[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill-nvidia-i4h-workflow-finetune":3,"mdc-fy8m95-key":34,"related-repo-nvidia-i4h-workflow-finetune":2297,"related-org-nvidia-i4h-workflow-finetune":2401},{"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},"i4h-workflow-finetune","fine-tune robot policies on LeRobot datasets","Fine-tune a GR00T or openpi PI0 policy on a LeRobot dataset. Use when asked to finetune, train, or post-train a policy on demos; not for evaluating a checkpoint (use [[i4h-workflow-validate]]).",{"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},"Robotics","robotics","tag",{"name":17,"slug":18,"type":15},"Machine Learning","machine-learning",{"name":9,"slug":8,"type":15},{"name":21,"slug":22,"type":15},"Simulation","simulation",2473,"https:\u002F\u002Fgithub.com\u002FNVIDIA\u002Fskills","2026-07-23T06:05:57.373594","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\u002Fi4h-workflow-finetune","---\nname: i4h-workflow-finetune\nversion: \"0.6.0\"\ndescription: Fine-tune a GR00T or openpi PI0 policy on a LeRobot dataset. Use when asked to finetune, train, or post-train a policy on demos; not for evaluating a checkpoint (use [[i4h-workflow-validate]]).\nlicense: Apache-2.0\nmetadata:\n  author: \"Isaac for Healthcare Team \u003Cisaac-for-healthcare-support@nvidia.com>\"\n  tags:\n    - isaac-for-healthcare\n    - i4h\n    - agentic-workflow\n    - finetune\n    - training\n---\n\n# i4h Workflow — Finetune\n\n## Purpose\n\nFine-tune a GR00T or openpi PI0 policy on an existing LeRobot dataset. Use when asked to finetune, train, or post-train a policy on recorded demos.\n\n## Base Code\n\nThese steps drive the i4h-workflows base code (the `workflows\u002Fagentic\u002F` tree). To reuse an existing checkout, set `I4H_WORKFLOWS` to its path (no clone happens). Otherwise this resolves the current repo, or clones to `~\u002Fi4h-workflows` — pick that default without prompting. Run every command below from the resolved root:\n\n```bash\n# Resolve the i4h-workflows base code (provides workflows\u002Fagentic\u002F).\nROOT=\"${I4H_WORKFLOWS:-$(git rev-parse --show-toplevel 2>\u002Fdev\u002Fnull)}\"\nif [ ! -d \"$ROOT\u002Fworkflows\u002Fagentic\" ]; then\n  ROOT=\"${I4H_WORKFLOWS:-$HOME\u002Fi4h-workflows}\"\n  [ -d \"$ROOT\u002Fworkflows\u002Fagentic\" ] || git clone https:\u002F\u002Fgithub.com\u002Fisaac-for-healthcare\u002Fi4h-workflows \"$ROOT\"\nfi\nexport I4H_WORKFLOWS=\"$ROOT\"; cd \"$ROOT\"\n```\n\n## Basics\n\n- The dataset path must be an existing LeRobot directory with `meta\u002Finfo.json`.\n- Train support is determined by `policy.train_module` in `workflows\u002Fagentic\u002Fconfig\u002Fenvironments\u002F\u003Cenv>.yaml`. A null value means inference-only.\n- `assemble_trocar` is inference-only.\n\n## Stack Map\n\n| Env | Stack | CLI |\n|---|---|---|\n| `scissor_pick_and_place` | `gr00t_n15` | `i4h-agentic-gr00t-n15-train` |\n| `locomanip_tray_pick_and_place` | `gr00t_n16` | `i4h-agentic-gr00t-n16-train` |\n| `locomanip_push_cart` | `gr00t_n16` | `i4h-agentic-gr00t-n16-train` |\n| `ultrasound_liver_scan` | `openpi_pi0` | `i4h-agentic-openpi-pi0-train` |\n\nN1.6 locomanip envs share `policy.locomanip.train`.\n\n## Preflight\n\n```bash\ntest -f \"${DATASET_PATH}\u002Fmeta\u002Finfo.json\"\nnvidia-smi --query-gpu=name --format=csv,noheader | wc -l\nworkflows\u002Fagentic\u002Fpolicy\u002F\u003Cstack>\u002Frun.sh --list-envs\n```\n\n## Run\n\nRun the steps below in order. Each step is a separate bash call; variables persist in the local agent's tmux session.\n\n### Step 1 — setup and resolve dataset\n\n```bash\nREPO_ROOT=\"${I4H_WORKFLOWS:-$(git rev-parse --show-toplevel 2>\u002Fdev\u002Fnull)}\"; [ -d \"$REPO_ROOT\u002Fworkflows\u002Fagentic\" ] || REPO_ROOT=\"$HOME\u002Fi4h-workflows\"\nENV_ID=scissor_pick_and_place\nSTACK_DIR=gr00t_n15\nTRAIN_CLI=i4h-agentic-gr00t-n15-train\nRUNS_ROOT=\"${REPO_ROOT}\u002Fworkflows\u002Fagentic\u002Fruns\"\n\n# Point DATASET_PATH at a converted LeRobot dataset dir (absolute; must contain meta\u002Finfo.json),\n# produced by [[i4h-workflow-dataset-convert]]. List candidates:\n#   find \"${RUNS_ROOT}\" \"${HF_LEROBOT_HOME:-$HOME\u002F.cache\u002Fhuggingface\u002Flerobot}\" -name info.json -path '*\u002Fmeta\u002F*' -printf '%h\\n' | sed 's#\u002Fmeta$##' | sort -u\nDATASET_PATH=\"${DATASET_PATH:-}\"\nif [ ! -f \"${DATASET_PATH%\u002F}\u002Fmeta\u002Finfo.json\" ]; then\n  echo \"finetune: set DATASET_PATH to a LeRobot dataset dir with meta\u002Finfo.json (got '${DATASET_PATH:-\u003Cunset>}'). Candidates:\" >&2\n  find \"${RUNS_ROOT}\" \"${HF_LEROBOT_HOME:-$HOME\u002F.cache\u002Fhuggingface\u002Flerobot}\" -name info.json -path '*\u002Fmeta\u002F*' -printf '%h\\n' 2>\u002Fdev\u002Fnull | sed 's#\u002Fmeta$##' | sort -u | head\n  exit 1\nfi\n\nRUN_DIR=\"${RUNS_ROOT}\u002Ffinetune_${ENV_ID}_$(date +%Y%m%d_%H%M%S)\"\nOUT=\"${RUN_DIR}\u002Fcheckpoint\"\nexport TMPDIR=\u002Ftmp   # short path: torch DataLoader FD-sharing socket must fit AF_UNIX's 108-byte limit\nmkdir -p \"${OUT}\" \"${RUN_DIR}\u002Flogs\"\nln -sfn \"${RUN_DIR}\" \"${RUNS_ROOT}\u002F.latest\"\n```\n\n### Step 2 — train\n\n```bash\nuv --directory \"${REPO_ROOT}\u002Fworkflows\u002Fagentic\u002Fpolicy\u002F${STACK_DIR}\" run \"${TRAIN_CLI}\" \\\n  --env \"${ENV_ID}\" \\\n  --dataset-path \"${DATASET_PATH}\" \\\n  --output-dir \"${OUT}\" \\\n  --max-steps 1000 \\\n  --save-steps 1000 \\\n  --num-gpus 1 \\\n  2>&1 | tee \"${RUN_DIR}\u002Flogs\u002Ffinetune.log\"\n```\n\nTyro flags use kebab case (`--max-steps`, not `--max_steps`).\n\n## Common Flags\n\n- `--dataset-path PATH` (required)\n- `--output-dir PATH`\n- `--base-model-path PATH_OR_REPO` overrides YAML `policy.model_repo`\n- `--max-steps N`, `--save-steps N`\n- `--batch-size N`, `--learning-rate FLOAT`\n- `--no-tune-visual` — freeze the vision backbone (trains the action head + projector only): ~2× faster, ~half the memory, less overfitting. Good default for small datasets; unfreeze only with lots of data + a real visual domain gap.\n- `--num-gpus N` — must not exceed visible GPUs\n- `--report-to tensorboard|wandb`\n\n## Verify\n\n- Checkpoint directory `${OUT}\u002Fcheckpoint-\u003CN>` contains `model-0000*-of-*.safetensors`, `experiment_cfg\u002F`, `processor\u002F`.\n- Log contains `train_loss` lines and a final `'train_runtime': ...` summary.\n\n## Prerequisites\n\n- Workflow set up via [[i4h-workflow-setup]] (the stack's `.venv` must exist).\n- An existing LeRobot dataset directory with `meta\u002Finfo.json`.\n- A train-capable env: `policy.train_module` non-null in `workflows\u002Fagentic\u002Fconfig\u002Fenvironments\u002F\u003Cenv>.yaml` (`assemble_trocar` is inference-only).\n- At least one visible GPU (`--num-gpus` must not exceed visible GPUs).\n\n## Limitations\n\n- Inference-only envs (null `policy.train_module`, e.g. `assemble_trocar`) cannot be fine-tuned.\n- Requires GPU(s); `--num-gpus` must not exceed the count from `nvidia-smi`.\n- N1.6 locomanip envs share `policy.locomanip.train`.\n- Each env maps to one stack\u002FCLI (see Stack Map); the dataset must match that env.\n\n## Troubleshooting\n\n- **Error:** train CLI \u002F module import fails - Cause: workflow not set up, stack `.venv` missing. Fix: run [[i4h-workflow-setup]] first.\n- **Error:** dataset path rejected \u002F missing `meta\u002Finfo.json` - Cause: `--dataset-path` is not a valid LeRobot directory. Fix: point to a converted LeRobot dataset (see Preflight `test -f`).\n- **Error:** env is inference-only \u002F no train support - Cause: `policy.train_module` is null for that env. Fix: choose a train-capable env from the Stack Map.\n- **Error:** unrecognized flag like `--max_steps` - Cause: Tyro flags use kebab case. Fix: use `--max-steps` form.\n\n## Final Response\n\nReport env, stack, dataset path, output checkpoint path, train_loss summary, and blockers.\n",{"data":35,"body":45},{"name":4,"version":36,"description":6,"license":26,"metadata":37},"0.6.0",{"author":38,"tags":39},"Isaac for Healthcare Team \u003Cisaac-for-healthcare-support@nvidia.com>",[40,41,42,43,44],"isaac-for-healthcare","i4h","agentic-workflow","finetune","training",{"type":46,"children":47},"root",[48,56,63,69,75,105,447,453,503,509,660,672,678,773,779,784,791,1618,1624,1855,1876,1882,1986,1992,2050,2056,2122,2128,2183,2189,2280,2286,2291],{"type":49,"tag":50,"props":51,"children":52},"element","h1",{"id":4},[53],{"type":54,"value":55},"text","i4h Workflow — Finetune",{"type":49,"tag":57,"props":58,"children":60},"h2",{"id":59},"purpose",[61],{"type":54,"value":62},"Purpose",{"type":49,"tag":64,"props":65,"children":66},"p",{},[67],{"type":54,"value":68},"Fine-tune a GR00T or openpi PI0 policy on an existing LeRobot dataset. Use when asked to finetune, train, or post-train a policy on recorded demos.",{"type":49,"tag":57,"props":70,"children":72},{"id":71},"base-code",[73],{"type":54,"value":74},"Base Code",{"type":49,"tag":64,"props":76,"children":77},{},[78,80,87,89,95,97,103],{"type":54,"value":79},"These steps drive the i4h-workflows base code (the ",{"type":49,"tag":81,"props":82,"children":84},"code",{"className":83},[],[85],{"type":54,"value":86},"workflows\u002Fagentic\u002F",{"type":54,"value":88}," tree). To reuse an existing checkout, set ",{"type":49,"tag":81,"props":90,"children":92},{"className":91},[],[93],{"type":54,"value":94},"I4H_WORKFLOWS",{"type":54,"value":96}," to its path (no clone happens). Otherwise this resolves the current repo, or clones to ",{"type":49,"tag":81,"props":98,"children":100},{"className":99},[],[101],{"type":54,"value":102},"~\u002Fi4h-workflows",{"type":54,"value":104}," — pick that default without prompting. Run every command below from the resolved root:",{"type":49,"tag":106,"props":107,"children":112},"pre",{"className":108,"code":109,"language":110,"meta":111,"style":111},"language-bash shiki shiki-themes material-theme-lighter material-theme material-theme-palenight","# Resolve the i4h-workflows base code (provides workflows\u002Fagentic\u002F).\nROOT=\"${I4H_WORKFLOWS:-$(git rev-parse --show-toplevel 2>\u002Fdev\u002Fnull)}\"\nif [ ! -d \"$ROOT\u002Fworkflows\u002Fagentic\" ]; then\n  ROOT=\"${I4H_WORKFLOWS:-$HOME\u002Fi4h-workflows}\"\n  [ -d \"$ROOT\u002Fworkflows\u002Fagentic\" ] || git clone https:\u002F\u002Fgithub.com\u002Fisaac-for-healthcare\u002Fi4h-workflows \"$ROOT\"\nfi\nexport I4H_WORKFLOWS=\"$ROOT\"; cd \"$ROOT\"\n","bash","",[113],{"type":49,"tag":81,"props":114,"children":115},{"__ignoreMap":111},[116,128,218,273,316,384,393],{"type":49,"tag":117,"props":118,"children":121},"span",{"class":119,"line":120},"line",1,[122],{"type":49,"tag":117,"props":123,"children":125},{"style":124},"--shiki-light:#90A4AE;--shiki-light-font-style:italic;--shiki-default:#546E7A;--shiki-default-font-style:italic;--shiki-dark:#676E95;--shiki-dark-font-style:italic",[126],{"type":54,"value":127},"# Resolve the i4h-workflows base code (provides workflows\u002Fagentic\u002F).\n",{"type":49,"tag":117,"props":129,"children":131},{"class":119,"line":130},2,[132,138,144,149,153,158,164,169,174,179,184,189,194,199,203,208,213],{"type":49,"tag":117,"props":133,"children":135},{"style":134},"--shiki-light:#90A4AE;--shiki-default:#EEFFFF;--shiki-dark:#BABED8",[136],{"type":54,"value":137},"ROOT",{"type":49,"tag":117,"props":139,"children":141},{"style":140},"--shiki-light:#39ADB5;--shiki-default:#89DDFF;--shiki-dark:#89DDFF",[142],{"type":54,"value":143},"=",{"type":49,"tag":117,"props":145,"children":146},{"style":140},[147],{"type":54,"value":148},"\"${",{"type":49,"tag":117,"props":150,"children":151},{"style":134},[152],{"type":54,"value":94},{"type":49,"tag":117,"props":154,"children":155},{"style":140},[156],{"type":54,"value":157},":-",{"type":49,"tag":117,"props":159,"children":161},{"style":160},"--shiki-light:#91B859;--shiki-default:#C3E88D;--shiki-dark:#C3E88D",[162],{"type":54,"value":163},"$(",{"type":49,"tag":117,"props":165,"children":166},{"style":134},[167],{"type":54,"value":168},"git",{"type":49,"tag":117,"props":170,"children":171},{"style":134},[172],{"type":54,"value":173}," rev-parse",{"type":49,"tag":117,"props":175,"children":176},{"style":134},[177],{"type":54,"value":178}," --show-toplevel",{"type":49,"tag":117,"props":180,"children":181},{"style":134},[182],{"type":54,"value":183}," 2",{"type":49,"tag":117,"props":185,"children":186},{"style":160},[187],{"type":54,"value":188},">",{"type":49,"tag":117,"props":190,"children":191},{"style":140},[192],{"type":54,"value":193},"\u002F",{"type":49,"tag":117,"props":195,"children":196},{"style":134},[197],{"type":54,"value":198},"dev",{"type":49,"tag":117,"props":200,"children":201},{"style":140},[202],{"type":54,"value":193},{"type":49,"tag":117,"props":204,"children":205},{"style":134},[206],{"type":54,"value":207},"null",{"type":49,"tag":117,"props":209,"children":210},{"style":160},[211],{"type":54,"value":212},")",{"type":49,"tag":117,"props":214,"children":215},{"style":140},[216],{"type":54,"value":217},"}\"\n",{"type":49,"tag":117,"props":219,"children":221},{"class":119,"line":220},3,[222,228,233,238,243,248,253,258,263,268],{"type":49,"tag":117,"props":223,"children":225},{"style":224},"--shiki-light:#39ADB5;--shiki-light-font-style:italic;--shiki-default:#89DDFF;--shiki-default-font-style:italic;--shiki-dark:#89DDFF;--shiki-dark-font-style:italic",[226],{"type":54,"value":227},"if",{"type":49,"tag":117,"props":229,"children":230},{"style":140},[231],{"type":54,"value":232}," [",{"type":49,"tag":117,"props":234,"children":235},{"style":140},[236],{"type":54,"value":237}," !",{"type":49,"tag":117,"props":239,"children":240},{"style":140},[241],{"type":54,"value":242}," -d",{"type":49,"tag":117,"props":244,"children":245},{"style":140},[246],{"type":54,"value":247}," \"",{"type":49,"tag":117,"props":249,"children":250},{"style":134},[251],{"type":54,"value":252},"$ROOT",{"type":49,"tag":117,"props":254,"children":255},{"style":160},[256],{"type":54,"value":257},"\u002Fworkflows\u002Fagentic",{"type":49,"tag":117,"props":259,"children":260},{"style":140},[261],{"type":54,"value":262},"\"",{"type":49,"tag":117,"props":264,"children":265},{"style":140},[266],{"type":54,"value":267}," ];",{"type":49,"tag":117,"props":269,"children":270},{"style":224},[271],{"type":54,"value":272}," then\n",{"type":49,"tag":117,"props":274,"children":276},{"class":119,"line":275},4,[277,282,286,290,294,298,303,307,312],{"type":49,"tag":117,"props":278,"children":279},{"style":134},[280],{"type":54,"value":281},"  ROOT",{"type":49,"tag":117,"props":283,"children":284},{"style":140},[285],{"type":54,"value":143},{"type":49,"tag":117,"props":287,"children":288},{"style":140},[289],{"type":54,"value":148},{"type":49,"tag":117,"props":291,"children":292},{"style":134},[293],{"type":54,"value":94},{"type":49,"tag":117,"props":295,"children":296},{"style":140},[297],{"type":54,"value":157},{"type":49,"tag":117,"props":299,"children":300},{"style":134},[301],{"type":54,"value":302},"$HOME",{"type":49,"tag":117,"props":304,"children":305},{"style":140},[306],{"type":54,"value":193},{"type":49,"tag":117,"props":308,"children":309},{"style":134},[310],{"type":54,"value":311},"i4h-workflows",{"type":49,"tag":117,"props":313,"children":314},{"style":140},[315],{"type":54,"value":217},{"type":49,"tag":117,"props":317,"children":319},{"class":119,"line":318},5,[320,325,329,333,337,341,345,350,355,361,366,371,375,379],{"type":49,"tag":117,"props":321,"children":322},{"style":140},[323],{"type":54,"value":324},"  [",{"type":49,"tag":117,"props":326,"children":327},{"style":140},[328],{"type":54,"value":242},{"type":49,"tag":117,"props":330,"children":331},{"style":140},[332],{"type":54,"value":247},{"type":49,"tag":117,"props":334,"children":335},{"style":134},[336],{"type":54,"value":252},{"type":49,"tag":117,"props":338,"children":339},{"style":160},[340],{"type":54,"value":257},{"type":49,"tag":117,"props":342,"children":343},{"style":140},[344],{"type":54,"value":262},{"type":49,"tag":117,"props":346,"children":347},{"style":140},[348],{"type":54,"value":349}," ]",{"type":49,"tag":117,"props":351,"children":352},{"style":140},[353],{"type":54,"value":354}," ||",{"type":49,"tag":117,"props":356,"children":358},{"style":357},"--shiki-light:#E2931D;--shiki-default:#FFCB6B;--shiki-dark:#FFCB6B",[359],{"type":54,"value":360}," git",{"type":49,"tag":117,"props":362,"children":363},{"style":160},[364],{"type":54,"value":365}," clone",{"type":49,"tag":117,"props":367,"children":368},{"style":160},[369],{"type":54,"value":370}," https:\u002F\u002Fgithub.com\u002Fisaac-for-healthcare\u002Fi4h-workflows",{"type":49,"tag":117,"props":372,"children":373},{"style":140},[374],{"type":54,"value":247},{"type":49,"tag":117,"props":376,"children":377},{"style":134},[378],{"type":54,"value":252},{"type":49,"tag":117,"props":380,"children":381},{"style":140},[382],{"type":54,"value":383},"\"\n",{"type":49,"tag":117,"props":385,"children":387},{"class":119,"line":386},6,[388],{"type":49,"tag":117,"props":389,"children":390},{"style":224},[391],{"type":54,"value":392},"fi\n",{"type":49,"tag":117,"props":394,"children":396},{"class":119,"line":395},7,[397,403,408,412,416,420,424,429,435,439,443],{"type":49,"tag":117,"props":398,"children":400},{"style":399},"--shiki-light:#9C3EDA;--shiki-default:#C792EA;--shiki-dark:#C792EA",[401],{"type":54,"value":402},"export",{"type":49,"tag":117,"props":404,"children":405},{"style":134},[406],{"type":54,"value":407}," I4H_WORKFLOWS",{"type":49,"tag":117,"props":409,"children":410},{"style":140},[411],{"type":54,"value":143},{"type":49,"tag":117,"props":413,"children":414},{"style":140},[415],{"type":54,"value":262},{"type":49,"tag":117,"props":417,"children":418},{"style":134},[419],{"type":54,"value":252},{"type":49,"tag":117,"props":421,"children":422},{"style":140},[423],{"type":54,"value":262},{"type":49,"tag":117,"props":425,"children":426},{"style":140},[427],{"type":54,"value":428},";",{"type":49,"tag":117,"props":430,"children":432},{"style":431},"--shiki-light:#6182B8;--shiki-default:#82AAFF;--shiki-dark:#82AAFF",[433],{"type":54,"value":434}," cd",{"type":49,"tag":117,"props":436,"children":437},{"style":140},[438],{"type":54,"value":247},{"type":49,"tag":117,"props":440,"children":441},{"style":134},[442],{"type":54,"value":252},{"type":49,"tag":117,"props":444,"children":445},{"style":140},[446],{"type":54,"value":383},{"type":49,"tag":57,"props":448,"children":450},{"id":449},"basics",[451],{"type":54,"value":452},"Basics",{"type":49,"tag":454,"props":455,"children":456},"ul",{},[457,471,492],{"type":49,"tag":458,"props":459,"children":460},"li",{},[461,463,469],{"type":54,"value":462},"The dataset path must be an existing LeRobot directory with ",{"type":49,"tag":81,"props":464,"children":466},{"className":465},[],[467],{"type":54,"value":468},"meta\u002Finfo.json",{"type":54,"value":470},".",{"type":49,"tag":458,"props":472,"children":473},{},[474,476,482,484,490],{"type":54,"value":475},"Train support is determined by ",{"type":49,"tag":81,"props":477,"children":479},{"className":478},[],[480],{"type":54,"value":481},"policy.train_module",{"type":54,"value":483}," in ",{"type":49,"tag":81,"props":485,"children":487},{"className":486},[],[488],{"type":54,"value":489},"workflows\u002Fagentic\u002Fconfig\u002Fenvironments\u002F\u003Cenv>.yaml",{"type":54,"value":491},". A null value means inference-only.",{"type":49,"tag":458,"props":493,"children":494},{},[495,501],{"type":49,"tag":81,"props":496,"children":498},{"className":497},[],[499],{"type":54,"value":500},"assemble_trocar",{"type":54,"value":502}," is inference-only.",{"type":49,"tag":57,"props":504,"children":506},{"id":505},"stack-map",[507],{"type":54,"value":508},"Stack Map",{"type":49,"tag":510,"props":511,"children":512},"table",{},[513,537],{"type":49,"tag":514,"props":515,"children":516},"thead",{},[517],{"type":49,"tag":518,"props":519,"children":520},"tr",{},[521,527,532],{"type":49,"tag":522,"props":523,"children":524},"th",{},[525],{"type":54,"value":526},"Env",{"type":49,"tag":522,"props":528,"children":529},{},[530],{"type":54,"value":531},"Stack",{"type":49,"tag":522,"props":533,"children":534},{},[535],{"type":54,"value":536},"CLI",{"type":49,"tag":538,"props":539,"children":540},"tbody",{},[541,572,602,630],{"type":49,"tag":518,"props":542,"children":543},{},[544,554,563],{"type":49,"tag":545,"props":546,"children":547},"td",{},[548],{"type":49,"tag":81,"props":549,"children":551},{"className":550},[],[552],{"type":54,"value":553},"scissor_pick_and_place",{"type":49,"tag":545,"props":555,"children":556},{},[557],{"type":49,"tag":81,"props":558,"children":560},{"className":559},[],[561],{"type":54,"value":562},"gr00t_n15",{"type":49,"tag":545,"props":564,"children":565},{},[566],{"type":49,"tag":81,"props":567,"children":569},{"className":568},[],[570],{"type":54,"value":571},"i4h-agentic-gr00t-n15-train",{"type":49,"tag":518,"props":573,"children":574},{},[575,584,593],{"type":49,"tag":545,"props":576,"children":577},{},[578],{"type":49,"tag":81,"props":579,"children":581},{"className":580},[],[582],{"type":54,"value":583},"locomanip_tray_pick_and_place",{"type":49,"tag":545,"props":585,"children":586},{},[587],{"type":49,"tag":81,"props":588,"children":590},{"className":589},[],[591],{"type":54,"value":592},"gr00t_n16",{"type":49,"tag":545,"props":594,"children":595},{},[596],{"type":49,"tag":81,"props":597,"children":599},{"className":598},[],[600],{"type":54,"value":601},"i4h-agentic-gr00t-n16-train",{"type":49,"tag":518,"props":603,"children":604},{},[605,614,622],{"type":49,"tag":545,"props":606,"children":607},{},[608],{"type":49,"tag":81,"props":609,"children":611},{"className":610},[],[612],{"type":54,"value":613},"locomanip_push_cart",{"type":49,"tag":545,"props":615,"children":616},{},[617],{"type":49,"tag":81,"props":618,"children":620},{"className":619},[],[621],{"type":54,"value":592},{"type":49,"tag":545,"props":623,"children":624},{},[625],{"type":49,"tag":81,"props":626,"children":628},{"className":627},[],[629],{"type":54,"value":601},{"type":49,"tag":518,"props":631,"children":632},{},[633,642,651],{"type":49,"tag":545,"props":634,"children":635},{},[636],{"type":49,"tag":81,"props":637,"children":639},{"className":638},[],[640],{"type":54,"value":641},"ultrasound_liver_scan",{"type":49,"tag":545,"props":643,"children":644},{},[645],{"type":49,"tag":81,"props":646,"children":648},{"className":647},[],[649],{"type":54,"value":650},"openpi_pi0",{"type":49,"tag":545,"props":652,"children":653},{},[654],{"type":49,"tag":81,"props":655,"children":657},{"className":656},[],[658],{"type":54,"value":659},"i4h-agentic-openpi-pi0-train",{"type":49,"tag":64,"props":661,"children":662},{},[663,665,671],{"type":54,"value":664},"N1.6 locomanip envs share ",{"type":49,"tag":81,"props":666,"children":668},{"className":667},[],[669],{"type":54,"value":670},"policy.locomanip.train",{"type":54,"value":470},{"type":49,"tag":57,"props":673,"children":675},{"id":674},"preflight",[676],{"type":54,"value":677},"Preflight",{"type":49,"tag":106,"props":679,"children":681},{"className":108,"code":680,"language":110,"meta":111,"style":111},"test -f \"${DATASET_PATH}\u002Fmeta\u002Finfo.json\"\nnvidia-smi --query-gpu=name --format=csv,noheader | wc -l\nworkflows\u002Fagentic\u002Fpolicy\u002F\u003Cstack>\u002Frun.sh --list-envs\n",[682],{"type":49,"tag":81,"props":683,"children":684},{"__ignoreMap":111},[685,722,755],{"type":49,"tag":117,"props":686,"children":687},{"class":119,"line":120},[688,693,698,703,708,713,718],{"type":49,"tag":117,"props":689,"children":690},{"style":431},[691],{"type":54,"value":692},"test",{"type":49,"tag":117,"props":694,"children":695},{"style":160},[696],{"type":54,"value":697}," -f",{"type":49,"tag":117,"props":699,"children":700},{"style":140},[701],{"type":54,"value":702}," \"${",{"type":49,"tag":117,"props":704,"children":705},{"style":134},[706],{"type":54,"value":707},"DATASET_PATH",{"type":49,"tag":117,"props":709,"children":710},{"style":140},[711],{"type":54,"value":712},"}",{"type":49,"tag":117,"props":714,"children":715},{"style":160},[716],{"type":54,"value":717},"\u002Fmeta\u002Finfo.json",{"type":49,"tag":117,"props":719,"children":720},{"style":140},[721],{"type":54,"value":383},{"type":49,"tag":117,"props":723,"children":724},{"class":119,"line":130},[725,730,735,740,745,750],{"type":49,"tag":117,"props":726,"children":727},{"style":357},[728],{"type":54,"value":729},"nvidia-smi",{"type":49,"tag":117,"props":731,"children":732},{"style":160},[733],{"type":54,"value":734}," --query-gpu=name",{"type":49,"tag":117,"props":736,"children":737},{"style":160},[738],{"type":54,"value":739}," --format=csv,noheader",{"type":49,"tag":117,"props":741,"children":742},{"style":140},[743],{"type":54,"value":744}," |",{"type":49,"tag":117,"props":746,"children":747},{"style":357},[748],{"type":54,"value":749}," wc",{"type":49,"tag":117,"props":751,"children":752},{"style":160},[753],{"type":54,"value":754}," -l\n",{"type":49,"tag":117,"props":756,"children":757},{"class":119,"line":220},[758,763,768],{"type":49,"tag":117,"props":759,"children":760},{"style":357},[761],{"type":54,"value":762},"workflows\u002Fagentic\u002Fpolicy\u002F",{"type":49,"tag":117,"props":764,"children":765},{"style":134},[766],{"type":54,"value":767},"\u003Cstack>\u002Frun.sh ",{"type":49,"tag":117,"props":769,"children":770},{"style":160},[771],{"type":54,"value":772},"--list-envs\n",{"type":49,"tag":57,"props":774,"children":776},{"id":775},"run",[777],{"type":54,"value":778},"Run",{"type":49,"tag":64,"props":780,"children":781},{},[782],{"type":54,"value":783},"Run the steps below in order. Each step is a separate bash call; variables persist in the local agent's tmux session.",{"type":49,"tag":785,"props":786,"children":788},"h3",{"id":787},"step-1-setup-and-resolve-dataset",[789],{"type":54,"value":790},"Step 1 — setup and resolve dataset",{"type":49,"tag":106,"props":792,"children":794},{"className":108,"code":793,"language":110,"meta":111,"style":111},"REPO_ROOT=\"${I4H_WORKFLOWS:-$(git rev-parse --show-toplevel 2>\u002Fdev\u002Fnull)}\"; [ -d \"$REPO_ROOT\u002Fworkflows\u002Fagentic\" ] || REPO_ROOT=\"$HOME\u002Fi4h-workflows\"\nENV_ID=scissor_pick_and_place\nSTACK_DIR=gr00t_n15\nTRAIN_CLI=i4h-agentic-gr00t-n15-train\nRUNS_ROOT=\"${REPO_ROOT}\u002Fworkflows\u002Fagentic\u002Fruns\"\n\n# Point DATASET_PATH at a converted LeRobot dataset dir (absolute; must contain meta\u002Finfo.json),\n# produced by [[i4h-workflow-dataset-convert]]. List candidates:\n#   find \"${RUNS_ROOT}\" \"${HF_LEROBOT_HOME:-$HOME\u002F.cache\u002Fhuggingface\u002Flerobot}\" -name info.json -path '*\u002Fmeta\u002F*' -printf '%h\\n' | sed 's#\u002Fmeta$##' | sort -u\nDATASET_PATH=\"${DATASET_PATH:-}\"\nif [ ! -f \"${DATASET_PATH%\u002F}\u002Fmeta\u002Finfo.json\" ]; then\n  echo \"finetune: set DATASET_PATH to a LeRobot dataset dir with meta\u002Finfo.json (got '${DATASET_PATH:-\u003Cunset>}'). Candidates:\" >&2\n  find \"${RUNS_ROOT}\" \"${HF_LEROBOT_HOME:-$HOME\u002F.cache\u002Fhuggingface\u002Flerobot}\" -name info.json -path '*\u002Fmeta\u002F*' -printf '%h\\n' 2>\u002Fdev\u002Fnull | sed 's#\u002Fmeta$##' | sort -u | head\n  exit 1\nfi\n\nRUN_DIR=\"${RUNS_ROOT}\u002Ffinetune_${ENV_ID}_$(date +%Y%m%d_%H%M%S)\"\nOUT=\"${RUN_DIR}\u002Fcheckpoint\"\nexport TMPDIR=\u002Ftmp   # short path: torch DataLoader FD-sharing socket must fit AF_UNIX's 108-byte limit\nmkdir -p \"${OUT}\" \"${RUN_DIR}\u002Flogs\"\nln -sfn \"${RUN_DIR}\" \"${RUNS_ROOT}\u002F.latest\"\n",[795],{"type":49,"tag":81,"props":796,"children":797},{"__ignoreMap":111},[798,934,951,968,985,1018,1027,1035,1044,1053,1078,1127,1190,1366,1381,1389,1397,1463,1497,1524,1571],{"type":49,"tag":117,"props":799,"children":800},{"class":119,"line":120},[801,806,810,814,818,822,826,830,834,838,842,846,850,854,858,862,866,871,875,879,883,887,892,896,900,904,908,913,917,921,925,930],{"type":49,"tag":117,"props":802,"children":803},{"style":134},[804],{"type":54,"value":805},"REPO_ROOT",{"type":49,"tag":117,"props":807,"children":808},{"style":140},[809],{"type":54,"value":143},{"type":49,"tag":117,"props":811,"children":812},{"style":140},[813],{"type":54,"value":148},{"type":49,"tag":117,"props":815,"children":816},{"style":134},[817],{"type":54,"value":94},{"type":49,"tag":117,"props":819,"children":820},{"style":140},[821],{"type":54,"value":157},{"type":49,"tag":117,"props":823,"children":824},{"style":160},[825],{"type":54,"value":163},{"type":49,"tag":117,"props":827,"children":828},{"style":134},[829],{"type":54,"value":168},{"type":49,"tag":117,"props":831,"children":832},{"style":134},[833],{"type":54,"value":173},{"type":49,"tag":117,"props":835,"children":836},{"style":134},[837],{"type":54,"value":178},{"type":49,"tag":117,"props":839,"children":840},{"style":134},[841],{"type":54,"value":183},{"type":49,"tag":117,"props":843,"children":844},{"style":160},[845],{"type":54,"value":188},{"type":49,"tag":117,"props":847,"children":848},{"style":140},[849],{"type":54,"value":193},{"type":49,"tag":117,"props":851,"children":852},{"style":134},[853],{"type":54,"value":198},{"type":49,"tag":117,"props":855,"children":856},{"style":140},[857],{"type":54,"value":193},{"type":49,"tag":117,"props":859,"children":860},{"style":134},[861],{"type":54,"value":207},{"type":49,"tag":117,"props":863,"children":864},{"style":160},[865],{"type":54,"value":212},{"type":49,"tag":117,"props":867,"children":868},{"style":140},[869],{"type":54,"value":870},"}\"",{"type":49,"tag":117,"props":872,"children":873},{"style":140},[874],{"type":54,"value":428},{"type":49,"tag":117,"props":876,"children":877},{"style":140},[878],{"type":54,"value":232},{"type":49,"tag":117,"props":880,"children":881},{"style":140},[882],{"type":54,"value":242},{"type":49,"tag":117,"props":884,"children":885},{"style":140},[886],{"type":54,"value":247},{"type":49,"tag":117,"props":888,"children":889},{"style":134},[890],{"type":54,"value":891},"$REPO_ROOT",{"type":49,"tag":117,"props":893,"children":894},{"style":160},[895],{"type":54,"value":257},{"type":49,"tag":117,"props":897,"children":898},{"style":140},[899],{"type":54,"value":262},{"type":49,"tag":117,"props":901,"children":902},{"style":140},[903],{"type":54,"value":349},{"type":49,"tag":117,"props":905,"children":906},{"style":140},[907],{"type":54,"value":354},{"type":49,"tag":117,"props":909,"children":910},{"style":134},[911],{"type":54,"value":912}," REPO_ROOT",{"type":49,"tag":117,"props":914,"children":915},{"style":140},[916],{"type":54,"value":143},{"type":49,"tag":117,"props":918,"children":919},{"style":140},[920],{"type":54,"value":262},{"type":49,"tag":117,"props":922,"children":923},{"style":134},[924],{"type":54,"value":302},{"type":49,"tag":117,"props":926,"children":927},{"style":160},[928],{"type":54,"value":929},"\u002Fi4h-workflows",{"type":49,"tag":117,"props":931,"children":932},{"style":140},[933],{"type":54,"value":383},{"type":49,"tag":117,"props":935,"children":936},{"class":119,"line":130},[937,942,946],{"type":49,"tag":117,"props":938,"children":939},{"style":134},[940],{"type":54,"value":941},"ENV_ID",{"type":49,"tag":117,"props":943,"children":944},{"style":140},[945],{"type":54,"value":143},{"type":49,"tag":117,"props":947,"children":948},{"style":160},[949],{"type":54,"value":950},"scissor_pick_and_place\n",{"type":49,"tag":117,"props":952,"children":953},{"class":119,"line":220},[954,959,963],{"type":49,"tag":117,"props":955,"children":956},{"style":134},[957],{"type":54,"value":958},"STACK_DIR",{"type":49,"tag":117,"props":960,"children":961},{"style":140},[962],{"type":54,"value":143},{"type":49,"tag":117,"props":964,"children":965},{"style":160},[966],{"type":54,"value":967},"gr00t_n15\n",{"type":49,"tag":117,"props":969,"children":970},{"class":119,"line":275},[971,976,980],{"type":49,"tag":117,"props":972,"children":973},{"style":134},[974],{"type":54,"value":975},"TRAIN_CLI",{"type":49,"tag":117,"props":977,"children":978},{"style":140},[979],{"type":54,"value":143},{"type":49,"tag":117,"props":981,"children":982},{"style":160},[983],{"type":54,"value":984},"i4h-agentic-gr00t-n15-train\n",{"type":49,"tag":117,"props":986,"children":987},{"class":119,"line":318},[988,993,997,1001,1005,1009,1014],{"type":49,"tag":117,"props":989,"children":990},{"style":134},[991],{"type":54,"value":992},"RUNS_ROOT",{"type":49,"tag":117,"props":994,"children":995},{"style":140},[996],{"type":54,"value":143},{"type":49,"tag":117,"props":998,"children":999},{"style":140},[1000],{"type":54,"value":148},{"type":49,"tag":117,"props":1002,"children":1003},{"style":134},[1004],{"type":54,"value":805},{"type":49,"tag":117,"props":1006,"children":1007},{"style":140},[1008],{"type":54,"value":712},{"type":49,"tag":117,"props":1010,"children":1011},{"style":160},[1012],{"type":54,"value":1013},"\u002Fworkflows\u002Fagentic\u002Fruns",{"type":49,"tag":117,"props":1015,"children":1016},{"style":140},[1017],{"type":54,"value":383},{"type":49,"tag":117,"props":1019,"children":1020},{"class":119,"line":386},[1021],{"type":49,"tag":117,"props":1022,"children":1024},{"emptyLinePlaceholder":1023},true,[1025],{"type":54,"value":1026},"\n",{"type":49,"tag":117,"props":1028,"children":1029},{"class":119,"line":395},[1030],{"type":49,"tag":117,"props":1031,"children":1032},{"style":124},[1033],{"type":54,"value":1034},"# Point DATASET_PATH at a converted LeRobot dataset dir (absolute; must contain meta\u002Finfo.json),\n",{"type":49,"tag":117,"props":1036,"children":1038},{"class":119,"line":1037},8,[1039],{"type":49,"tag":117,"props":1040,"children":1041},{"style":124},[1042],{"type":54,"value":1043},"# produced by [[i4h-workflow-dataset-convert]]. List candidates:\n",{"type":49,"tag":117,"props":1045,"children":1047},{"class":119,"line":1046},9,[1048],{"type":49,"tag":117,"props":1049,"children":1050},{"style":124},[1051],{"type":54,"value":1052},"#   find \"${RUNS_ROOT}\" \"${HF_LEROBOT_HOME:-$HOME\u002F.cache\u002Fhuggingface\u002Flerobot}\" -name info.json -path '*\u002Fmeta\u002F*' -printf '%h\\n' | sed 's#\u002Fmeta$##' | sort -u\n",{"type":49,"tag":117,"props":1054,"children":1056},{"class":119,"line":1055},10,[1057,1061,1065,1069,1073],{"type":49,"tag":117,"props":1058,"children":1059},{"style":134},[1060],{"type":54,"value":707},{"type":49,"tag":117,"props":1062,"children":1063},{"style":140},[1064],{"type":54,"value":143},{"type":49,"tag":117,"props":1066,"children":1067},{"style":140},[1068],{"type":54,"value":148},{"type":49,"tag":117,"props":1070,"children":1071},{"style":134},[1072],{"type":54,"value":707},{"type":49,"tag":117,"props":1074,"children":1075},{"style":140},[1076],{"type":54,"value":1077},":-}\"\n",{"type":49,"tag":117,"props":1079,"children":1081},{"class":119,"line":1080},11,[1082,1086,1090,1094,1098,1102,1106,1111,1115,1119,1123],{"type":49,"tag":117,"props":1083,"children":1084},{"style":224},[1085],{"type":54,"value":227},{"type":49,"tag":117,"props":1087,"children":1088},{"style":140},[1089],{"type":54,"value":232},{"type":49,"tag":117,"props":1091,"children":1092},{"style":140},[1093],{"type":54,"value":237},{"type":49,"tag":117,"props":1095,"children":1096},{"style":140},[1097],{"type":54,"value":697},{"type":49,"tag":117,"props":1099,"children":1100},{"style":140},[1101],{"type":54,"value":702},{"type":49,"tag":117,"props":1103,"children":1104},{"style":134},[1105],{"type":54,"value":707},{"type":49,"tag":117,"props":1107,"children":1108},{"style":140},[1109],{"type":54,"value":1110},"%\u002F}",{"type":49,"tag":117,"props":1112,"children":1113},{"style":160},[1114],{"type":54,"value":717},{"type":49,"tag":117,"props":1116,"children":1117},{"style":140},[1118],{"type":54,"value":262},{"type":49,"tag":117,"props":1120,"children":1121},{"style":140},[1122],{"type":54,"value":267},{"type":49,"tag":117,"props":1124,"children":1125},{"style":224},[1126],{"type":54,"value":272},{"type":49,"tag":117,"props":1128,"children":1130},{"class":119,"line":1129},12,[1131,1136,1140,1145,1150,1154,1158,1163,1168,1172,1176,1181,1185],{"type":49,"tag":117,"props":1132,"children":1133},{"style":431},[1134],{"type":54,"value":1135},"  echo",{"type":49,"tag":117,"props":1137,"children":1138},{"style":140},[1139],{"type":54,"value":247},{"type":49,"tag":117,"props":1141,"children":1142},{"style":160},[1143],{"type":54,"value":1144},"finetune: set DATASET_PATH to a LeRobot dataset dir with meta\u002Finfo.json (got '",{"type":49,"tag":117,"props":1146,"children":1147},{"style":140},[1148],{"type":54,"value":1149},"${",{"type":49,"tag":117,"props":1151,"children":1152},{"style":134},[1153],{"type":54,"value":707},{"type":49,"tag":117,"props":1155,"children":1156},{"style":140},[1157],{"type":54,"value":157},{"type":49,"tag":117,"props":1159,"children":1160},{"style":160},[1161],{"type":54,"value":1162},"\u003C",{"type":49,"tag":117,"props":1164,"children":1165},{"style":134},[1166],{"type":54,"value":1167},"unset",{"type":49,"tag":117,"props":1169,"children":1170},{"style":160},[1171],{"type":54,"value":188},{"type":49,"tag":117,"props":1173,"children":1174},{"style":140},[1175],{"type":54,"value":712},{"type":49,"tag":117,"props":1177,"children":1178},{"style":160},[1179],{"type":54,"value":1180},"'). Candidates:",{"type":49,"tag":117,"props":1182,"children":1183},{"style":140},[1184],{"type":54,"value":262},{"type":49,"tag":117,"props":1186,"children":1187},{"style":140},[1188],{"type":54,"value":1189}," >&2\n",{"type":49,"tag":117,"props":1191,"children":1193},{"class":119,"line":1192},13,[1194,1199,1203,1207,1211,1215,1220,1224,1228,1232,1236,1241,1245,1250,1254,1259,1263,1268,1273,1278,1283,1288,1293,1298,1302,1307,1311,1316,1321,1325,1330,1334,1339,1343,1347,1352,1357,1361],{"type":49,"tag":117,"props":1195,"children":1196},{"style":357},[1197],{"type":54,"value":1198},"  find",{"type":49,"tag":117,"props":1200,"children":1201},{"style":140},[1202],{"type":54,"value":702},{"type":49,"tag":117,"props":1204,"children":1205},{"style":134},[1206],{"type":54,"value":992},{"type":49,"tag":117,"props":1208,"children":1209},{"style":140},[1210],{"type":54,"value":870},{"type":49,"tag":117,"props":1212,"children":1213},{"style":140},[1214],{"type":54,"value":702},{"type":49,"tag":117,"props":1216,"children":1217},{"style":134},[1218],{"type":54,"value":1219},"HF_LEROBOT_HOME",{"type":49,"tag":117,"props":1221,"children":1222},{"style":140},[1223],{"type":54,"value":157},{"type":49,"tag":117,"props":1225,"children":1226},{"style":134},[1227],{"type":54,"value":302},{"type":49,"tag":117,"props":1229,"children":1230},{"style":140},[1231],{"type":54,"value":193},{"type":49,"tag":117,"props":1233,"children":1234},{"style":160},[1235],{"type":54,"value":470},{"type":49,"tag":117,"props":1237,"children":1238},{"style":134},[1239],{"type":54,"value":1240},"cache",{"type":49,"tag":117,"props":1242,"children":1243},{"style":140},[1244],{"type":54,"value":193},{"type":49,"tag":117,"props":1246,"children":1247},{"style":134},[1248],{"type":54,"value":1249},"huggingface",{"type":49,"tag":117,"props":1251,"children":1252},{"style":140},[1253],{"type":54,"value":193},{"type":49,"tag":117,"props":1255,"children":1256},{"style":134},[1257],{"type":54,"value":1258},"lerobot",{"type":49,"tag":117,"props":1260,"children":1261},{"style":140},[1262],{"type":54,"value":870},{"type":49,"tag":117,"props":1264,"children":1265},{"style":160},[1266],{"type":54,"value":1267}," -name",{"type":49,"tag":117,"props":1269,"children":1270},{"style":160},[1271],{"type":54,"value":1272}," info.json",{"type":49,"tag":117,"props":1274,"children":1275},{"style":160},[1276],{"type":54,"value":1277}," -path",{"type":49,"tag":117,"props":1279,"children":1280},{"style":140},[1281],{"type":54,"value":1282}," '",{"type":49,"tag":117,"props":1284,"children":1285},{"style":160},[1286],{"type":54,"value":1287},"*\u002Fmeta\u002F*",{"type":49,"tag":117,"props":1289,"children":1290},{"style":140},[1291],{"type":54,"value":1292},"'",{"type":49,"tag":117,"props":1294,"children":1295},{"style":160},[1296],{"type":54,"value":1297}," -printf",{"type":49,"tag":117,"props":1299,"children":1300},{"style":140},[1301],{"type":54,"value":1282},{"type":49,"tag":117,"props":1303,"children":1304},{"style":160},[1305],{"type":54,"value":1306},"%h\\n",{"type":49,"tag":117,"props":1308,"children":1309},{"style":140},[1310],{"type":54,"value":1292},{"type":49,"tag":117,"props":1312,"children":1313},{"style":140},[1314],{"type":54,"value":1315}," 2>",{"type":49,"tag":117,"props":1317,"children":1318},{"style":160},[1319],{"type":54,"value":1320},"\u002Fdev\u002Fnull",{"type":49,"tag":117,"props":1322,"children":1323},{"style":140},[1324],{"type":54,"value":744},{"type":49,"tag":117,"props":1326,"children":1327},{"style":357},[1328],{"type":54,"value":1329}," sed",{"type":49,"tag":117,"props":1331,"children":1332},{"style":140},[1333],{"type":54,"value":1282},{"type":49,"tag":117,"props":1335,"children":1336},{"style":160},[1337],{"type":54,"value":1338},"s#\u002Fmeta$##",{"type":49,"tag":117,"props":1340,"children":1341},{"style":140},[1342],{"type":54,"value":1292},{"type":49,"tag":117,"props":1344,"children":1345},{"style":140},[1346],{"type":54,"value":744},{"type":49,"tag":117,"props":1348,"children":1349},{"style":357},[1350],{"type":54,"value":1351}," sort",{"type":49,"tag":117,"props":1353,"children":1354},{"style":160},[1355],{"type":54,"value":1356}," -u",{"type":49,"tag":117,"props":1358,"children":1359},{"style":140},[1360],{"type":54,"value":744},{"type":49,"tag":117,"props":1362,"children":1363},{"style":357},[1364],{"type":54,"value":1365}," head\n",{"type":49,"tag":117,"props":1367,"children":1369},{"class":119,"line":1368},14,[1370,1375],{"type":49,"tag":117,"props":1371,"children":1372},{"style":431},[1373],{"type":54,"value":1374},"  exit",{"type":49,"tag":117,"props":1376,"children":1378},{"style":1377},"--shiki-light:#F76D47;--shiki-default:#F78C6C;--shiki-dark:#F78C6C",[1379],{"type":54,"value":1380}," 1\n",{"type":49,"tag":117,"props":1382,"children":1384},{"class":119,"line":1383},15,[1385],{"type":49,"tag":117,"props":1386,"children":1387},{"style":224},[1388],{"type":54,"value":392},{"type":49,"tag":117,"props":1390,"children":1392},{"class":119,"line":1391},16,[1393],{"type":49,"tag":117,"props":1394,"children":1395},{"emptyLinePlaceholder":1023},[1396],{"type":54,"value":1026},{"type":49,"tag":117,"props":1398,"children":1400},{"class":119,"line":1399},17,[1401,1406,1410,1414,1418,1422,1427,1431,1435,1439,1444,1448,1453,1458],{"type":49,"tag":117,"props":1402,"children":1403},{"style":134},[1404],{"type":54,"value":1405},"RUN_DIR",{"type":49,"tag":117,"props":1407,"children":1408},{"style":140},[1409],{"type":54,"value":143},{"type":49,"tag":117,"props":1411,"children":1412},{"style":140},[1413],{"type":54,"value":148},{"type":49,"tag":117,"props":1415,"children":1416},{"style":134},[1417],{"type":54,"value":992},{"type":49,"tag":117,"props":1419,"children":1420},{"style":140},[1421],{"type":54,"value":712},{"type":49,"tag":117,"props":1423,"children":1424},{"style":160},[1425],{"type":54,"value":1426},"\u002Ffinetune_",{"type":49,"tag":117,"props":1428,"children":1429},{"style":140},[1430],{"type":54,"value":1149},{"type":49,"tag":117,"props":1432,"children":1433},{"style":134},[1434],{"type":54,"value":941},{"type":49,"tag":117,"props":1436,"children":1437},{"style":140},[1438],{"type":54,"value":712},{"type":49,"tag":117,"props":1440,"children":1441},{"style":160},[1442],{"type":54,"value":1443},"_",{"type":49,"tag":117,"props":1445,"children":1446},{"style":140},[1447],{"type":54,"value":163},{"type":49,"tag":117,"props":1449,"children":1450},{"style":357},[1451],{"type":54,"value":1452},"date",{"type":49,"tag":117,"props":1454,"children":1455},{"style":160},[1456],{"type":54,"value":1457}," +%Y%m%d_%H%M%S",{"type":49,"tag":117,"props":1459,"children":1460},{"style":140},[1461],{"type":54,"value":1462},")\"\n",{"type":49,"tag":117,"props":1464,"children":1466},{"class":119,"line":1465},18,[1467,1472,1476,1480,1484,1488,1493],{"type":49,"tag":117,"props":1468,"children":1469},{"style":134},[1470],{"type":54,"value":1471},"OUT",{"type":49,"tag":117,"props":1473,"children":1474},{"style":140},[1475],{"type":54,"value":143},{"type":49,"tag":117,"props":1477,"children":1478},{"style":140},[1479],{"type":54,"value":148},{"type":49,"tag":117,"props":1481,"children":1482},{"style":134},[1483],{"type":54,"value":1405},{"type":49,"tag":117,"props":1485,"children":1486},{"style":140},[1487],{"type":54,"value":712},{"type":49,"tag":117,"props":1489,"children":1490},{"style":160},[1491],{"type":54,"value":1492},"\u002Fcheckpoint",{"type":49,"tag":117,"props":1494,"children":1495},{"style":140},[1496],{"type":54,"value":383},{"type":49,"tag":117,"props":1498,"children":1500},{"class":119,"line":1499},19,[1501,1505,1510,1514,1519],{"type":49,"tag":117,"props":1502,"children":1503},{"style":399},[1504],{"type":54,"value":402},{"type":49,"tag":117,"props":1506,"children":1507},{"style":134},[1508],{"type":54,"value":1509}," TMPDIR",{"type":49,"tag":117,"props":1511,"children":1512},{"style":140},[1513],{"type":54,"value":143},{"type":49,"tag":117,"props":1515,"children":1516},{"style":134},[1517],{"type":54,"value":1518},"\u002Ftmp   ",{"type":49,"tag":117,"props":1520,"children":1521},{"style":124},[1522],{"type":54,"value":1523},"# short path: torch DataLoader FD-sharing socket must fit AF_UNIX's 108-byte limit\n",{"type":49,"tag":117,"props":1525,"children":1527},{"class":119,"line":1526},20,[1528,1533,1538,1542,1546,1550,1554,1558,1562,1567],{"type":49,"tag":117,"props":1529,"children":1530},{"style":357},[1531],{"type":54,"value":1532},"mkdir",{"type":49,"tag":117,"props":1534,"children":1535},{"style":160},[1536],{"type":54,"value":1537}," -p",{"type":49,"tag":117,"props":1539,"children":1540},{"style":140},[1541],{"type":54,"value":702},{"type":49,"tag":117,"props":1543,"children":1544},{"style":134},[1545],{"type":54,"value":1471},{"type":49,"tag":117,"props":1547,"children":1548},{"style":140},[1549],{"type":54,"value":870},{"type":49,"tag":117,"props":1551,"children":1552},{"style":140},[1553],{"type":54,"value":702},{"type":49,"tag":117,"props":1555,"children":1556},{"style":134},[1557],{"type":54,"value":1405},{"type":49,"tag":117,"props":1559,"children":1560},{"style":140},[1561],{"type":54,"value":712},{"type":49,"tag":117,"props":1563,"children":1564},{"style":160},[1565],{"type":54,"value":1566},"\u002Flogs",{"type":49,"tag":117,"props":1568,"children":1569},{"style":140},[1570],{"type":54,"value":383},{"type":49,"tag":117,"props":1572,"children":1574},{"class":119,"line":1573},21,[1575,1580,1585,1589,1593,1597,1601,1605,1609,1614],{"type":49,"tag":117,"props":1576,"children":1577},{"style":357},[1578],{"type":54,"value":1579},"ln",{"type":49,"tag":117,"props":1581,"children":1582},{"style":160},[1583],{"type":54,"value":1584}," -sfn",{"type":49,"tag":117,"props":1586,"children":1587},{"style":140},[1588],{"type":54,"value":702},{"type":49,"tag":117,"props":1590,"children":1591},{"style":134},[1592],{"type":54,"value":1405},{"type":49,"tag":117,"props":1594,"children":1595},{"style":140},[1596],{"type":54,"value":870},{"type":49,"tag":117,"props":1598,"children":1599},{"style":140},[1600],{"type":54,"value":702},{"type":49,"tag":117,"props":1602,"children":1603},{"style":134},[1604],{"type":54,"value":992},{"type":49,"tag":117,"props":1606,"children":1607},{"style":140},[1608],{"type":54,"value":712},{"type":49,"tag":117,"props":1610,"children":1611},{"style":160},[1612],{"type":54,"value":1613},"\u002F.latest",{"type":49,"tag":117,"props":1615,"children":1616},{"style":140},[1617],{"type":54,"value":383},{"type":49,"tag":785,"props":1619,"children":1621},{"id":1620},"step-2-train",[1622],{"type":54,"value":1623},"Step 2 — train",{"type":49,"tag":106,"props":1625,"children":1627},{"className":108,"code":1626,"language":110,"meta":111,"style":111},"uv --directory \"${REPO_ROOT}\u002Fworkflows\u002Fagentic\u002Fpolicy\u002F${STACK_DIR}\" run \"${TRAIN_CLI}\" \\\n  --env \"${ENV_ID}\" \\\n  --dataset-path \"${DATASET_PATH}\" \\\n  --output-dir \"${OUT}\" \\\n  --max-steps 1000 \\\n  --save-steps 1000 \\\n  --num-gpus 1 \\\n  2>&1 | tee \"${RUN_DIR}\u002Flogs\u002Ffinetune.log\"\n",[1628],{"type":49,"tag":81,"props":1629,"children":1630},{"__ignoreMap":111},[1631,1695,1719,1743,1767,1784,1800,1817],{"type":49,"tag":117,"props":1632,"children":1633},{"class":119,"line":120},[1634,1639,1644,1648,1652,1656,1661,1665,1669,1673,1678,1682,1686,1690],{"type":49,"tag":117,"props":1635,"children":1636},{"style":357},[1637],{"type":54,"value":1638},"uv",{"type":49,"tag":117,"props":1640,"children":1641},{"style":160},[1642],{"type":54,"value":1643}," --directory",{"type":49,"tag":117,"props":1645,"children":1646},{"style":140},[1647],{"type":54,"value":702},{"type":49,"tag":117,"props":1649,"children":1650},{"style":134},[1651],{"type":54,"value":805},{"type":49,"tag":117,"props":1653,"children":1654},{"style":140},[1655],{"type":54,"value":712},{"type":49,"tag":117,"props":1657,"children":1658},{"style":160},[1659],{"type":54,"value":1660},"\u002Fworkflows\u002Fagentic\u002Fpolicy\u002F",{"type":49,"tag":117,"props":1662,"children":1663},{"style":140},[1664],{"type":54,"value":1149},{"type":49,"tag":117,"props":1666,"children":1667},{"style":134},[1668],{"type":54,"value":958},{"type":49,"tag":117,"props":1670,"children":1671},{"style":140},[1672],{"type":54,"value":870},{"type":49,"tag":117,"props":1674,"children":1675},{"style":160},[1676],{"type":54,"value":1677}," run",{"type":49,"tag":117,"props":1679,"children":1680},{"style":140},[1681],{"type":54,"value":702},{"type":49,"tag":117,"props":1683,"children":1684},{"style":134},[1685],{"type":54,"value":975},{"type":49,"tag":117,"props":1687,"children":1688},{"style":140},[1689],{"type":54,"value":870},{"type":49,"tag":117,"props":1691,"children":1692},{"style":134},[1693],{"type":54,"value":1694}," \\\n",{"type":49,"tag":117,"props":1696,"children":1697},{"class":119,"line":130},[1698,1703,1707,1711,1715],{"type":49,"tag":117,"props":1699,"children":1700},{"style":160},[1701],{"type":54,"value":1702},"  --env",{"type":49,"tag":117,"props":1704,"children":1705},{"style":140},[1706],{"type":54,"value":702},{"type":49,"tag":117,"props":1708,"children":1709},{"style":134},[1710],{"type":54,"value":941},{"type":49,"tag":117,"props":1712,"children":1713},{"style":140},[1714],{"type":54,"value":870},{"type":49,"tag":117,"props":1716,"children":1717},{"style":134},[1718],{"type":54,"value":1694},{"type":49,"tag":117,"props":1720,"children":1721},{"class":119,"line":220},[1722,1727,1731,1735,1739],{"type":49,"tag":117,"props":1723,"children":1724},{"style":160},[1725],{"type":54,"value":1726},"  --dataset-path",{"type":49,"tag":117,"props":1728,"children":1729},{"style":140},[1730],{"type":54,"value":702},{"type":49,"tag":117,"props":1732,"children":1733},{"style":134},[1734],{"type":54,"value":707},{"type":49,"tag":117,"props":1736,"children":1737},{"style":140},[1738],{"type":54,"value":870},{"type":49,"tag":117,"props":1740,"children":1741},{"style":134},[1742],{"type":54,"value":1694},{"type":49,"tag":117,"props":1744,"children":1745},{"class":119,"line":275},[1746,1751,1755,1759,1763],{"type":49,"tag":117,"props":1747,"children":1748},{"style":160},[1749],{"type":54,"value":1750},"  --output-dir",{"type":49,"tag":117,"props":1752,"children":1753},{"style":140},[1754],{"type":54,"value":702},{"type":49,"tag":117,"props":1756,"children":1757},{"style":134},[1758],{"type":54,"value":1471},{"type":49,"tag":117,"props":1760,"children":1761},{"style":140},[1762],{"type":54,"value":870},{"type":49,"tag":117,"props":1764,"children":1765},{"style":134},[1766],{"type":54,"value":1694},{"type":49,"tag":117,"props":1768,"children":1769},{"class":119,"line":318},[1770,1775,1780],{"type":49,"tag":117,"props":1771,"children":1772},{"style":160},[1773],{"type":54,"value":1774},"  --max-steps",{"type":49,"tag":117,"props":1776,"children":1777},{"style":1377},[1778],{"type":54,"value":1779}," 1000",{"type":49,"tag":117,"props":1781,"children":1782},{"style":134},[1783],{"type":54,"value":1694},{"type":49,"tag":117,"props":1785,"children":1786},{"class":119,"line":386},[1787,1792,1796],{"type":49,"tag":117,"props":1788,"children":1789},{"style":160},[1790],{"type":54,"value":1791},"  --save-steps",{"type":49,"tag":117,"props":1793,"children":1794},{"style":1377},[1795],{"type":54,"value":1779},{"type":49,"tag":117,"props":1797,"children":1798},{"style":134},[1799],{"type":54,"value":1694},{"type":49,"tag":117,"props":1801,"children":1802},{"class":119,"line":395},[1803,1808,1813],{"type":49,"tag":117,"props":1804,"children":1805},{"style":160},[1806],{"type":54,"value":1807},"  --num-gpus",{"type":49,"tag":117,"props":1809,"children":1810},{"style":1377},[1811],{"type":54,"value":1812}," 1",{"type":49,"tag":117,"props":1814,"children":1815},{"style":134},[1816],{"type":54,"value":1694},{"type":49,"tag":117,"props":1818,"children":1819},{"class":119,"line":1037},[1820,1825,1829,1834,1838,1842,1846,1851],{"type":49,"tag":117,"props":1821,"children":1822},{"style":140},[1823],{"type":54,"value":1824},"  2>&1",{"type":49,"tag":117,"props":1826,"children":1827},{"style":140},[1828],{"type":54,"value":744},{"type":49,"tag":117,"props":1830,"children":1831},{"style":357},[1832],{"type":54,"value":1833}," tee",{"type":49,"tag":117,"props":1835,"children":1836},{"style":140},[1837],{"type":54,"value":702},{"type":49,"tag":117,"props":1839,"children":1840},{"style":134},[1841],{"type":54,"value":1405},{"type":49,"tag":117,"props":1843,"children":1844},{"style":140},[1845],{"type":54,"value":712},{"type":49,"tag":117,"props":1847,"children":1848},{"style":160},[1849],{"type":54,"value":1850},"\u002Flogs\u002Ffinetune.log",{"type":49,"tag":117,"props":1852,"children":1853},{"style":140},[1854],{"type":54,"value":383},{"type":49,"tag":64,"props":1856,"children":1857},{},[1858,1860,1866,1868,1874],{"type":54,"value":1859},"Tyro flags use kebab case (",{"type":49,"tag":81,"props":1861,"children":1863},{"className":1862},[],[1864],{"type":54,"value":1865},"--max-steps",{"type":54,"value":1867},", not ",{"type":49,"tag":81,"props":1869,"children":1871},{"className":1870},[],[1872],{"type":54,"value":1873},"--max_steps",{"type":54,"value":1875},").",{"type":49,"tag":57,"props":1877,"children":1879},{"id":1878},"common-flags",[1880],{"type":54,"value":1881},"Common Flags",{"type":49,"tag":454,"props":1883,"children":1884},{},[1885,1896,1905,1922,1939,1955,1966,1977],{"type":49,"tag":458,"props":1886,"children":1887},{},[1888,1894],{"type":49,"tag":81,"props":1889,"children":1891},{"className":1890},[],[1892],{"type":54,"value":1893},"--dataset-path PATH",{"type":54,"value":1895}," (required)",{"type":49,"tag":458,"props":1897,"children":1898},{},[1899],{"type":49,"tag":81,"props":1900,"children":1902},{"className":1901},[],[1903],{"type":54,"value":1904},"--output-dir PATH",{"type":49,"tag":458,"props":1906,"children":1907},{},[1908,1914,1916],{"type":49,"tag":81,"props":1909,"children":1911},{"className":1910},[],[1912],{"type":54,"value":1913},"--base-model-path PATH_OR_REPO",{"type":54,"value":1915}," overrides YAML ",{"type":49,"tag":81,"props":1917,"children":1919},{"className":1918},[],[1920],{"type":54,"value":1921},"policy.model_repo",{"type":49,"tag":458,"props":1923,"children":1924},{},[1925,1931,1933],{"type":49,"tag":81,"props":1926,"children":1928},{"className":1927},[],[1929],{"type":54,"value":1930},"--max-steps N",{"type":54,"value":1932},", ",{"type":49,"tag":81,"props":1934,"children":1936},{"className":1935},[],[1937],{"type":54,"value":1938},"--save-steps N",{"type":49,"tag":458,"props":1940,"children":1941},{},[1942,1948,1949],{"type":49,"tag":81,"props":1943,"children":1945},{"className":1944},[],[1946],{"type":54,"value":1947},"--batch-size N",{"type":54,"value":1932},{"type":49,"tag":81,"props":1950,"children":1952},{"className":1951},[],[1953],{"type":54,"value":1954},"--learning-rate FLOAT",{"type":49,"tag":458,"props":1956,"children":1957},{},[1958,1964],{"type":49,"tag":81,"props":1959,"children":1961},{"className":1960},[],[1962],{"type":54,"value":1963},"--no-tune-visual",{"type":54,"value":1965}," — freeze the vision backbone (trains the action head + projector only): ~2× faster, ~half the memory, less overfitting. Good default for small datasets; unfreeze only with lots of data + a real visual domain gap.",{"type":49,"tag":458,"props":1967,"children":1968},{},[1969,1975],{"type":49,"tag":81,"props":1970,"children":1972},{"className":1971},[],[1973],{"type":54,"value":1974},"--num-gpus N",{"type":54,"value":1976}," — must not exceed visible GPUs",{"type":49,"tag":458,"props":1978,"children":1979},{},[1980],{"type":49,"tag":81,"props":1981,"children":1983},{"className":1982},[],[1984],{"type":54,"value":1985},"--report-to tensorboard|wandb",{"type":49,"tag":57,"props":1987,"children":1989},{"id":1988},"verify",[1990],{"type":54,"value":1991},"Verify",{"type":49,"tag":454,"props":1993,"children":1994},{},[1995,2029],{"type":49,"tag":458,"props":1996,"children":1997},{},[1998,2000,2006,2008,2014,2015,2021,2022,2028],{"type":54,"value":1999},"Checkpoint directory ",{"type":49,"tag":81,"props":2001,"children":2003},{"className":2002},[],[2004],{"type":54,"value":2005},"${OUT}\u002Fcheckpoint-\u003CN>",{"type":54,"value":2007}," contains ",{"type":49,"tag":81,"props":2009,"children":2011},{"className":2010},[],[2012],{"type":54,"value":2013},"model-0000*-of-*.safetensors",{"type":54,"value":1932},{"type":49,"tag":81,"props":2016,"children":2018},{"className":2017},[],[2019],{"type":54,"value":2020},"experiment_cfg\u002F",{"type":54,"value":1932},{"type":49,"tag":81,"props":2023,"children":2025},{"className":2024},[],[2026],{"type":54,"value":2027},"processor\u002F",{"type":54,"value":470},{"type":49,"tag":458,"props":2030,"children":2031},{},[2032,2034,2040,2042,2048],{"type":54,"value":2033},"Log contains ",{"type":49,"tag":81,"props":2035,"children":2037},{"className":2036},[],[2038],{"type":54,"value":2039},"train_loss",{"type":54,"value":2041}," lines and a final ",{"type":49,"tag":81,"props":2043,"children":2045},{"className":2044},[],[2046],{"type":54,"value":2047},"'train_runtime': ...",{"type":54,"value":2049}," summary.",{"type":49,"tag":57,"props":2051,"children":2053},{"id":2052},"prerequisites",[2054],{"type":54,"value":2055},"Prerequisites",{"type":49,"tag":454,"props":2057,"children":2058},{},[2059,2072,2083,2109],{"type":49,"tag":458,"props":2060,"children":2061},{},[2062,2064,2070],{"type":54,"value":2063},"Workflow set up via [[i4h-workflow-setup]] (the stack's ",{"type":49,"tag":81,"props":2065,"children":2067},{"className":2066},[],[2068],{"type":54,"value":2069},".venv",{"type":54,"value":2071}," must exist).",{"type":49,"tag":458,"props":2073,"children":2074},{},[2075,2077,2082],{"type":54,"value":2076},"An existing LeRobot dataset directory with ",{"type":49,"tag":81,"props":2078,"children":2080},{"className":2079},[],[2081],{"type":54,"value":468},{"type":54,"value":470},{"type":49,"tag":458,"props":2084,"children":2085},{},[2086,2088,2093,2095,2100,2102,2107],{"type":54,"value":2087},"A train-capable env: ",{"type":49,"tag":81,"props":2089,"children":2091},{"className":2090},[],[2092],{"type":54,"value":481},{"type":54,"value":2094}," non-null in ",{"type":49,"tag":81,"props":2096,"children":2098},{"className":2097},[],[2099],{"type":54,"value":489},{"type":54,"value":2101}," (",{"type":49,"tag":81,"props":2103,"children":2105},{"className":2104},[],[2106],{"type":54,"value":500},{"type":54,"value":2108}," is inference-only).",{"type":49,"tag":458,"props":2110,"children":2111},{},[2112,2114,2120],{"type":54,"value":2113},"At least one visible GPU (",{"type":49,"tag":81,"props":2115,"children":2117},{"className":2116},[],[2118],{"type":54,"value":2119},"--num-gpus",{"type":54,"value":2121}," must not exceed visible GPUs).",{"type":49,"tag":57,"props":2123,"children":2125},{"id":2124},"limitations",[2126],{"type":54,"value":2127},"Limitations",{"type":49,"tag":454,"props":2129,"children":2130},{},[2131,2150,2168,2178],{"type":49,"tag":458,"props":2132,"children":2133},{},[2134,2136,2141,2143,2148],{"type":54,"value":2135},"Inference-only envs (null ",{"type":49,"tag":81,"props":2137,"children":2139},{"className":2138},[],[2140],{"type":54,"value":481},{"type":54,"value":2142},", e.g. ",{"type":49,"tag":81,"props":2144,"children":2146},{"className":2145},[],[2147],{"type":54,"value":500},{"type":54,"value":2149},") cannot be fine-tuned.",{"type":49,"tag":458,"props":2151,"children":2152},{},[2153,2155,2160,2162,2167],{"type":54,"value":2154},"Requires GPU(s); ",{"type":49,"tag":81,"props":2156,"children":2158},{"className":2157},[],[2159],{"type":54,"value":2119},{"type":54,"value":2161}," must not exceed the count from ",{"type":49,"tag":81,"props":2163,"children":2165},{"className":2164},[],[2166],{"type":54,"value":729},{"type":54,"value":470},{"type":49,"tag":458,"props":2169,"children":2170},{},[2171,2172,2177],{"type":54,"value":664},{"type":49,"tag":81,"props":2173,"children":2175},{"className":2174},[],[2176],{"type":54,"value":670},{"type":54,"value":470},{"type":49,"tag":458,"props":2179,"children":2180},{},[2181],{"type":54,"value":2182},"Each env maps to one stack\u002FCLI (see Stack Map); the dataset must match that env.",{"type":49,"tag":57,"props":2184,"children":2186},{"id":2185},"troubleshooting",[2187],{"type":54,"value":2188},"Troubleshooting",{"type":49,"tag":454,"props":2190,"children":2191},{},[2192,2210,2241,2257],{"type":49,"tag":458,"props":2193,"children":2194},{},[2195,2201,2203,2208],{"type":49,"tag":2196,"props":2197,"children":2198},"strong",{},[2199],{"type":54,"value":2200},"Error:",{"type":54,"value":2202}," train CLI \u002F module import fails - Cause: workflow not set up, stack ",{"type":49,"tag":81,"props":2204,"children":2206},{"className":2205},[],[2207],{"type":54,"value":2069},{"type":54,"value":2209}," missing. Fix: run [[i4h-workflow-setup]] first.",{"type":49,"tag":458,"props":2211,"children":2212},{},[2213,2217,2219,2224,2226,2232,2234,2240],{"type":49,"tag":2196,"props":2214,"children":2215},{},[2216],{"type":54,"value":2200},{"type":54,"value":2218}," dataset path rejected \u002F missing ",{"type":49,"tag":81,"props":2220,"children":2222},{"className":2221},[],[2223],{"type":54,"value":468},{"type":54,"value":2225}," - Cause: ",{"type":49,"tag":81,"props":2227,"children":2229},{"className":2228},[],[2230],{"type":54,"value":2231},"--dataset-path",{"type":54,"value":2233}," is not a valid LeRobot directory. Fix: point to a converted LeRobot dataset (see Preflight ",{"type":49,"tag":81,"props":2235,"children":2237},{"className":2236},[],[2238],{"type":54,"value":2239},"test -f",{"type":54,"value":1875},{"type":49,"tag":458,"props":2242,"children":2243},{},[2244,2248,2250,2255],{"type":49,"tag":2196,"props":2245,"children":2246},{},[2247],{"type":54,"value":2200},{"type":54,"value":2249}," env is inference-only \u002F no train support - Cause: ",{"type":49,"tag":81,"props":2251,"children":2253},{"className":2252},[],[2254],{"type":54,"value":481},{"type":54,"value":2256}," is null for that env. Fix: choose a train-capable env from the Stack Map.",{"type":49,"tag":458,"props":2258,"children":2259},{},[2260,2264,2266,2271,2273,2278],{"type":49,"tag":2196,"props":2261,"children":2262},{},[2263],{"type":54,"value":2200},{"type":54,"value":2265}," unrecognized flag like ",{"type":49,"tag":81,"props":2267,"children":2269},{"className":2268},[],[2270],{"type":54,"value":1873},{"type":54,"value":2272}," - Cause: Tyro flags use kebab case. Fix: use ",{"type":49,"tag":81,"props":2274,"children":2276},{"className":2275},[],[2277],{"type":54,"value":1865},{"type":54,"value":2279}," form.",{"type":49,"tag":57,"props":2281,"children":2283},{"id":2282},"final-response",[2284],{"type":54,"value":2285},"Final Response",{"type":49,"tag":64,"props":2287,"children":2288},{},[2289],{"type":54,"value":2290},"Report env, stack, dataset path, output checkpoint path, train_loss summary, and blockers.",{"type":49,"tag":2292,"props":2293,"children":2294},"style",{},[2295],{"type":54,"value":2296},"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":2298,"total":2400},[2299,2316,2330,2344,2356,2373,2388],{"slug":2300,"name":2300,"fn":2301,"description":2302,"org":2303,"tags":2304,"stars":23,"repoUrl":24,"updatedAt":2315},"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},[2305,2308,2311,2312],{"name":2306,"slug":2307,"type":15},"Data Analysis","data-analysis",{"name":2309,"slug":2310,"type":15},"Data Engineering","data-engineering",{"name":9,"slug":8,"type":15},{"name":2313,"slug":2314,"type":15},"Performance","performance","2026-07-14T05:28:43.176466",{"slug":2317,"name":2317,"fn":2318,"description":2319,"org":2320,"tags":2321,"stars":23,"repoUrl":24,"updatedAt":2329},"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},[2322,2325,2328],{"name":2323,"slug":2324,"type":15},"Deployment","deployment",{"name":2326,"slug":2327,"type":15},"Infrastructure","infrastructure",{"name":9,"slug":8,"type":15},"2026-07-14T05:29:06.667109",{"slug":2331,"name":2331,"fn":2332,"description":2333,"org":2334,"tags":2335,"stars":23,"repoUrl":24,"updatedAt":2343},"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},[2336,2339,2340],{"name":2337,"slug":2338,"type":15},"Agents","agents",{"name":9,"slug":8,"type":15},{"name":2341,"slug":2342,"type":15},"Research","research","2026-07-14T05:28:06.816956",{"slug":2345,"name":2345,"fn":2346,"description":2347,"org":2348,"tags":2349,"stars":23,"repoUrl":24,"updatedAt":2355},"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},[2350,2351,2352],{"name":2306,"slug":2307,"type":15},{"name":9,"slug":8,"type":15},{"name":2353,"slug":2354,"type":15},"Testing","testing","2026-07-17T05:29:03.913266",{"slug":2357,"name":2357,"fn":2358,"description":2359,"org":2360,"tags":2361,"stars":23,"repoUrl":24,"updatedAt":2372},"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},[2362,2365,2368,2369],{"name":2363,"slug":2364,"type":15},"Automation","automation",{"name":2366,"slug":2367,"type":15},"Imaging","imaging",{"name":9,"slug":8,"type":15},{"name":2370,"slug":2371,"type":15},"Video","video","2026-07-17T05:28:53.905004",{"slug":2374,"name":2374,"fn":2375,"description":2376,"org":2377,"tags":2378,"stars":23,"repoUrl":24,"updatedAt":2387},"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},[2379,2380,2383,2384],{"name":2323,"slug":2324,"type":15},{"name":2381,"slug":2382,"type":15},"Docker","docker",{"name":9,"slug":8,"type":15},{"name":2385,"slug":2386,"type":15},"Operations","operations","2026-07-17T05:28:56.913999",{"slug":2389,"name":2389,"fn":2390,"description":2391,"org":2392,"tags":2393,"stars":23,"repoUrl":24,"updatedAt":2399},"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},[2394,2395,2398],{"name":9,"slug":8,"type":15},{"name":2396,"slug":2397,"type":15},"Quantum Computing","quantum-computing",{"name":21,"slug":22,"type":15},"2026-07-14T05:26:58.898253",305,{"items":2402,"total":2551},[2403,2421,2437,2448,2460,2474,2487,2499,2510,2519,2533,2542],{"slug":2404,"name":2404,"fn":2405,"description":2406,"org":2407,"tags":2408,"stars":2418,"repoUrl":2419,"updatedAt":2420},"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},[2409,2412,2415],{"name":2410,"slug":2411,"type":15},"Documentation","documentation",{"name":2413,"slug":2414,"type":15},"MCP","mcp",{"name":2416,"slug":2417,"type":15},"Search","search",21777,"https:\u002F\u002Fgithub.com\u002FNVIDIA\u002FNemoClaw","2026-07-20T06:00:01.461044",{"slug":2422,"name":2422,"fn":2423,"description":2424,"org":2425,"tags":2426,"stars":2434,"repoUrl":2435,"updatedAt":2436},"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},[2427,2430,2431],{"name":2428,"slug":2429,"type":15},"Containers","containers",{"name":2323,"slug":2324,"type":15},{"name":2432,"slug":2433,"type":15},"Python","python",17049,"https:\u002F\u002Fgithub.com\u002FNVIDIA\u002FMegatron-LM","2026-07-27T06:06:11.249662",{"slug":2438,"name":2438,"fn":2439,"description":2440,"org":2441,"tags":2442,"stars":2434,"repoUrl":2435,"updatedAt":2447},"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},[2443,2446],{"name":2444,"slug":2445,"type":15},"CI\u002FCD","ci-cd",{"name":2323,"slug":2324,"type":15},"2026-07-14T05:25:59.97109",{"slug":2449,"name":2449,"fn":2450,"description":2451,"org":2452,"tags":2453,"stars":2434,"repoUrl":2435,"updatedAt":2459},"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},[2454,2455,2456],{"name":2444,"slug":2445,"type":15},{"name":2323,"slug":2324,"type":15},{"name":2457,"slug":2458,"type":15},"GitHub","github","2026-07-27T06:06:12.278222",{"slug":2461,"name":2461,"fn":2462,"description":2463,"org":2464,"tags":2465,"stars":2434,"repoUrl":2435,"updatedAt":2473},"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},[2466,2469,2470],{"name":2467,"slug":2468,"type":15},"Debugging","debugging",{"name":2457,"slug":2458,"type":15},{"name":2471,"slug":2472,"type":15},"Triage","triage","2026-07-14T05:25:57.442089",{"slug":2475,"name":2475,"fn":2476,"description":2477,"org":2478,"tags":2479,"stars":2434,"repoUrl":2435,"updatedAt":2486},"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},[2480,2483],{"name":2481,"slug":2482,"type":15},"Best Practices","best-practices",{"name":2484,"slug":2485,"type":15},"Code Analysis","code-analysis","2026-07-14T05:25:56.18433",{"slug":2488,"name":2488,"fn":2489,"description":2490,"org":2491,"tags":2492,"stars":2434,"repoUrl":2435,"updatedAt":2498},"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},[2493,2494,2497],{"name":17,"slug":18,"type":15},{"name":2495,"slug":2496,"type":15},"Migration","migration",{"name":9,"slug":8,"type":15},"2026-07-17T06:07:11.777011",{"slug":2500,"name":2500,"fn":2501,"description":2502,"org":2503,"tags":2504,"stars":2434,"repoUrl":2435,"updatedAt":2509},"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},[2505,2508],{"name":2506,"slug":2507,"type":15},"QA","qa",{"name":2353,"slug":2354,"type":15},"2026-07-14T05:25:53.673039",{"slug":2511,"name":2511,"fn":2512,"description":2513,"org":2514,"tags":2515,"stars":2434,"repoUrl":2435,"updatedAt":2518},"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},[2516,2517],{"name":2323,"slug":2324,"type":15},{"name":2326,"slug":2327,"type":15},"2026-07-14T05:25:49.362534",{"slug":2520,"name":2520,"fn":2521,"description":2522,"org":2523,"tags":2524,"stars":2434,"repoUrl":2435,"updatedAt":2532},"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},[2525,2528,2529],{"name":2526,"slug":2527,"type":15},"Code Review","code-review",{"name":2457,"slug":2458,"type":15},{"name":2530,"slug":2531,"type":15},"Pull Requests","pull-requests","2026-07-14T05:26:01.226578",{"slug":2534,"name":2534,"fn":2535,"description":2536,"org":2537,"tags":2538,"stars":2434,"repoUrl":2435,"updatedAt":2541},"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},[2539,2540],{"name":2506,"slug":2507,"type":15},{"name":2353,"slug":2354,"type":15},"2026-07-14T05:25:54.928983",{"slug":2543,"name":2543,"fn":2544,"description":2545,"org":2546,"tags":2547,"stars":2434,"repoUrl":2435,"updatedAt":2550},"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},[2548,2549],{"name":2363,"slug":2364,"type":15},{"name":2444,"slug":2445,"type":15},"2026-07-30T05:29:03.275638",496]