"value":928},"    \"dataset.segment.root_dir\": SEG_TRAIN_ROOT,\n",{"type":47,"tag":818,"props":930,"children":932},{"class":820,"line":931},8,[933],{"type":47,"tag":818,"props":934,"children":935},{},[936],{"type":53,"value":937},"}\n",{"type":47,"tag":56,"props":939,"children":940},{},[941],{"type":47,"tag":500,"props":942,"children":943},{},[944],{"type":53,"value":945},"evaluate (mandatory data sources):",{"type":47,"tag":807,"props":947,"children":949},{"className":809,"code":948,"language":811,"meta":812,"style":812},"{\n    \"evaluate.batch_size\": 4,\n    \"dataset.segment.root_dir\": SEG_EVAL_ROOT,\n    \"evaluate.checkpoint\": CHECKPOINT,\n}\n",[950],{"type":47,"tag":72,"props":951,"children":952},{"__ignoreMap":812},[953,960,968,976,984],{"type":47,"tag":818,"props":954,"children":955},{"class":820,"line":821},[956],{"type":47,"tag":818,"props":957,"children":958},{},[959],{"type":53,"value":877},{"type":47,"tag":818,"props":961,"children":962},{"class":820,"line":830},[963],{"type":47,"tag":818,"props":964,"children":965},{},[966],{"type":53,"value":967},"    \"evaluate.batch_size\": 4,\n",{"type":47,"tag":818,"props":969,"children":970},{"class":820,"line":839},[971],{"type":47,"tag":818,"props":972,"children":973},{},[974],{"type":53,"value":975},"    \"dataset.segment.root_dir\": SEG_EVAL_ROOT,\n",{"type":47,"tag":818,"props":977,"children":978},{"class":820,"line":848},[979],{"type":47,"tag":818,"props":980,"children":981},{},[982],{"type":53,"value":983},"    \"evaluate.checkpoint\": CHECKPOINT,\n",{"type":47,"tag":818,"props":985,"children":986},{"class":820,"line":904},[987],{"type":47,"tag":818,"props":988,"children":989},{},[990],{"type":53,"value":937},{"type":47,"tag":56,"props":992,"children":993},{},[994],{"type":47,"tag":500,"props":995,"children":996},{},[997],{"type":53,"value":998},"inference (mandatory data sources):",{"type":47,"tag":807,"props":1000,"children":1002},{"className":809,"code":1001,"language":811,"meta":812,"style":812},"{\n    \"dataset.segment.batch_size\": 1,\n    \"dataset.segment.root_dir\": SEG_INFER_ROOT,\n    \"inference.checkpoint\": CHECKPOINT,\n}\n",[1003],{"type":47,"tag":72,"props":1004,"children":1005},{"__ignoreMap":812},[1006,1013,1021,1029,1037],{"type":47,"tag":818,"props":1007,"children":1008},{"class":820,"line":821},[1009],{"type":47,"tag":818,"props":1010,"children":1011},{},[1012],{"type":53,"value":877},{"type":47,"tag":818,"props":1014,"children":1015},{"class":820,"line":830},[1016],{"type":47,"tag":818,"props":1017,"children":1018},{},[1019],{"type":53,"value":1020},"    \"dataset.segment.batch_size\": 1,\n",{"type":47,"tag":818,"props":1022,"children":1023},{"class":820,"line":839},[1024],{"type":47,"tag":818,"props":1025,"children":1026},{},[1027],{"type":53,"value":1028},"    \"dataset.segment.root_dir\": SEG_INFER_ROOT,\n",{"type":47,"tag":818,"props":1030,"children":1031},{"class":820,"line":848},[1032],{"type":47,"tag":818,"props":1033,"children":1034},{},[1035],{"type":53,"value":1036},"    \"inference.checkpoint\": CHECKPOINT,\n",{"type":47,"tag":818,"props":1038,"children":1039},{"class":820,"line":904},[1040],{"type":47,"tag":818,"props":1041,"children":1042},{},[1043],{"type":53,"value":937},{"type":47,"tag":56,"props":1045,"children":1046},{},[1047],{"type":47,"tag":500,"props":1048,"children":1049},{},[1050],{"type":53,"value":1051},"export (mandatory data sources):",{"type":47,"tag":807,"props":1053,"children":1055},{"className":809,"code":1054,"language":811,"meta":812,"style":812},"{\n    \"dataset.segment.root_dir\": SEG_TRAIN_ROOT,\n    \"export.checkpoint\": CHECKPOINT,\n    \"export.input_height\": 256,\n    \"export.input_width\": 256,\n    \"export.onnx_file\": ONNX_FILE,\n}\n",[1056],{"type":47,"tag":72,"props":1057,"children":1058},{"__ignoreMap":812},[1059,1066,1073,1081,1089,1097,1105],{"type":47,"tag":818,"props":1060,"children":1061},{"class":820,"line":821},[1062],{"type":47,"tag":818,"props":1063,"children":1064},{},[1065],{"type":53,"value":877},{"type":47,"tag":818,"props":1067,"children":1068},{"class":820,"line":830},[1069],{"type":47,"tag":818,"props":1070,"children":1071},{},[1072],{"type":53,"value":928},{"type":47,"tag":818,"props":1074,"children":1075},{"class":820,"line":839},[1076],{"type":47,"tag":818,"props":1077,"children":1078},{},[1079],{"type":53,"value":1080},"    \"export.checkpoint\": CHECKPOINT,\n",{"type":47,"tag":818,"props":1082,"children":1083},{"class":820,"line":848},[1084],{"type":47,"tag":818,"props":1085,"children":1086},{},[1087],{"type":53,"value":1088},"    \"export.input_height\": 256,\n",{"type":47,"tag":818,"props":1090,"children":1091},{"class":820,"line":904},[1092],{"type":47,"tag":818,"props":1093,"children":1094},{},[1095],{"type":53,"value":1096},"    \"export.input_width\": 256,\n",{"type":47,"tag":818,"props":1098,"children":1099},{"class":820,"line":913},[1100],{"type":47,"tag":818,"props":1101,"children":1102},{},[1103],{"type":53,"value":1104},"    \"export.onnx_file\": ONNX_FILE,\n",{"type":47,"tag":818,"props":1106,"children":1107},{"class":820,"line":922},[1108],{"type":47,"tag":818,"props":1109,"children":1110},{},[1111],{"type":53,"value":937},{"type":47,"tag":56,"props":1113,"children":1114},{},[1115],{"type":47,"tag":500,"props":1116,"children":1117},{},[1118],{"type":53,"value":1119},"quantize (mandatory data sources):",{"type":47,"tag":807,"props":1121,"children":1123},{"className":809,"code":1122,"language":811,"meta":812,"style":812},"{\n    \"dataset.segment.root_dir\": SEG_TRAIN_ROOT,\n    \"dataset.segment.quant_calibration_dataset.images_dir\": CAL_IMAGES,\n    \"quantize.model_path\": CHECKPOINT,\n}\n",[1124],{"type":47,"tag":72,"props":1125,"children":1126},{"__ignoreMap":812},[1127,1134,1141,1149,1157],{"type":47,"tag":818,"props":1128,"children":1129},{"class":820,"line":821},[1130],{"type":47,"tag":818,"props":1131,"children":1132},{},[1133],{"type":53,"value":877},{"type":47,"tag":818,"props":1135,"children":1136},{"class":820,"line":830},[1137],{"type":47,"tag":818,"props":1138,"children":1139},{},[1140],{"type":53,"value":928},{"type":47,"tag":818,"props":1142,"children":1143},{"class":820,"line":839},[1144],{"type":47,"tag":818,"props":1145,"children":1146},{},[1147],{"type":53,"value":1148},"    \"dataset.segment.quant_calibration_dataset.images_dir\": CAL_IMAGES,\n",{"type":47,"tag":818,"props":1150,"children":1151},{"class":820,"line":848},[1152],{"type":47,"tag":818,"props":1153,"children":1154},{},[1155],{"type":53,"value":1156},"    \"quantize.model_path\": CHECKPOINT,\n",{"type":47,"tag":818,"props":1158,"children":1159},{"class":820,"line":904},[1160],{"type":47,"tag":818,"props":1161,"children":1162},{},[1163],{"type":53,"value":937},{"type":47,"tag":56,"props":1165,"children":1166},{},[1167,1169,1175,1177,1183,1185,1191,1193,1199,1200,1206,1207,1213,1214,1220,1222,1228,1230,1235],{"type":53,"value":1168},"If the source dataset is delivered as separate ",{"type":47,"tag":72,"props":1170,"children":1172},{"className":1171},[],[1173],{"type":53,"value":1174},"images\u002F*.tar.gz",{"type":53,"value":1176}," and\n",{"type":47,"tag":72,"props":1178,"children":1180},{"className":1179},[],[1181],{"type":53,"value":1182},"masks\u002F*.tar.gz",{"type":53,"value":1184}," archives, extract them before launch so ",{"type":47,"tag":72,"props":1186,"children":1188},{"className":1187},[],[1189],{"type":53,"value":1190},"root_dir",{"type":53,"value":1192}," contains\ndirectories such as ",{"type":47,"tag":72,"props":1194,"children":1196},{"className":1195},[],[1197],{"type":53,"value":1198},"images\u002Ftrain",{"type":53,"value":203},{"type":47,"tag":72,"props":1201,"children":1203},{"className":1202},[],[1204],{"type":53,"value":1205},"images\u002Fval",{"type":53,"value":203},{"type":47,"tag":72,"props":1208,"children":1210},{"className":1209},[],[1211],{"type":53,"value":1212},"images\u002Ftest",{"type":53,"value":203},{"type":47,"tag":72,"props":1215,"children":1217},{"className":1216},[],[1218],{"type":53,"value":1219},"masks\u002Ftrain",{"type":53,"value":1221},",\nand ",{"type":47,"tag":72,"props":1223,"children":1225},{"className":1224},[],[1226],{"type":53,"value":1227},"masks\u002Fval",{"type":53,"value":1229},". Do not point ",{"type":47,"tag":72,"props":1231,"children":1233},{"className":1232},[],[1234],{"type":53,"value":588},{"type":53,"value":1236}," at an archive staging\nfolder that still contains only tarballs.",{"type":47,"tag":121,"props":1238,"children":1240},{"id":1239},"eval-dataset",[1241],{"type":53,"value":1242},"Eval Dataset",{"type":47,"tag":56,"props":1244,"children":1245},{},[1246],{"type":53,"value":1247},"Optional. Validation data is typically part of the root_dir structure.",{"type":47,"tag":121,"props":1249,"children":1251},{"id":1250},"important-parameters",[1252],{"type":53,"value":1253},"Important Parameters",{"type":47,"tag":492,"props":1255,"children":1256},{},[1257,1267,1277,1286,1296,1306,1316,1326,1344],{"type":47,"tag":496,"props":1258,"children":1259},{},[1260,1265],{"type":47,"tag":500,"props":1261,"children":1262},{},[1263],{"type":53,"value":1264},"dataset.segment.num_classes",{"type":53,"value":1266},": Number of segmentation classes. Default 2 (binary). Must match the number of classes in your mask annotations.",{"type":47,"tag":496,"props":1268,"children":1269},{},[1270,1275],{"type":47,"tag":500,"props":1271,"children":1272},{},[1273],{"type":53,"value":1274},"model.backbone.type",{"type":53,"value":1276},": Default fan_small_12_p4_hybrid. Supported includes FAN variants, SegFormer MIT variants, and others.",{"type":47,"tag":496,"props":1278,"children":1279},{},[1280,1284],{"type":47,"tag":500,"props":1281,"children":1282},{},[1283],{"type":53,"value":588},{"type":53,"value":1285},": Root directory of the segmentation dataset.",{"type":47,"tag":496,"props":1287,"children":1288},{},[1289,1294],{"type":47,"tag":500,"props":1290,"children":1291},{},[1292],{"type":53,"value":1293},"dataset.segment.img_size",{"type":53,"value":1295},": Input image size. Default 256. Increase for finer segmentation at the cost of memory.",{"type":47,"tag":496,"props":1297,"children":1298},{},[1299,1304],{"type":47,"tag":500,"props":1300,"children":1301},{},[1302],{"type":53,"value":1303},"train.optim.lr",{"type":53,"value":1305},": Learning rate. Default 6e-5.",{"type":47,"tag":496,"props":1307,"children":1308},{},[1309,1314],{"type":47,"tag":500,"props":1310,"children":1311},{},[1312],{"type":53,"value":1313},"model.freeze_backbone",{"type":53,"value":1315},": Whether to freeze the backbone during training. Useful for fine-tuning with limited data.",{"type":47,"tag":496,"props":1317,"children":1318},{},[1319,1324],{"type":47,"tag":500,"props":1320,"children":1321},{},[1322],{"type":53,"value":1323},"dataset.segment.batch_size",{"type":53,"value":1325},": Per-GPU batch size. Default 8.",{"type":47,"tag":496,"props":1327,"children":1328},{},[1329,1334,1336,1342],{"type":47,"tag":500,"props":1330,"children":1331},{},[1332],{"type":53,"value":1333},"dataset.segment.label_transform",{"type":53,"value":1335},": Use the string ",{"type":47,"tag":72,"props":1337,"children":1339},{"className":1338},[],[1340],{"type":53,"value":1341},"\"None\"",{"type":53,"value":1343}," when no label\ntransform is desired. Do not set this to JSON\u002FYAML null; strict schema merge\ntreats the field as a string enum.",{"type":47,"tag":496,"props":1345,"children":1346},{},[1347,1352,1354,1360],{"type":47,"tag":500,"props":1348,"children":1349},{},[1350],{"type":53,"value":1351},"dataset.segment.palette",{"type":53,"value":1353},": For grayscale masks, use one integer per RGB\nentry, for example ",{"type":47,"tag":72,"props":1355,"children":1357},{"className":1356},[],[1358],{"type":53,"value":1359},"rgb: [85]",{"type":53,"value":1361},". Preserve the dataset's actual label ids and\nclass names rather than normalizing them unless the user explicitly asks for a\nconversion.",{"type":47,"tag":121,"props":1363,"children":1365},{"id":1364},"multi-gpu-multi-node",[1366],{"type":53,"value":1367},"Multi-GPU \u002F Multi-Node",{"type":47,"tag":56,"props":1369,"children":1370},{},[1371,1376,1378,1383],{"type":47,"tag":500,"props":1372,"children":1373},{},[1374],{"type":53,"value":1375},"Launch method:",{"type":53,"value":1377}," Lightning-managed (single ",{"type":47,"tag":72,"props":1379,"children":1381},{"className":1380},[],[1382],{"type":53,"value":811},{"type":53,"value":1384}," process, Lightning spawns workers).",{"type":47,"tag":535,"props":1386,"children":1387},{},[1388,1408],{"type":47,"tag":539,"props":1389,"children":1390},{},[1391],{"type":47,"tag":543,"props":1392,"children":1393},{},[1394,1398,1403],{"type":47,"tag":547,"props":1395,"children":1396},{},[1397],{"type":53,"value":556},{"type":47,"tag":547,"props":1399,"children":1400},{},[1401],{"type":53,"value":1402},"Description",{"type":47,"tag":547,"props":1404,"children":1405},{},[1406],{"type":53,"value":1407},"Default",{"type":47,"tag":573,"props":1409,"children":1410},{},[1411,1433,1458,1479,1501],{"type":47,"tag":543,"props":1412,"children":1413},{},[1414,1423,1428],{"type":47,"tag":580,"props":1415,"children":1416},{},[1417],{"type":47,"tag":72,"props":1418,"children":1420},{"className":1419},[],[1421],{"type":53,"value":1422},"train.num_gpus",{"type":47,"tag":580,"props":1424,"children":1425},{},[1426],{"type":53,"value":1427},"Number of GPUs",{"type":47,"tag":580,"props":1429,"children":1430},{},[1431],{"type":53,"value":1432},"1",{"type":47,"tag":543,"props":1434,"children":1435},{},[1436,1445,1450],{"type":47,"tag":580,"props":1437,"children":1438},{},[1439],{"type":47,"tag":72,"props":1440,"children":1442},{"className":1441},[],[1443],{"type":53,"value":1444},"train.gpu_ids",{"type":47,"tag":580,"props":1446,"children":1447},{},[1448],{"type":53,"value":1449},"GPU device indices",{"type":47,"tag":580,"props":1451,"children":1452},{},[1453],{"type":47,"tag":818,"props":1454,"children":1455},{},[1456],{"type":53,"value":1457},"0",{"type":47,"tag":543,"props":1459,"children":1460},{},[1461,1470,1475],{"type":47,"tag":580,"props":1462,"children":1463},{},[1464],{"type":47,"tag":72,"props":1465,"children":1467},{"className":1466},[],[1468],{"type":53,"value":1469},"train.num_nodes",{"type":47,"tag":580,"props":1471,"children":1472},{},[1473],{"type":53,"value":1474},"Number of nodes",{"type":47,"tag":580,"props":1476,"children":1477},{},[1478],{"type":53,"value":1432},{"type":47,"tag":543,"props":1480,"children":1481},{},[1482,1491,1496],{"type":47,"tag":580,"props":1483,"children":1484},{},[1485],{"type":47,"tag":72,"props":1486,"children":1488},{"className":1487},[],[1489],{"type":53,"value":1490},"train.sync_batchnorm",{"type":47,"tag":580,"props":1492,"children":1493},{},[1494],{"type":53,"value":1495},"Sync BN across GPUs",{"type":47,"tag":580,"props":1497,"children":1498},{},[1499],{"type":53,"value":1500},"configurable",{"type":47,"tag":543,"props":1502,"children":1503},{},[1504,1513,1518],{"type":47,"tag":580,"props":1505,"children":1506},{},[1507],{"type":47,"tag":72,"props":1508,"children":1510},{"className":1509},[],[1511],{"type":53,"value":1512},"train.use_distributed_sampler",{"type":47,"tag":580,"props":1514,"children":1515},{},[1516],{"type":53,"value":1517},"Use distributed sampler",{"type":47,"tag":580,"props":1519,"children":1520},{},[1521],{"type":53,"value":1500},{"type":47,"tag":492,"props":1523,"children":1524},{},[1525,1536],{"type":47,"tag":496,"props":1526,"children":1527},{},[1528,1530],{"type":53,"value":1529},"Multi-GPU strategy: ",{"type":47,"tag":72,"props":1531,"children":1533},{"className":1532},[],[1534],{"type":53,"value":1535},"ddp_find_unused_parameters_true",{"type":47,"tag":496,"props":1537,"children":1538},{},[1539],{"type":53,"value":1540},"No fsdp support",{"type":47,"tag":56,"props":1542,"children":1543},{},[1544,1549,1551,1557,1558,1564,1565,1571,1572,1578,1579,1585],{"type":47,"tag":500,"props":1545,"children":1546},{},[1547],{"type":53,"value":1548},"Multi-node env vars",{"type":53,"value":1550}," (set by orchestrator): ",{"type":47,"tag":72,"props":1552,"children":1554},{"className":1553},[],[1555],{"type":53,"value":1556},"WORLD_SIZE",{"type":53,"value":203},{"type":47,"tag":72,"props":1559,"children":1561},{"className":1560},[],[1562],{"type":53,"value":1563},"NODE_RANK",{"type":53,"value":203},{"type":47,"tag":72,"props":1566,"children":1568},{"className":1567},[],[1569],{"type":53,"value":1570},"MASTER_ADDR",{"type":53,"value":203},{"type":47,"tag":72,"props":1573,"children":1575},{"className":1574},[],[1576],{"type":53,"value":1577},"MASTER_PORT",{"type":53,"value":203},{"type":47,"tag":72,"props":1580,"children":1582},{"className":1581},[],[1583],{"type":53,"value":1584},"NUM_GPU_PER_NODE",{"type":53,"value":464},{"type":47,"tag":121,"props":1587,"children":1589},{"id":1588},"hardware",[1590],{"type":53,"value":1591},"Hardware",{"type":47,"tag":56,"props":1593,"children":1594},{},[1595],{"type":53,"value":1596},"Minimum 1 GPU(s), recommended 2 GPU(s). 16GB+ (V100 or A100) VRAM per GPU. SegFormer is relatively lightweight. Default img_size=256 is memory-friendly. Increase img_size for higher resolution at the cost of memory and speed.",{"type":47,"tag":121,"props":1598,"children":1600},{"id":1599},"error-patterns",[1601],{"type":53,"value":1602},"Error Patterns",{"type":47,"tag":56,"props":1604,"children":1605},{},[1606,1611],{"type":47,"tag":500,"props":1607,"children":1608},{},[1609],{"type":53,"value":1610},"CUDA out of memory",{"type":53,"value":1612},": Reduce batch_size or img_size. SegFormer memory scales quadratically with image size.",{"type":47,"tag":56,"props":1614,"children":1615},{},[1616,1621],{"type":47,"tag":500,"props":1617,"children":1618},{},[1619],{"type":53,"value":1620},"num_classes mismatch",{"type":53,"value":1622},": Ensure dataset.segment.num_classes matches the actual number of classes in your mask annotations.",{"type":47,"tag":56,"props":1624,"children":1625},{},[1626,1631,1633,1639],{"type":47,"tag":500,"props":1627,"children":1628},{},[1629],{"type":53,"value":1630},"TensorBoard unsupported for segmentation training",{"type":53,"value":1632},": Keep ",{"type":47,"tag":72,"props":1634,"children":1636},{"className":1635},[],[1637],{"type":53,"value":1638},"train.tensorboard.enabled: false",{"type":53,"value":1640},". The SegFormer training entrypoint asserts that TensorBoard visualization is not supported for segmentation, so do not enable TensorBoard just to extract AutoML metrics; use log parsing or a post-train evaluator instead.",{"type":47,"tag":56,"props":1642,"children":1643},{},[1644,1649,1651,1657,1659,1665,1666,1672,1674,1679,1680,1686,1688,1693],{"type":47,"tag":500,"props":1645,"children":1646},{},[1647],{"type":53,"value":1648},"AutoML metric extraction",{"type":53,"value":1650},": SegFormer train status files report ",{"type":47,"tag":72,"props":1652,"children":1654},{"className":1653},[],[1655],{"type":53,"value":1656},"val_miou",{"type":53,"value":1658}," alongside ",{"type":47,"tag":72,"props":1660,"children":1662},{"className":1661},[],[1663],{"type":53,"value":1664},"val_loss",{"type":53,"value":203},{"type":47,"tag":72,"props":1667,"children":1669},{"className":1668},[],[1670],{"type":53,"value":1671},"val_acc",{"type":53,"value":1673},", and other validation KPIs. Default AutoML train launches must optimize ",{"type":47,"tag":72,"props":1675,"children":1677},{"className":1676},[],[1678],{"type":53,"value":1656},{"type":53,"value":456},{"type":47,"tag":72,"props":1681,"children":1683},{"className":1682},[],[1684],{"type":53,"value":1685},"direction: maximize",{"type":53,"value":1687},"; do not optimize ",{"type":47,"tag":72,"props":1689,"children":1691},{"className":1690},[],[1692],{"type":53,"value":1664},{"type":53,"value":1694}," for default model invocations.",{"type":47,"tag":56,"props":1696,"children":1697},{},[1698,1700,1705,1707,1713,1715,1720],{"type":53,"value":1699},"For AutoML or long segmentation sweeps, read ",{"type":47,"tag":72,"props":1701,"children":1703},{"className":1702},[],[1704],{"type":53,"value":1656},{"type":53,"value":1706}," from\n",{"type":47,"tag":72,"props":1708,"children":1710},{"className":1709},[],[1711],{"type":53,"value":1712},"results_dir\u002Ftrain\u002Fstatus.json",{"type":53,"value":1714}," first. If the wrapper reports a terminal\nfailure but the structured status file reached the configured training budget\nand contains finite ",{"type":47,"tag":72,"props":1716,"children":1718},{"className":1717},[],[1719],{"type":53,"value":1656},{"type":53,"value":1721},", report the recovered metric with the wrapper\nfailure noted instead of discarding the measurement.",{"type":47,"tag":56,"props":1723,"children":1724},{},[1725,1727,1733,1734,1740,1741,1747,1748,1754,1755,1761,1762,1768,1770,1776],{"type":53,"value":1726},"For high-resolution custom segmentation targets, keep dataset paths as per-run\ninputs. Do not add customer\u002Fuser-specific roots to this reusable skill. When the\nuser asks for a fixed full-budget search, remember that bracket algorithms\n(",{"type":47,"tag":72,"props":1728,"children":1730},{"className":1729},[],[1731],{"type":53,"value":1732},"asha",{"type":53,"value":203},{"type":47,"tag":72,"props":1735,"children":1737},{"className":1736},[],[1738],{"type":53,"value":1739},"bohb",{"type":53,"value":203},{"type":47,"tag":72,"props":1742,"children":1744},{"className":1743},[],[1745],{"type":53,"value":1746},"dehb",{"type":53,"value":203},{"type":47,"tag":72,"props":1749,"children":1751},{"className":1750},[],[1752],{"type":53,"value":1753},"hyperband",{"type":53,"value":203},{"type":47,"tag":72,"props":1756,"children":1758},{"className":1757},[],[1759],{"type":53,"value":1760},"hyperband_es",{"type":53,"value":203},{"type":47,"tag":72,"props":1763,"children":1765},{"className":1764},[],[1766],{"type":53,"value":1767},"pbt",{"type":53,"value":1769},") may intentionally\nlower ",{"type":47,"tag":72,"props":1771,"children":1773},{"className":1772},[],[1774],{"type":53,"value":1775},"train.num_epochs",{"type":53,"value":1777}," for some recommendations; use Bayesian\u002FBFBO or lock the\nbudget if every recommendation must run the full epoch count.",{"type":47,"tag":56,"props":1779,"children":1780},{},[1781,1786,1788,1794,1796,1802,1804,1810,1812,1818,1820,1825,1826,1831,1832,1837],{"type":47,"tag":500,"props":1782,"children":1783},{},[1784],{"type":53,"value":1785},"Checkpoint handoff",{"type":53,"value":1787},": For evaluate\u002Fexport\u002Finference\u002Fquantize\u002Fresume, use the checkpoint resolver on the best AutoML child job's ",{"type":47,"tag":72,"props":1789,"children":1791},{"className":1790},[],[1792],{"type":53,"value":1793},"results_dir\u002Ftrain\u002F",{"type":53,"value":1795}," folder and select the action-appropriate ",{"type":47,"tag":72,"props":1797,"children":1799},{"className":1798},[],[1800],{"type":53,"value":1801},"model_epoch_*.pth",{"type":53,"value":1803}," checkpoint, such as ",{"type":47,"tag":72,"props":1805,"children":1807},{"className":1806},[],[1808],{"type":53,"value":1809},"model_epoch_000_step_00010.pth",{"type":53,"value":1811},". SegFormer may also write ",{"type":47,"tag":72,"props":1813,"children":1815},{"className":1814},[],[1816],{"type":53,"value":1817},"segformer_model_latest.pth",{"type":53,"value":1819},", but that should only be used when a caller explicitly requests latest. Preserve ",{"type":47,"tag":72,"props":1821,"children":1823},{"className":1822},[],[1824],{"type":53,"value":1264},{"type":53,"value":203},{"type":47,"tag":72,"props":1827,"children":1829},{"className":1828},[],[1830],{"type":53,"value":1293},{"type":53,"value":410},{"type":47,"tag":72,"props":1833,"children":1835},{"className":1834},[],[1836],{"type":53,"value":588},{"type":53,"value":1838}," overrides for downstream actions.",{"type":47,"tag":56,"props":1840,"children":1841},{},[1842,1847,1849,1854,1856,1862,1864,1870],{"type":47,"tag":500,"props":1843,"children":1844},{},[1845],{"type":53,"value":1846},"Resume\u002Fretrain checkpoint",{"type":53,"value":1848},": Resume uses ",{"type":47,"tag":72,"props":1850,"children":1852},{"className":1851},[],[1853],{"type":53,"value":462},{"type":53,"value":1855},".\nPass the exact resolved checkpoint from the previous train output, not a guessed\n",{"type":47,"tag":72,"props":1857,"children":1859},{"className":1858},[],[1860],{"type":53,"value":1861},"model.pth",{"type":53,"value":1863}," path. A resumed one-epoch run should produce the next checkpoint in\nthe new results directory, for example ",{"type":47,"tag":72,"props":1865,"children":1867},{"className":1866},[],[1868],{"type":53,"value":1869},"model_epoch_001_step_00020.pth",{"type":53,"value":464},{"type":47,"tag":56,"props":1872,"children":1873},{},[1874,1879,1880,1886,1887,1893,1895,1900,1902,1908],{"type":47,"tag":500,"props":1875,"children":1876},{},[1877],{"type":53,"value":1878},"Export \u002F TensorRT shape alignment",{"type":53,"value":1632},{"type":47,"tag":72,"props":1881,"children":1883},{"className":1882},[],[1884],{"type":53,"value":1885},"export.input_height",{"type":53,"value":1176},{"type":47,"tag":72,"props":1888,"children":1890},{"className":1889},[],[1891],{"type":53,"value":1892},"export.input_width",{"type":53,"value":1894}," aligned with ",{"type":47,"tag":72,"props":1896,"children":1898},{"className":1897},[],[1899],{"type":53,"value":1293},{"type":53,"value":1901}," unless the trained\nmodel and deploy specs have been validated at another resolution. The packaged\nfresh-install path is validated at ",{"type":47,"tag":72,"props":1903,"children":1905},{"className":1904},[],[1906],{"type":53,"value":1907},"256x256",{"type":53,"value":1909},", matching the default SegFormer\ndataset and deploy templates.",{"type":47,"tag":56,"props":1911,"children":1912},{},[1913,1926,1928,1933,1935,1940],{"type":47,"tag":500,"props":1914,"children":1915},{},[1916,1918,1924],{"type":53,"value":1917},"Parent ",{"type":47,"tag":72,"props":1919,"children":1921},{"className":1920},[],[1922],{"type":53,"value":1923},"segformer gen_trt_engine",{"type":53,"value":1925}," rejected by the PyT CLI",{"type":53,"value":1927},": In the validated 7.0.0 PyT container, ",{"type":47,"tag":72,"props":1929,"children":1931},{"className":1930},[],[1932],{"type":53,"value":1923},{"type":53,"value":1934}," is not a valid parent-model subtask. Use the SegFormer deploy workflow (",{"type":47,"tag":72,"props":1936,"children":1938},{"className":1937},[],[1939],{"type":53,"value":101},{"type":53,"value":1941},") for TensorRT engine generation, TensorRT evaluation, and TensorRT inference.",{"type":47,"tag":121,"props":1943,"children":1945},{"id":1944},"spec-param-parent-model-inference",[1946],{"type":53,"value":1947},"Spec Param \u002F Parent Model Inference",{"type":47,"tag":56,"props":1949,"children":1950},{},[1951,1953,1959,1961,1967,1969,1975],{"type":53,"value":1952},"Model-specific inference mappings belong in this MD file, not in ",{"type":47,"tag":72,"props":1954,"children":1956},{"className":1955},[],[1957],{"type":53,"value":1958},"config.json",{"type":53,"value":1960},". Generated runners should read this section and apply the mappings with SDK helpers before ",{"type":47,"tag":72,"props":1962,"children":1964},{"className":1963},[],[1965],{"type":53,"value":1966},"create_job()",{"type":53,"value":1968},". This mirrors the old microservices ",{"type":47,"tag":72,"props":1970,"children":1972},{"className":1971},[],[1973],{"type":53,"value":1974},"infer_params.py",{"type":53,"value":1976}," flow.",{"type":47,"tag":56,"props":1978,"children":1979},{},[1980,1982,1988],{"type":53,"value":1981},"Inference mappings from TAO Core ",{"type":47,"tag":72,"props":1983,"children":1985},{"className":1984},[],[1986],{"type":53,"value":1987},"segformer.config.json",{"type":53,"value":1989},":",{"type":47,"tag":535,"props":1991,"children":1992},{},[1993,2018],{"type":47,"tag":539,"props":1994,"children":1995},{},[1996],{"type":47,"tag":543,"props":1997,"children":1998},{},[1999,2003,2008,2013],{"type":47,"tag":547,"props":2000,"children":2001},{},[2002],{"type":53,"value":551},{"type":47,"tag":547,"props":2004,"children":2005},{},[2006],{"type":53,"value":2007},"Spec Field",{"type":47,"tag":547,"props":2009,"children":2010},{},[2011],{"type":53,"value":2012},"Inference Function",{"type":47,"tag":547,"props":2014,"children":2015},{},[2016],{"type":53,"value":2017},"Meaning",{"type":47,"tag":573,"props":2019,"children":2020},{},[2021,2051,2081,2109,2139,2166,2194,2224,2251,2278,2306,2334,2361,2388,2416,2443,2470,2500,2527,2555],{"type":47,"tag":543,"props":2022,"children":2023},{},[2024,2028,2037,2046],{"type":47,"tag":580,"props":2025,"children":2026},{},[2027],{"type":53,"value":85},{"type":47,"tag":580,"props":2029,"children":2030},{},[2031],{"type":47,"tag":72,"props":2032,"children":2034},{"className":2033},[],[2035],{"type":53,"value":2036},"encryption_key",{"type":47,"tag":580,"props":2038,"children":2039},{},[2040],{"type":47,"tag":72,"props":2041,"children":2043},{"className":2042},[],[2044],{"type":53,"value":2045},"key",{"type":47,"tag":580,"props":2047,"children":2048},{},[2049],{"type":53,"value":2050},"encryption key",{"type":47,"tag":543,"props":2052,"children":2053},{},[2054,2058,2067,2076],{"type":47,"tag":580,"props":2055,"children":2056},{},[2057],{"type":53,"value":85},{"type":47,"tag":580,"props":2059,"children":2060},{},[2061],{"type":47,"tag":72,"props":2062,"children":2064},{"className":2063},[],[2065],{"type":53,"value":2066},"evaluate.checkpoint",{"type":47,"tag":580,"props":2068,"children":2069},{},[2070],{"type":47,"tag":72,"props":2071,"children":2073},{"className":2072},[],[2074],{"type":53,"value":2075},"parent_model",{"type":47,"tag":580,"props":2077,"children":2078},{},[2079],{"type":53,"value":2080},"model file inferred from the parent job results folder",{"type":47,"tag":543,"props":2082,"children":2083},{},[2084,2088,2097,2105],{"type":47,"tag":580,"props":2085,"children":2086},{},[2087],{"type":53,"value":85},{"type":47,"tag":580,"props":2089,"children":2090},{},[2091],{"type":47,"tag":72,"props":2092,"children":2094},{"className":2093},[],[2095],{"type":53,"value":2096},"evaluate.trt_engine",{"type":47,"tag":580,"props":2098,"children":2099},{},[2100],{"type":47,"tag":72,"props":2101,"children":2103},{"className":2102},[],[2104],{"type":53,"value":2075},{"type":47,"tag":580,"props":2106,"children":2107},{},[2108],{"type":53,"value":2080},{"type":47,"tag":543,"props":2110,"children":2111},{},[2112,2116,2125,2134],{"type":47,"tag":580,"props":2113,"children":2114},{},[2115],{"type":53,"value":85},{"type":47,"tag":580,"props":2117,"children":2118},{},[2119],{"type":47,"tag":72,"props":2120,"children":2122},{"className":2121},[],[2123],{"type":53,"value":2124},"results_dir",{"type":47,"tag":580,"props":2126,"children":2127},{},[2128],{"type":47,"tag":72,"props":2129,"children":2131},{"className":2130},[],[2132],{"type":53,"value":2133},"output_dir",{"type":47,"tag":580,"props":2135,"children":2136},{},[2137],{"type":53,"value":2138},"current job results directory",{"type":47,"tag":543,"props":2140,"children":2141},{},[2142,2146,2154,2162],{"type":47,"tag":580,"props":2143,"children":2144},{},[2145],{"type":53,"value":357},{"type":47,"tag":580,"props":2147,"children":2148},{},[2149],{"type":47,"tag":72,"props":2150,"children":2152},{"className":2151},[],[2153],{"type":53,"value":2036},{"type":47,"tag":580,"props":2155,"children":2156},{},[2157],{"type":47,"tag":72,"props":2158,"children":2160},{"className":2159},[],[2161],{"type":53,"value":2045},{"type":47,"tag":580,"props":2163,"children":2164},{},[2165],{"type":53,"value":2050},{"type":47,"tag":543,"props":2167,"children":2168},{},[2169,2173,2182,2190],{"type":47,"tag":580,"props":2170,"children":2171},{},[2172],{"type":53,"value":357},{"type":47,"tag":580,"props":2174,"children":2175},{},[2176],{"type":47,"tag":72,"props":2177,"children":2179},{"className":2178},[],[2180],{"type":53,"value":2181},"export.checkpoint",{"type":47,"tag":580,"props":2183,"children":2184},{},[2185],{"type":47,"tag":72,"props":2186,"children":2188},{"className":2187},[],[2189],{"type":53,"value":2075},{"type":47,"tag":580,"props":2191,"children":2192},{},[2193],{"type":53,"value":2080},{"type":47,"tag":543,"props":2195,"children":2196},{},[2197,2201,2210,2219],{"type":47,"tag":580,"props":2198,"children":2199},{},[2200],{"type":53,"value":357},{"type":47,"tag":580,"props":2202,"children":2203},{},[2204],{"type":47,"tag":72,"props":2205,"children":2207},{"className":2206},[],[2208],{"type":53,"value":2209},"export.onnx_file",{"type":47,"tag":580,"props":2211,"children":2212},{},[2213],{"type":47,"tag":72,"props":2214,"children":2216},{"className":2215},[],[2217],{"type":53,"value":2218},"create_onnx_file",{"type":47,"tag":580,"props":2220,"children":2221},{},[2222],{"type":53,"value":2223},"output ONNX path",{"type":47,"tag":543,"props":2225,"children":2226},{},[2227,2231,2239,2247],{"type":47,"tag":580,"props":2228,"children":2229},{},[2230],{"type":53,"value":357},{"type":47,"tag":580,"props":2232,"children":2233},{},[2234],{"type":47,"tag":72,"props":2235,"children":2237},{"className":2236},[],[2238],{"type":53,"value":2124},{"type":47,"tag":580,"props":2240,"children":2241},{},[2242],{"type":47,"tag":72,"props":2243,"children":2245},{"className":2244},[],[2246],{"type":53,"value":2133},{"type":47,"tag":580,"props":2248,"children":2249},{},[2250],{"type":53,"value":2138},{"type":47,"tag":543,"props":2252,"children":2253},{},[2254,2258,2266,2274],{"type":47,"tag":580,"props":2255,"children":2256},{},[2257],{"type":53,"value":93},{"type":47,"tag":580,"props":2259,"children":2260},{},[2261],{"type":47,"tag":72,"props":2262,"children":2264},{"className":2263},[],[2265],{"type":53,"value":2036},{"type":47,"tag":580,"props":2267,"children":2268},{},[2269],{"type":47,"tag":72,"props":2270,"children":2272},{"className":2271},[],[2273],{"type":53,"value":2045},{"type":47,"tag":580,"props":2275,"children":2276},{},[2277],{"type":53,"value":2050},{"type":47,"tag":543,"props":2279,"children":2280},{},[2281,2285,2294,2302],{"type":47,"tag":580,"props":2282,"children":2283},{},[2284],{"type":53,"value":93},{"type":47,"tag":580,"props":2286,"children":2287},{},[2288],{"type":47,"tag":72,"props":2289,"children":2291},{"className":2290},[],[2292],{"type":53,"value":2293},"inference.checkpoint",{"type":47,"tag":580,"props":2295,"children":2296},{},[2297],{"type":47,"tag":72,"props":2298,"children":2300},{"className":2299},[],[2301],{"type":53,"value":2075},{"type":47,"tag":580,"props":2303,"children":2304},{},[2305],{"type":53,"value":2080},{"type":47,"tag":543,"props":2307,"children":2308},{},[2309,2313,2322,2330],{"type":47,"tag":580,"props":2310,"children":2311},{},[2312],{"type":53,"value":93},{"type":47,"tag":580,"props":2314,"children":2315},{},[2316],{"type":47,"tag":72,"props":2317,"children":2319},{"className":2318},[],[2320],{"type":53,"value":2321},"inference.trt_engine",{"type":47,"tag":580,"props":2323,"children":2324},{},[2325],{"type":47,"tag":72,"props":2326,"children":2328},{"className":2327},[],[2329],{"type":53,"value":2075},{"type":47,"tag":580,"props":2331,"children":2332},{},[2333],{"type":53,"value":2080},{"type":47,"tag":543,"props":2335,"children":2336},{},[2337,2341,2349,2357],{"type":47,"tag":580,"props":2338,"children":2339},{},[2340],{"type":53,"value":93},{"type":47,"tag":580,"props":2342,"children":2343},{},[2344],{"type":47,"tag":72,"props":2345,"children":2347},{"className":2346},[],[2348],{"type":53,"value":2124},{"type":47,"tag":580,"props":2350,"children":2351},{},[2352],{"type":47,"tag":72,"props":2353,"children":2355},{"className":2354},[],[2356],{"type":53,"value":2133},{"type":47,"tag":580,"props":2358,"children":2359},{},[2360],{"type":53,"value":2138},{"type":47,"tag":543,"props":2362,"children":2363},{},[2364,2368,2376,2384],{"type":47,"tag":580,"props":2365,"children":2366},{},[2367],{"type":53,"value":408},{"type":47,"tag":580,"props":2369,"children":2370},{},[2371],{"type":47,"tag":72,"props":2372,"children":2374},{"className":2373},[],[2375],{"type":53,"value":2036},{"type":47,"tag":580,"props":2377,"children":2378},{},[2379],{"type":47,"tag":72,"props":2380,"children":2382},{"className":2381},[],[2383],{"type":53,"value":2045},{"type":47,"tag":580,"props":2385,"children":2386},{},[2387],{"type":53,"value":2050},{"type":47,"tag":543,"props":2389,"children":2390},{},[2391,2395,2404,2412],{"type":47,"tag":580,"props":2392,"children":2393},{},[2394],{"type":53,"value":408},{"type":47,"tag":580,"props":2396,"children":2397},{},[2398],{"type":47,"tag":72,"props":2399,"children":2401},{"className":2400},[],[2402],{"type":53,"value":2403},"quantize.model_path",{"type":47,"tag":580,"props":2405,"children":2406},{},[2407],{"type":47,"tag":72,"props":2408,"children":2410},{"className":2409},[],[2411],{"type":53,"value":2075},{"type":47,"tag":580,"props":2413,"children":2414},{},[2415],{"type":53,"value":2080},{"type":47,"tag":543,"props":2417,"children":2418},{},[2419,2423,2431,2439],{"type":47,"tag":580,"props":2420,"children":2421},{},[2422],{"type":53,"value":408},{"type":47,"tag":580,"props":2424,"children":2425},{},[2426],{"type":47,"tag":72,"props":2427,"children":2429},{"className":2428},[],[2430],{"type":53,"value":2124},{"type":47,"tag":580,"props":2432,"children":2433},{},[2434],{"type":47,"tag":72,"props":2435,"children":2437},{"className":2436},[],[2438],{"type":53,"value":2133},{"type":47,"tag":580,"props":2440,"children":2441},{},[2442],{"type":53,"value":2138},{"type":47,"tag":543,"props":2444,"children":2445},{},[2446,2450,2458,2466],{"type":47,"tag":580,"props":2447,"children":2448},{},[2449],{"type":53,"value":383},{"type":47,"tag":580,"props":2451,"children":2452},{},[2453],{"type":47,"tag":72,"props":2454,"children":2456},{"className":2455},[],[2457],{"type":53,"value":2036},{"type":47,"tag":580,"props":2459,"children":2460},{},[2461],{"type":47,"tag":72,"props":2462,"children":2464},{"className":2463},[],[2465],{"type":53,"value":2045},{"type":47,"tag":580,"props":2467,"children":2468},{},[2469],{"type":53,"value":2050},{"type":47,"tag":543,"props":2471,"children":2472},{},[2473,2477,2486,2495],{"type":47,"tag":580,"props":2474,"children":2475},{},[2476],{"type":53,"value":383},{"type":47,"tag":580,"props":2478,"children":2479},{},[2480],{"type":47,"tag":72,"props":2481,"children":2483},{"className":2482},[],[2484],{"type":53,"value":2485},"model.backbone.pretrained_backbone_path",{"type":47,"tag":580,"props":2487,"children":2488},{},[2489],{"type":47,"tag":72,"props":2490,"children":2492},{"className":2491},[],[2493],{"type":53,"value":2494},"ptm_if_no_resume_model",{"type":47,"tag":580,"props":2496,"children":2497},{},[2498],{"type":53,"value":2499},"PTM when no resume checkpoint exists",{"type":47,"tag":543,"props":2501,"children":2502},{},[2503,2507,2515,2523],{"type":47,"tag":580,"props":2504,"children":2505},{},[2506],{"type":53,"value":383},{"type":47,"tag":580,"props":2508,"children":2509},{},[2510],{"type":47,"tag":72,"props":2511,"children":2513},{"className":2512},[],[2514],{"type":53,"value":2124},{"type":47,"tag":580,"props":2516,"children":2517},{},[2518],{"type":47,"tag":72,"props":2519,"children":2521},{"className":2520},[],[2522],{"type":53,"value":2133},{"type":47,"tag":580,"props":2524,"children":2525},{},[2526],{"type":53,"value":2138},{"type":47,"tag":543,"props":2528,"children":2529},{},[2530,2534,2543,2551],{"type":47,"tag":580,"props":2531,"children":2532},{},[2533],{"type":53,"value":383},{"type":47,"tag":580,"props":2535,"children":2536},{},[2537],{"type":47,"tag":72,"props":2538,"children":2540},{"className":2539},[],[2541],{"type":53,"value":2542},"train.pretrained_model_path",{"type":47,"tag":580,"props":2544,"children":2545},{},[2546],{"type":47,"tag":72,"props":2547,"children":2549},{"className":2548},[],[2550],{"type":53,"value":2494},{"type":47,"tag":580,"props":2552,"children":2553},{},[2554],{"type":53,"value":2499},{"type":47,"tag":543,"props":2556,"children":2557},{},[2558,2562,2570,2579],{"type":47,"tag":580,"props":2559,"children":2560},{},[2561],{"type":53,"value":383},{"type":47,"tag":580,"props":2563,"children":2564},{},[2565],{"type":47,"tag":72,"props":2566,"children":2568},{"className":2567},[],[2569],{"type":53,"value":462},{"type":47,"tag":580,"props":2571,"children":2572},{},[2573],{"type":47,"tag":72,"props":2574,"children":2576},{"className":2575},[],[2577],{"type":53,"value":2578},"resume_model",{"type":47,"tag":580,"props":2580,"children":2581},{},[2582],{"type":53,"value":2583},"model file inferred from the current job results folder",{"type":47,"tag":56,"props":2585,"children":2586},{},[2587,2589,2594,2596,2602,2604,2610,2612,2617],{"type":53,"value":2588},"For ",{"type":47,"tag":72,"props":2590,"children":2592},{"className":2591},[],[2593],{"type":53,"value":2075},{"type":53,"value":2595}," or ",{"type":47,"tag":72,"props":2597,"children":2599},{"className":2598},[],[2600],{"type":53,"value":2601},"parent_model_folder",{"type":53,"value":2603},", pass the upstream train\u002Fexport\u002FAutoML child job id as ",{"type":47,"tag":72,"props":2605,"children":2607},{"className":2606},[],[2608],{"type":53,"value":2609},"parent_job_id",{"type":53,"value":2611},". 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":47,"tag":72,"props":2613,"children":2615},{"className":2614},[],[2616],{"type":53,"value":1958},{"type":53,"value":2618}," and do not patch generated runner scripts to guess checkpoint paths.",{"type":47,"tag":121,"props":2620,"children":2622},{"id":2621},"deployment",[2623],{"type":53,"value":2624},"Deployment",{"type":47,"tag":492,"props":2626,"children":2627},{},[2628],{"type":47,"tag":496,"props":2629,"children":2630},{},[2631],{"type":47,"tag":2632,"props":2633,"children":2634},"a",{"href":101},[2635],{"type":53,"value":2636},"tao-deploy-segformer",{"type":47,"tag":2638,"props":2639,"children":2640},"style",{},[2641],{"type":53,"value":2642},"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":2644,"total":2798},[2645,2663,2678,2689,2701,2715,2728,2740,2753,2764,2778,2787],{"slug":2646,"name":2646,"fn":2647,"description":2648,"org":2649,"tags":2650,"stars":2660,"repoUrl":2661,"updatedAt":2662},"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},[2651,2654,2657],{"name":2652,"slug":2653,"type":15},"Documentation","documentation",{"name":2655,"slug":2656,"type":15},"MCP","mcp",{"name":2658,"slug":2659,"type":15},"Search","search",21777,"https:\u002F\u002Fgithub.com\u002FNVIDIA\u002FNemoClaw","2026-07-20T06:00:01.461044",{"slug":2664,"name":2664,"fn":2665,"description":2666,"org":2667,"tags":2668,"stars":2675,"repoUrl":2676,"updatedAt":2677},"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},[2669,2672,2673],{"name":2670,"slug":2671,"type":15},"Containers","containers",{"name":2624,"slug":2621,"type":15},{"name":2674,"slug":811,"type":15},"Python",17049,"https:\u002F\u002Fgithub.com\u002FNVIDIA\u002FMegatron-LM","2026-07-27T06:06:11.249662",{"slug":2679,"name":2679,"fn":2680,"description":2681,"org":2682,"tags":2683,"stars":2675,"repoUrl":2676,"updatedAt":2688},"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},[2684,2687],{"name":2685,"slug":2686,"type":15},"CI\u002FCD","ci-cd",{"name":2624,"slug":2621,"type":15},"2026-07-14T05:25:59.97109",{"slug":2690,"name":2690,"fn":2691,"description":2692,"org":2693,"tags":2694,"stars":2675,"repoUrl":2676,"updatedAt":2700},"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},[2695,2696,2697],{"name":2685,"slug":2686,"type":15},{"name":2624,"slug":2621,"type":15},{"name":2698,"slug":2699,"type":15},"GitHub","github","2026-07-27T06:06:12.278222",{"slug":2702,"name":2702,"fn":2703,"description":2704,"org":2705,"tags":2706,"stars":2675,"repoUrl":2676,"updatedAt":2714},"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},[2707,2710,2711],{"name":2708,"slug":2709,"type":15},"Debugging","debugging",{"name":2698,"slug":2699,"type":15},{"name":2712,"slug":2713,"type":15},"Triage","triage","2026-07-14T05:25:57.442089",{"slug":2716,"name":2716,"fn":2717,"description":2718,"org":2719,"tags":2720,"stars":2675,"repoUrl":2676,"updatedAt":2727},"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},[2721,2724],{"name":2722,"slug":2723,"type":15},"Best Practices","best-practices",{"name":2725,"slug":2726,"type":15},"Code Analysis","code-analysis","2026-07-14T05:25:56.18433",{"slug":2729,"name":2729,"fn":2730,"description":2731,"org":2732,"tags":2733,"stars":2675,"repoUrl":2676,"updatedAt":2739},"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},[2734,2735,2738],{"name":17,"slug":18,"type":15},{"name":2736,"slug":2737,"type":15},"Migration","migration",{"name":9,"slug":8,"type":15},"2026-07-17T06:07:11.777011",{"slug":2741,"name":2741,"fn":2742,"description":2743,"org":2744,"tags":2745,"stars":2675,"repoUrl":2676,"updatedAt":2752},"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},[2746,2749],{"name":2747,"slug":2748,"type":15},"QA","qa",{"name":2750,"slug":2751,"type":15},"Testing","testing","2026-07-14T05:25:53.673039",{"slug":2754,"name":2754,"fn":2755,"description":2756,"org":2757,"tags":2758,"stars":2675,"repoUrl":2676,"updatedAt":2763},"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},[2759,2760],{"name":2624,"slug":2621,"type":15},{"name":2761,"slug":2762,"type":15},"Infrastructure","infrastructure","2026-07-14T05:25:49.362534",{"slug":2765,"name":2765,"fn":2766,"description":2767,"org":2768,"tags":2769,"stars":2675,"repoUrl":2676,"updatedAt":2777},"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},[2770,2773,2774],{"name":2771,"slug":2772,"type":15},"Code Review","code-review",{"name":2698,"slug":2699,"type":15},{"name":2775,"slug":2776,"type":15},"Pull Requests","pull-requests","2026-07-14T05:26:01.226578",{"slug":2779,"name":2779,"fn":2780,"description":2781,"org":2782,"tags":2783,"stars":2675,"repoUrl":2676,"updatedAt":2786},"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},[2784,2785],{"name":2747,"slug":2748,"type":15},{"name":2750,"slug":2751,"type":15},"2026-07-14T05:25:54.928983",{"slug":2788,"name":2788,"fn":2789,"description":2790,"org":2791,"tags":2792,"stars":2675,"repoUrl":2676,"updatedAt":2797},"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},[2793,2796],{"name":2794,"slug":2795,"type":15},"Automation","automation",{"name":2685,"slug":2686,"type":15},"2026-07-30T05:29:03.275638",496,{"items":2800,"total":2896},[2801,2818,2828,2842,2852,2867,2882],{"slug":2802,"name":2802,"fn":2803,"description":2804,"org":2805,"tags":2806,"stars":23,"repoUrl":24,"updatedAt":2817},"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},[2807,2810,2813,2814],{"name":2808,"slug":2809,"type":15},"Data Analysis","data-analysis",{"name":2811,"slug":2812,"type":15},"Data Engineering","data-engineering",{"name":9,"slug":8,"type":15},{"name":2815,"slug":2816,"type":15},"Performance","performance","2026-07-14T05:28:43.176466",{"slug":2819,"name":2819,"fn":2820,"description":2821,"org":2822,"tags":2823,"stars":23,"repoUrl":24,"updatedAt":2827},"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},[2824,2825,2826],{"name":2624,"slug":2621,"type":15},{"name":2761,"slug":2762,"type":15},{"name":9,"slug":8,"type":15},"2026-07-14T05:29:06.667109",{"slug":2829,"name":2829,"fn":2830,"description":2831,"org":2832,"tags":2833,"stars":23,"repoUrl":24,"updatedAt":2841},"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},[2834,2837,2838],{"name":2835,"slug":2836,"type":15},"Agents","agents",{"name":9,"slug":8,"type":15},{"name":2839,"slug":2840,"type":15},"Research","research","2026-07-14T05:28:06.816956",{"slug":2843,"name":2843,"fn":2844,"description":2845,"org":2846,"tags":2847,"stars":23,"repoUrl":24,"updatedAt":2851},"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},[2848,2849,2850],{"name":2808,"slug":2809,"type":15},{"name":9,"slug":8,"type":15},{"name":2750,"slug":2751,"type":15},"2026-07-17T05:29:03.913266",{"slug":2853,"name":2853,"fn":2854,"description":2855,"org":2856,"tags":2857,"stars":23,"repoUrl":24,"updatedAt":2866},"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},[2858,2859,2862,2863],{"name":2794,"slug":2795,"type":15},{"name":2860,"slug":2861,"type":15},"Imaging","imaging",{"name":9,"slug":8,"type":15},{"name":2864,"slug":2865,"type":15},"Video","video","2026-07-17T05:28:53.905004",{"slug":2868,"name":2868,"fn":2869,"description":2870,"org":2871,"tags":2872,"stars":23,"repoUrl":24,"updatedAt":2881},"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},[2873,2874,2877,2878],{"name":2624,"slug":2621,"type":15},{"name":2875,"slug":2876,"type":15},"Docker","docker",{"name":9,"slug":8,"type":15},{"name":2879,"slug":2880,"type":15},"Operations","operations","2026-07-17T05:28:56.913999",{"slug":2883,"name":2883,"fn":2884,"description":2885,"org":2886,"tags":2887,"stars":23,"repoUrl":24,"updatedAt":2895},"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},[2888,2889,2892],{"name":9,"slug":8,"type":15},{"name":2890,"slug":2891,"type":15},"Quantum Computing","quantum-computing",{"name":2893,"slug":2894,"type":15},"Simulation","simulation","2026-07-14T05:26:58.898253",305]