[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M78DC1SM637QFN6DHK6MFSYN":3,"skill-same:01M78DC1SM637QFN6DHK6MFSYN":159},{"skillNo":4,"name":5,"nameZh":6,"description":7,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":16,"categories":17,"keywords":18,"publishedAt":19,"linkOnly":20,"markdownContent":21,"markdownBody":22,"frontMatter":23,"refName":33,"commitSha":34,"skillDirectory":35,"forksCount":14,"repositoryHtmlUrl":36,"repositoryHomepageUrl":37,"ownerHtmlUrl":38,"ownerBlogUrl":39,"ownerDescription":40,"updatedAt":41,"install":42,"resources":55},"01M78DC1SM637QFN6DHK6MFSYN","tao-train-foundation-stereo",null,"Stereo depth estimation using FoundationStereo. Predicts disparity maps from stereo image pairs for 3D reconstruction. Use when training, evaluating, exporting, or running inference for a TAO FoundationStereo model. Trigger phrases include \"train stereo depth\", \"FoundationStereo\", \"stereo disparity estimation\", \"3D reconstruction from stereo\".","Solizardking\u002Fskillhub-main","Solizardking","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F108096810?v=4","01M786KDB10AYV042N0F9BD5B5","Clawd","01M786NYYXSZK1D05WM8JQPC48",0,"Apache-2.0","https:\u002F\u002Fgithub.com\u002FSolizardking\u002Fskillhub-main\u002Fblob\u002Fmain\u002Fskills\u002Fnvidia\u002Ftao-train-foundation-stereo\u002FSKILL.md",[],[],"2026-11-15 14:19:17",false,"---\nname: tao-train-foundation-stereo\ndescription: Stereo depth estimation using FoundationStereo. Predicts disparity maps from stereo image pairs for 3D\n  reconstruction. Use when training, evaluating, exporting, or running inference for a TAO FoundationStereo model. Trigger\n  phrases include \"train stereo depth\", \"FoundationStereo\", \"stereo disparity estimation\", \"3D reconstruction from stereo\".\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- stereo\n- depth\n- estimation\n---\n\n# Depth Net Stereo\n\nStereo depth estimation using FoundationStereo architecture. Predicts disparity maps from stereo image pairs for 3D reconstruction.\n\nUses pretrained Depth Anything v2 and EdgeNeXt encoders. Set `model.stereo_backbone.depth_anything_v2_pretrained_path` and `model.stereo_backbone.edgenext_pretrained_path`.\n\nThe mono and stereo skills both invoke the unified TAO `depth_net` CLI inside the container; the mono\u002Fstereo family is selected via `model.model_type` (e.g., `FoundationStereo`).\n\nPyT actions packaged by this model skill: `train`, `evaluate`, `inference`, `export`, and `quantize`. The PyT `depth_net` entrypoint does not accept a `gen_trt_engine` action in the current TAO image; build TensorRT engines only through the deploy workflow.\n\nFor TAO Deploy TensorRT actions (`gen_trt_engine`, TensorRT `evaluate`, and TensorRT `inference`), read `references\u002Ftao-deploy-foundation-stereo.md` first. The deploy spec template lives in this skill's `references\u002Fspec_template_deploy.yaml`.\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## Workflow\n\n### Prerequisites — data accessibility\n\nYour dataset (left + right images + GT disparity) must be reachable from inside the container:\n- **SDK runner**: place files at the S3 paths the runner resolves (the `S3_TRAIN` \u002F `S3_EVAL` placeholders shown in the spec overrides). The runner handles S3 → container-path mounting transparently.\n- **Direct `docker run`** (e.g. local testing): mount the host dataset root read-only at the same in-container path:\n\n```\ndocker run ... -v \u003Chost_data_root>:\u003Chost_data_root>:ro \u003Ccontainer> ...\n```\n\nThe same accessibility requirement applies to the `\u003Coutput_dir>` written by all actions.\n\n### Step 1 — Annotation file\n\nPer-line annotation file referenced by `data_sources[*].data_file`:\n\n| Columns | Format | Use |\n|---|---|---|\n| 2 | `\u003Cleft> \u003Cright>` | Stereo inference (no GT) |\n| 3 | `\u003Cleft> \u003Cright> \u003Cdisparity>` | Stereo with GT |\n| 4 | `\u003Cleft> \u003Cright> \u003Cdisparity> \u003Cocclusion_mask>` | Stereo with GT and occlusion mask |\n\nIf you already have one, point to it. Otherwise generate via `depth_net convert`:\n\n```\ndepth_net convert -e \u003Cconvert_spec.yaml>\n```\n\n`convert_spec.yaml` template (stereo):\n\n```yaml\nresults_dir: \u003Cdirectory where generated annotation files are written>\ndata_root: \u003Cdirectory whose immediate children are scene folders that contain your image+depth files; convert walks data_root recursively but expects per-scene subdirectories at one level below>\nimage_dir_pattern: [\u003Csubstring matching left image paths>]\nright_dir_pattern: [\u003Csubstring matching right image paths>]\ndepth_dir_pattern: [\u003Csubstring matching GT disparity paths>]\nnocc_dir_pattern: []                 # optional, occlusion mask paths\nimage_extension: '.png'  # always include the leading dot\ndepth_extension: '.png'  # form must match image_extension (the swap is a substring replace)\nnocc_extension: ''\nsplit_ratio: 0.0        # 0.0\u002F1.0 = test-only; 0.8 = 80\u002F20 train+val\n```\n\n`convert` walks `data_root` recursively, selects paths whose path-string contains *all* substrings in `image_dir_pattern` (AND-filter), then derives right \u002F depth \u002F mask paths by replacing `image_dir_pattern[0]` with the corresponding pattern's first element plus extension swap. Inspect your dataset's directory layout and identify the substrings distinguishing left, right, and GT (e.g. `im0` vs `im1` vs `disp0GT` for Middlebury).\n\n### Step 2 — Pair `model_type` and `dataset_name` based on your data\n\nPrefer the dataset-specific class when your layout matches a supported one — it applies class-specific path conventions, evaluation crops, and (where applicable) occlusion-mask handling. Fall back to `GenericDataset` only for layouts that do not match any registered class.\n\n| Data category | `model_type` | `dataset_name` |\n|---|---|---|\n| Middlebury data | `FoundationStereo` | `Middlebury` |\n| KITTI data | `FoundationStereo` | `Kitti` |\n| ETH3D data | `FoundationStereo` | `Eth3d` |\n| FSD synthetic data | `FoundationStereo` | `FSD` |\n| IsaacReal synthetic data | `FoundationStereo` | `IsaacRealDataset` |\n| Crestereo synthetic data | `FoundationStereo` | `Crestereo` |\n| Other \u002F non-canonical layout | `FoundationStereo` | `GenericDataset` |\n\nValid `dataset_name` values for stereo `data_sources` (case-insensitive): `FSD`, `IsaacRealDataset`, `Crestereo`, `Middlebury`, `Eth3d`, `Kitti`, `GenericDataset`.\n\nThe same `dataset_name` value applies across train and evaluate actions (all of which use 3-column or 4-column annotations with GT disparity). The deploy-side `evaluate` action follows the same rule — see `references\u002Ftao-deploy-foundation-stereo.md`. For inference with 2-column annotations (left + right, no GT), use `dataset_name: GenericDataset` regardless of data layout — the dataset-specific classes (`Middlebury` \u002F `Kitti` \u002F `Eth3d` \u002F `FSD` \u002F `IsaacRealDataset` \u002F `Crestereo`) require 3-column input and reject 2-column annotations at the dataloader level. For inference with 3-column annotations (left + right + GT), the dataset-specific class is fine.\n\n### Step 3 — Write spec yaml from the spec overrides\n\nCopy the action block from `references\u002Fspec-overrides-foundation-stereo.md`. Replace:\n- `model.model_type` from Step 2 (typically `FoundationStereo`)\n- `dataset.\u003C...>.data_sources[*].dataset_name` from Step 2\n- `dataset.\u003C...>.data_sources[*].data_file` with the path from Step 1\n- For deploy-side `evaluate`: enforce `dataset.test_dataset.batch_size: 1` (see `references\u002Ftao-deploy-foundation-stereo.md`).\n\nShape consistency: the `crop_size` in `dataset.test_dataset.augmentation.crop_size` should match `export.input_height` \u002F `input_width` so the trained-model evaluator and the deploy-side TensorRT evaluator operate at the same shape. Note that `crop_size` is decorative on the pyt `evaluate` path but authoritative on the deploy `evaluate` side — see `references\u002Ftroubleshooting-foundation-stereo.md` and `references\u002Ftao-deploy-foundation-stereo.md`.\n\nFresh-install smoke runs are validated at `crop_size: [128, 128]` with `dataset.max_disparity: 128` and `model.max_disparity: 128`. Avoid 112×112 crops and avoid setting `max_disparity` smaller than the square crop side for smoke tests: those combinations can fail inside FoundationStereo with feature-map or loss-mask shape mismatches before a checkpoint is produced.\n\nData source overrides are **mandatory for every action**. Each `data_sources` entry is a dict with two mandatory fields: `data_file` and `dataset_name`. See `references\u002Fspec-overrides-foundation-stereo.md` for the per-action dataset-requirements table, every action's override block, and the `quantize` known-issue note.\n\n### Step 4 — Run\n\nCreate writable home\u002Fcache directories inside the mounted output path before using\n`--user`. Some TAO containers do not have an `\u002Fetc\u002Fpasswd` entry for the host UID,\nand PyTorch \u002F matplotlib need writable cache paths when running as that UID.\n\n```bash\nmkdir -p \u003Coutput_dir>\u002Fhome \\\n         \u003Coutput_dir>\u002F.cache\u002Fmatplotlib \\\n         \u003Coutput_dir>\u002F.cache\u002Ftorchinductor \\\n         \u003Coutput_dir>\u002F.cache\u002Fxdg\n```\n\n```\ndocker run --gpus 'device=0' --shm-size 16G --ipc=host \\\n  --user \"$(id -u):$(id -g)\" \\\n  -e USER=\"$(id -un)\" \\\n  -e LOGNAME=\"$(id -un)\" \\\n  -e HOME=\u003Coutput_dir>\u002Fhome \\\n  -e MPLCONFIGDIR=\u003Coutput_dir>\u002F.cache\u002Fmatplotlib \\\n  -e TORCHINDUCTOR_CACHE_DIR=\u003Coutput_dir>\u002F.cache\u002Ftorchinductor \\\n  -e XDG_CACHE_HOME=\u003Coutput_dir>\u002F.cache\u002Fxdg \\\n  -v \u003Cdata_root>:\u003Cdata_root>:ro \\\n  -v \u003Coutput_dir>:\u003Coutput_dir> \\\n  \u003Ccontainer> \\\n  depth_net \u003Caction> -e \u003Cspec.yaml>\n```\n\nWithout `--user \"$(id -u):$(id -g)\"` the container writes outputs as `nobody:nogroup`, blocking host-side cleanup \u002F retry.\n\n### Step 5 — Verify\n\n- Container exit code 0\n- `status.json` `kpi` block populated\n- For `train`: inspect per-step `train_loss` directly (the entrypoint reports `Execution status: PASS` even when loss is NaN)\n- For `evaluate`: rely on `epe` \u002F `bp1` \u002F `bp2` \u002F `bp3` \u002F `d1` \u002F `rmse` (the evaluator also emits `abs_rel` \u002F `sq_rel` \u002F `rmse_log` which are non-meaningful for stereo — see `references\u002Fparameters-foundation-stereo.md`)\n- For `inference`: artifacts under `results_dir`\n\nFor TAO Deploy TensorRT actions (`gen_trt_engine`, TensorRT `evaluate`, and TensorRT `inference`), read `references\u002Ftao-deploy-foundation-stereo.md` first. Deploy spec templates live in this skill's `references\u002F` folder with the `spec_template_deploy_*.yaml` prefix.\n\n## Training Requirements\n\n- **Monitoring metric:** val\u002Floss\n- **Eval dataset:** optional. Val dataset configured via `dataset.val_dataset.data_sources` (each entry needs `data_file` and `dataset_name`).\n\nSee `references\u002Fspec-overrides-foundation-stereo.md` for the per-action dataset-requirements table and every action's mandatory data-source override block.\n\n## Parameters, Metrics, Multi-GPU, Export\u002FTRT, Hardware\n\nSee `references\u002Fparameters-foundation-stereo.md` for the full Important Parameters list (incl. `model.encoder` `vits` override, `model.max_disparity` default 416, `model.volume_dim` no-op note, `dataset.baseline`, `dataset.focal_x`, `train.precision`, `export.batch_size`), the Evaluation Metrics table, Multi-GPU \u002F Multi-Node launch keys, Export \u002F TRT Defaults (`opset_version`\u002F`on_cpu` pairing, NGC 576×960 settings), and Hardware requirements.\n\n## Error Patterns and Troubleshooting\n\nSee `references\u002Ftroubleshooting-foundation-stereo.md` for disparity overflow, smoke-test shape mismatch, missing pretrained paths, the `encoder` \u002F `dataset_name` struct errors, the `depth_net_stereo: not found` entrypoint note, the pyt-vs-deploy `crop_size` discussion, and the deploy `evaluate` scalar-conversion failure.\n\n## Spec Param \u002F Parent Model Inference\n\nSee `references\u002Fcheckpoint-inference-mappings-foundation-stereo.md` for the checkpoint-resolution rules (`model_epoch_\u003Cepoch>_step_\u003Cstep>.pth`, `dn_model_latest.pth` policy), the absence of parent PyT `gen_trt_engine`, and the full per-action inference-mapping table from `depth_net_stereo.config.json` (including `parent_model` \u002F `parent_job_id` resolution).\n\n## Deployment\n\n- [tao-deploy-foundation-stereo](references\u002Ftao-deploy-foundation-stereo.md)\n","# Depth Net Stereo\n\nStereo depth estimation using FoundationStereo architecture. Predicts disparity maps from stereo image pairs for 3D reconstruction.\n\nUses pretrained Depth Anything v2 and EdgeNeXt encoders. Set `model.stereo_backbone.depth_anything_v2_pretrained_path` and `model.stereo_backbone.edgenext_pretrained_path`.\n\nThe mono and stereo skills both invoke the unified TAO `depth_net` CLI inside the container; the mono\u002Fstereo family is selected via `model.model_type` (e.g., `FoundationStereo`).\n\nPyT actions packaged by this model skill: `train`, `evaluate`, `inference`, `export`, and `quantize`. The PyT `depth_net` entrypoint does not accept a `gen_trt_engine` action in the current TAO image; build TensorRT engines only through the deploy workflow.\n\nFor TAO Deploy TensorRT actions (`gen_trt_engine`, TensorRT `evaluate`, and TensorRT `inference`), read `references\u002Ftao-deploy-foundation-stereo.md` first. The deploy spec template lives in this skill's `references\u002Fspec_template_deploy.yaml`.\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## Workflow\n\n### Prerequisites — data accessibility\n\nYour dataset (left + right images + GT disparity) must be reachable from inside the container:\n- **SDK runner**: place files at the S3 paths the runner resolves (the `S3_TRAIN` \u002F `S3_EVAL` placeholders shown in the spec overrides). The runner handles S3 → container-path mounting transparently.\n- **Direct `docker run`** (e.g. local testing): mount the host dataset root read-only at the same in-container path:\n\n```\ndocker run ... -v \u003Chost_data_root>:\u003Chost_data_root>:ro \u003Ccontainer> ...\n```\n\nThe same accessibility requirement applies to the `\u003Coutput_dir>` written by all actions.\n\n### Step 1 — Annotation file\n\nPer-line annotation file referenced by `data_sources[*].data_file`:\n\n| Columns | Format | Use |\n|---|---|---|\n| 2 | `\u003Cleft> \u003Cright>` | Stereo inference (no GT) |\n| 3 | `\u003Cleft> \u003Cright> \u003Cdisparity>` | Stereo with GT |\n| 4 | `\u003Cleft> \u003Cright> \u003Cdisparity> \u003Cocclusion_mask>` | Stereo with GT and occlusion mask |\n\nIf you already have one, point to it. Otherwise generate via `depth_net convert`:\n\n```\ndepth_net convert -e \u003Cconvert_spec.yaml>\n```\n\n`convert_spec.yaml` template (stereo):\n\n```yaml\nresults_dir: \u003Cdirectory where generated annotation files are written>\ndata_root: \u003Cdirectory whose immediate children are scene folders that contain your image+depth files; convert walks data_root recursively but expects per-scene subdirectories at one level below>\nimage_dir_pattern: [\u003Csubstring matching left image paths>]\nright_dir_pattern: [\u003Csubstring matching right image paths>]\ndepth_dir_pattern: [\u003Csubstring matching GT disparity paths>]\nnocc_dir_pattern: []                 # optional, occlusion mask paths\nimage_extension: '.png'  # always include the leading dot\ndepth_extension: '.png'  # form must match image_extension (the swap is a substring replace)\nnocc_extension: ''\nsplit_ratio: 0.0        # 0.0\u002F1.0 = test-only; 0.8 = 80\u002F20 train+val\n```\n\n`convert` walks `data_root` recursively, selects paths whose path-string contains *all* substrings in `image_dir_pattern` (AND-filter), then derives right \u002F depth \u002F mask paths by replacing `image_dir_pattern[0]` with the corresponding pattern's first element plus extension swap. Inspect your dataset's directory layout and identify the substrings distinguishing left, right, and GT (e.g. `im0` vs `im1` vs `disp0GT` for Middlebury).\n\n### Step 2 — Pair `model_type` and `dataset_name` based on your data\n\nPrefer the dataset-specific class when your layout matches a supported one — it applies class-specific path conventions, evaluation crops, and (where applicable) occlusion-mask handling. Fall back to `GenericDataset` only for layouts that do not match any registered class.\n\n| Data category | `model_type` | `dataset_name` |\n|---|---|---|\n| Middlebury data | `FoundationStereo` | `Middlebury` |\n| KITTI data | `FoundationStereo` | `Kitti` |\n| ETH3D data | `FoundationStereo` | `Eth3d` |\n| FSD synthetic data | `FoundationStereo` | `FSD` |\n| IsaacReal synthetic data | `FoundationStereo` | `IsaacRealDataset` |\n| Crestereo synthetic data | `FoundationStereo` | `Crestereo` |\n| Other \u002F non-canonical layout | `FoundationStereo` | `GenericDataset` |\n\nValid `dataset_name` values for stereo `data_sources` (case-insensitive): `FSD`, `IsaacRealDataset`, `Crestereo`, `Middlebury`, `Eth3d`, `Kitti`, `GenericDataset`.\n\nThe same `dataset_name` value applies across train and evaluate actions (all of which use 3-column or 4-column annotations with GT disparity). The deploy-side `evaluate` action follows the same rule — see `references\u002Ftao-deploy-foundation-stereo.md`. For inference with 2-column annotations (left + right, no GT), use `dataset_name: GenericDataset` regardless of data layout — the dataset-specific classes (`Middlebury` \u002F `Kitti` \u002F `Eth3d` \u002F `FSD` \u002F `IsaacRealDataset` \u002F `Crestereo`) require 3-column input and reject 2-column annotations at the dataloader level. For inference with 3-column annotations (left + right + GT), the dataset-specific class is fine.\n\n### Step 3 — Write spec yaml from the spec overrides\n\nCopy the action block from `references\u002Fspec-overrides-foundation-stereo.md`. Replace:\n- `model.model_type` from Step 2 (typically `FoundationStereo`)\n- `dataset.\u003C...>.data_sources[*].dataset_name` from Step 2\n- `dataset.\u003C...>.data_sources[*].data_file` with the path from Step 1\n- For deploy-side `evaluate`: enforce `dataset.test_dataset.batch_size: 1` (see `references\u002Ftao-deploy-foundation-stereo.md`).\n\nShape consistency: the `crop_size` in `dataset.test_dataset.augmentation.crop_size` should match `export.input_height` \u002F `input_width` so the trained-model evaluator and the deploy-side TensorRT evaluator operate at the same shape. Note that `crop_size` is decorative on the pyt `evaluate` path but authoritative on the deploy `evaluate` side — see `references\u002Ftroubleshooting-foundation-stereo.md` and `references\u002Ftao-deploy-foundation-stereo.md`.\n\nFresh-install smoke runs are validated at `crop_size: [128, 128]` with `dataset.max_disparity: 128` and `model.max_disparity: 128`. Avoid 112×112 crops and avoid setting `max_disparity` smaller than the square crop side for smoke tests: those combinations can fail inside FoundationStereo with feature-map or loss-mask shape mismatches before a checkpoint is produced.\n\nData source overrides are **mandatory for every action**. Each `data_sources` entry is a dict with two mandatory fields: `data_file` and `dataset_name`. See `references\u002Fspec-overrides-foundation-stereo.md` for the per-action dataset-requirements table, every action's override block, and the `quantize` known-issue note.\n\n### Step 4 — Run\n\nCreate writable home\u002Fcache directories inside the mounted output path before using\n`--user`. Some TAO containers do not have an `\u002Fetc\u002Fpasswd` entry for the host UID,\nand PyTorch \u002F matplotlib need writable cache paths when running as that UID.\n\n```bash\nmkdir -p \u003Coutput_dir>\u002Fhome \\\n         \u003Coutput_dir>\u002F.cache\u002Fmatplotlib \\\n         \u003Coutput_dir>\u002F.cache\u002Ftorchinductor \\\n         \u003Coutput_dir>\u002F.cache\u002Fxdg\n```\n\n```\ndocker run --gpus 'device=0' --shm-size 16G --ipc=host \\\n  --user \"$(id -u):$(id -g)\" \\\n  -e USER=\"$(id -un)\" \\\n  -e LOGNAME=\"$(id -un)\" \\\n  -e HOME=\u003Coutput_dir>\u002Fhome \\\n  -e MPLCONFIGDIR=\u003Coutput_dir>\u002F.cache\u002Fmatplotlib \\\n  -e TORCHINDUCTOR_CACHE_DIR=\u003Coutput_dir>\u002F.cache\u002Ftorchinductor \\\n  -e XDG_CACHE_HOME=\u003Coutput_dir>\u002F.cache\u002Fxdg \\\n  -v \u003Cdata_root>:\u003Cdata_root>:ro \\\n  -v \u003Coutput_dir>:\u003Coutput_dir> \\\n  \u003Ccontainer> \\\n  depth_net \u003Caction> -e \u003Cspec.yaml>\n```\n\nWithout `--user \"$(id -u):$(id -g)\"` the container writes outputs as `nobody:nogroup`, blocking host-side cleanup \u002F retry.\n\n### Step 5 — Verify\n\n- Container exit code 0\n- `status.json` `kpi` block populated\n- For `train`: inspect per-step `train_loss` directly (the entrypoint reports `Execution status: PASS` even when loss is NaN)\n- For `evaluate`: rely on `epe` \u002F `bp1` \u002F `bp2` \u002F `bp3` \u002F `d1` \u002F `rmse` (the evaluator also emits `abs_rel` \u002F `sq_rel` \u002F `rmse_log` which are non-meaningful for stereo — see `references\u002Fparameters-foundation-stereo.md`)\n- For `inference`: artifacts under `results_dir`\n\nFor TAO Deploy TensorRT actions (`gen_trt_engine`, TensorRT `evaluate`, and TensorRT `inference`), read `references\u002Ftao-deploy-foundation-stereo.md` first. Deploy spec templates live in this skill's `references\u002F` folder with the `spec_template_deploy_*.yaml` prefix.\n\n## Training Requirements\n\n- **Monitoring metric:** val\u002Floss\n- **Eval dataset:** optional. Val dataset configured via `dataset.val_dataset.data_sources` (each entry needs `data_file` and `dataset_name`).\n\nSee `references\u002Fspec-overrides-foundation-stereo.md` for the per-action dataset-requirements table and every action's mandatory data-source override block.\n\n## Parameters, Metrics, Multi-GPU, Export\u002FTRT, Hardware\n\nSee `references\u002Fparameters-foundation-stereo.md` for the full Important Parameters list (incl. `model.encoder` `vits` override, `model.max_disparity` default 416, `model.volume_dim` no-op note, `dataset.baseline`, `dataset.focal_x`, `train.precision`, `export.batch_size`), the Evaluation Metrics table, Multi-GPU \u002F Multi-Node launch keys, Export \u002F TRT Defaults (`opset_version`\u002F`on_cpu` pairing, NGC 576×960 settings), and Hardware requirements.\n\n## Error Patterns and Troubleshooting\n\nSee `references\u002Ftroubleshooting-foundation-stereo.md` for disparity overflow, smoke-test shape mismatch, missing pretrained paths, the `encoder` \u002F `dataset_name` struct errors, the `depth_net_stereo: not found` entrypoint note, the pyt-vs-deploy `crop_size` discussion, and the deploy `evaluate` scalar-conversion failure.\n\n## Spec Param \u002F Parent Model Inference\n\nSee `references\u002Fcheckpoint-inference-mappings-foundation-stereo.md` for the checkpoint-resolution rules (`model_epoch_\u003Cepoch>_step_\u003Cstep>.pth`, `dn_model_latest.pth` policy), the absence of parent PyT `gen_trt_engine`, and the full per-action inference-mapping table from `depth_net_stereo.config.json` (including `parent_model` \u002F `parent_job_id` resolution).\n\n## Deployment\n\n- [tao-deploy-foundation-stereo](references\u002Ftao-deploy-foundation-stereo.md)\n",{"name":5,"tags":24,"license":15,"metadata":28,"description":7,"allowed-tools":31,"compatibility":32},[25,26,27],"stereo","depth","estimation",{"author":29,"version":30},"NVIDIA Corporation","0.1.0","Read Bash","Requires docker + nvidia-container-toolkit.","main","be3de0244dc1a57e323d8160787ebb40fb64930e","skills\u002Fnvidia\u002Ftao-train-foundation-stereo","https:\u002F\u002Fgithub.com\u002FSolizardking\u002Fskillhub-main","https:\u002F\u002Fskills.x402.wtf","https:\u002F\u002Fgithub.com\u002FSolizardking","https:\u002F\u002Fmusebook.trade","Got a mile high and never came down","2026-10-08 23:20:13",{"source":8,"skillName":5,"commands":43,"prompt":53,"zipAvailable":54},[44,47,50],{"tool":45,"command":46},"npx","npx skills add Solizardking\u002Fskillhub-main --skill tao-train-foundation-stereo",{"tool":48,"command":49},"bunx","bunx skills add Solizardking\u002Fskillhub-main --skill tao-train-foundation-stereo",{"tool":51,"command":52},"pnpm","pnpm dlx skills add Solizardking\u002Fskillhub-main --skill tao-train-foundation-stereo","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002FSolizardking\u002Fskillhub-main\u002Fblob\u002Fmain\u002Fskills\u002Fnvidia\u002Ftao-train-foundation-stereo\u002FSKILL.md",true,[56,62,66,70,74,78,82,86,90,94,98,102,105,109,113,117,121,125,129,133,137,141,145,149,153],{"path":57,"resourceType":58,"resourceTypeDesc":59,"sizeBytes":60,"sourceHtmlUrl":61},"BENCHMARK.md","REFERENCE","参考文档",4129,"https:\u002F\u002Fgithub.com\u002FSolizardking\u002Fskillhub-main\u002Fblob\u002Fmain\u002Fskills\u002Fnvidia\u002Ftao-train-foundation-stereo\u002FBENCHMARK.md",{"path":63,"resourceType":58,"resourceTypeDesc":59,"sizeBytes":64,"sourceHtmlUrl":65},"evals\u002Fevals.json",819,"https:\u002F\u002Fgithub.com\u002FSolizardking\u002Fskillhub-main\u002Fblob\u002Fmain\u002Fskills\u002Fnvidia\u002Ftao-train-foundation-stereo\u002Fevals\u002Fevals.json",{"path":67,"resourceType":58,"resourceTypeDesc":59,"sizeBytes":68,"sourceHtmlUrl":69},"references\u002Fcheckpoint-inference-mappings-foundation-stereo.md",3974,"https:\u002F\u002Fgithub.com\u002FSolizardking\u002Fskillhub-main\u002Fblob\u002Fmain\u002Fskills\u002Fnvidia\u002Ftao-train-foundation-stereo\u002Freferences\u002Fcheckpoint-inference-mappings-foundation-stereo.md",{"path":71,"resourceType":58,"resourceTypeDesc":59,"sizeBytes":72,"sourceHtmlUrl":73},"references\u002Fparameters-foundation-stereo.md",5755,"https:\u002F\u002Fgithub.com\u002FSolizardking\u002Fskillhub-main\u002Fblob\u002Fmain\u002Fskills\u002Fnvidia\u002Ftao-train-foundation-stereo\u002Freferences\u002Fparameters-foundation-stereo.md",{"path":75,"resourceType":58,"resourceTypeDesc":59,"sizeBytes":76,"sourceHtmlUrl":77},"references\u002Fskill_info.yaml",5594,"https:\u002F\u002Fgithub.com\u002FSolizardking\u002Fskillhub-main\u002Fblob\u002Fmain\u002Fskills\u002Fnvidia\u002Ftao-train-foundation-stereo\u002Freferences\u002Fskill_info.yaml",{"path":79,"resourceType":58,"resourceTypeDesc":59,"sizeBytes":80,"sourceHtmlUrl":81},"references\u002Fspec_template_deploy.yaml",2029,"https:\u002F\u002Fgithub.com\u002FSolizardking\u002Fskillhub-main\u002Fblob\u002Fmain\u002Fskills\u002Fnvidia\u002Ftao-train-foundation-stereo\u002Freferences\u002Fspec_template_deploy.yaml",{"path":83,"resourceType":58,"resourceTypeDesc":59,"sizeBytes":84,"sourceHtmlUrl":85},"references\u002Fspec_template_evaluate.yaml",5676,"https:\u002F\u002Fgithub.com\u002FSolizardking\u002Fskillhub-main\u002Fblob\u002Fmain\u002Fskills\u002Fnvidia\u002Ftao-train-foundation-stereo\u002Freferences\u002Fspec_template_evaluate.yaml",{"path":87,"resourceType":58,"resourceTypeDesc":59,"sizeBytes":88,"sourceHtmlUrl":89},"references\u002Fspec_template_export.yaml",5745,"https:\u002F\u002Fgithub.com\u002FSolizardking\u002Fskillhub-main\u002Fblob\u002Fmain\u002Fskills\u002Fnvidia\u002Ftao-train-foundation-stereo\u002Freferences\u002Fspec_template_export.yaml",{"path":91,"resourceType":58,"resourceTypeDesc":59,"sizeBytes":92,"sourceHtmlUrl":93},"references\u002Fspec_template_gen_trt_engine.yaml",5764,"https:\u002F\u002Fgithub.com\u002FSolizardking\u002Fskillhub-main\u002Fblob\u002Fmain\u002Fskills\u002Fnvidia\u002Ftao-train-foundation-stereo\u002Freferences\u002Fspec_template_gen_trt_engine.yaml",{"path":95,"resourceType":58,"resourceTypeDesc":59,"sizeBytes":96,"sourceHtmlUrl":97},"references\u002Fspec_template_inference.yaml",5698,"https:\u002F\u002Fgithub.com\u002FSolizardking\u002Fskillhub-main\u002Fblob\u002Fmain\u002Fskills\u002Fnvidia\u002Ftao-train-foundation-stereo\u002Freferences\u002Fspec_template_inference.yaml",{"path":99,"resourceType":58,"resourceTypeDesc":59,"sizeBytes":100,"sourceHtmlUrl":101},"references\u002Fspec_template_quantize.yaml",5511,"https:\u002F\u002Fgithub.com\u002FSolizardking\u002Fskillhub-main\u002Fblob\u002Fmain\u002Fskills\u002Fnvidia\u002Ftao-train-foundation-stereo\u002Freferences\u002Fspec_template_quantize.yaml",{"path":103,"resourceType":58,"resourceTypeDesc":59,"sizeBytes":100,"sourceHtmlUrl":104},"references\u002Fspec_template_train.yaml","https:\u002F\u002Fgithub.com\u002FSolizardking\u002Fskillhub-main\u002Fblob\u002Fmain\u002Fskills\u002Fnvidia\u002Ftao-train-foundation-stereo\u002Freferences\u002Fspec_template_train.yaml",{"path":106,"resourceType":58,"resourceTypeDesc":59,"sizeBytes":107,"sourceHtmlUrl":108},"references\u002Fspec-overrides-foundation-stereo.md",4173,"https:\u002F\u002Fgithub.com\u002FSolizardking\u002Fskillhub-main\u002Fblob\u002Fmain\u002Fskills\u002Fnvidia\u002Ftao-train-foundation-stereo\u002Freferences\u002Fspec-overrides-foundation-stereo.md",{"path":110,"resourceType":58,"resourceTypeDesc":59,"sizeBytes":111,"sourceHtmlUrl":112},"references\u002Ftao-deploy-foundation-stereo.md",11857,"https:\u002F\u002Fgithub.com\u002FSolizardking\u002Fskillhub-main\u002Fblob\u002Fmain\u002Fskills\u002Fnvidia\u002Ftao-train-foundation-stereo\u002Freferences\u002Ftao-deploy-foundation-stereo.md",{"path":114,"resourceType":58,"resourceTypeDesc":59,"sizeBytes":115,"sourceHtmlUrl":116},"references\u002Ftao-deploy-foundation-stereo.skill_info.yaml",3265,"https:\u002F\u002Fgithub.com\u002FSolizardking\u002Fskillhub-main\u002Fblob\u002Fmain\u002Fskills\u002Fnvidia\u002Ftao-train-foundation-stereo\u002Freferences\u002Ftao-deploy-foundation-stereo.skill_info.yaml",{"path":118,"resourceType":58,"resourceTypeDesc":59,"sizeBytes":119,"sourceHtmlUrl":120},"references\u002Ftroubleshooting-foundation-stereo.md",2463,"https:\u002F\u002Fgithub.com\u002FSolizardking\u002Fskillhub-main\u002Fblob\u002Fmain\u002Fskills\u002Fnvidia\u002Ftao-train-foundation-stereo\u002Freferences\u002Ftroubleshooting-foundation-stereo.md",{"path":122,"resourceType":58,"resourceTypeDesc":59,"sizeBytes":123,"sourceHtmlUrl":124},"schemas\u002Fevaluate.schema.json",106996,"https:\u002F\u002Fgithub.com\u002FSolizardking\u002Fskillhub-main\u002Fblob\u002Fmain\u002Fskills\u002Fnvidia\u002Ftao-train-foundation-stereo\u002Fschemas\u002Fevaluate.schema.json",{"path":126,"resourceType":58,"resourceTypeDesc":59,"sizeBytes":127,"sourceHtmlUrl":128},"schemas\u002Fexport.schema.json",107660,"https:\u002F\u002Fgithub.com\u002FSolizardking\u002Fskillhub-main\u002Fblob\u002Fmain\u002Fskills\u002Fnvidia\u002Ftao-train-foundation-stereo\u002Fschemas\u002Fexport.schema.json",{"path":130,"resourceType":58,"resourceTypeDesc":59,"sizeBytes":131,"sourceHtmlUrl":132},"schemas\u002Fgen_trt_engine.schema.json",109561,"https:\u002F\u002Fgithub.com\u002FSolizardking\u002Fskillhub-main\u002Fblob\u002Fmain\u002Fskills\u002Fnvidia\u002Ftao-train-foundation-stereo\u002Fschemas\u002Fgen_trt_engine.schema.json",{"path":134,"resourceType":58,"resourceTypeDesc":59,"sizeBytes":135,"sourceHtmlUrl":136},"schemas\u002Finference.schema.json",107441,"https:\u002F\u002Fgithub.com\u002FSolizardking\u002Fskillhub-main\u002Fblob\u002Fmain\u002Fskills\u002Fnvidia\u002Ftao-train-foundation-stereo\u002Fschemas\u002Finference.schema.json",{"path":138,"resourceType":58,"resourceTypeDesc":59,"sizeBytes":139,"sourceHtmlUrl":140},"schemas\u002Fmanifest.json",30587,"https:\u002F\u002Fgithub.com\u002FSolizardking\u002Fskillhub-main\u002Fblob\u002Fmain\u002Fskills\u002Fnvidia\u002Ftao-train-foundation-stereo\u002Fschemas\u002Fmanifest.json",{"path":142,"resourceType":58,"resourceTypeDesc":59,"sizeBytes":143,"sourceHtmlUrl":144},"schemas\u002Fquantize.schema.json",103806,"https:\u002F\u002Fgithub.com\u002FSolizardking\u002Fskillhub-main\u002Fblob\u002Fmain\u002Fskills\u002Fnvidia\u002Ftao-train-foundation-stereo\u002Fschemas\u002Fquantize.schema.json",{"path":146,"resourceType":58,"resourceTypeDesc":59,"sizeBytes":147,"sourceHtmlUrl":148},"schemas\u002Ftrain.schema.json",103800,"https:\u002F\u002Fgithub.com\u002FSolizardking\u002Fskillhub-main\u002Fblob\u002Fmain\u002Fskills\u002Fnvidia\u002Ftao-train-foundation-stereo\u002Fschemas\u002Ftrain.schema.json",{"path":150,"resourceType":58,"resourceTypeDesc":59,"sizeBytes":151,"sourceHtmlUrl":152},"skill-card.md",4055,"https:\u002F\u002Fgithub.com\u002FSolizardking\u002Fskillhub-main\u002Fblob\u002Fmain\u002Fskills\u002Fnvidia\u002Ftao-train-foundation-stereo\u002Fskill-card.md",{"path":154,"resourceType":155,"resourceTypeDesc":156,"sizeBytes":157,"sourceHtmlUrl":158},"skill.oms.sig","OTHER","其他",9921,"https:\u002F\u002Fgithub.com\u002FSolizardking\u002Fskillhub-main\u002Fblob\u002Fmain\u002Fskills\u002Fnvidia\u002Ftao-train-foundation-stereo\u002Fskill.oms.sig",[160,168,175,182,189,197],{"skillNo":161,"name":162,"nameZh":6,"description":163,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":164,"licenseFileUrl":6,"sourceHtmlUrl":165,"categories":166,"keywords":167,"publishedAt":19},"01M78E0TGCAMNPS207BA3G07S7","zkrouter","Self-hosted OpenAI-compatible LLM router — smart tier-based routing, ZK-stamped routing decisions, mode overrides, free OpenRouter at install via the Birth bot. Save 60-80% on inference costs by routing to the cheapest capable model across 12+ OpenRouter models.","MIT","https:\u002F\u002Fgithub.com\u002FSolizardking\u002Fskillhub-main\u002Fblob\u002Fmain\u002Fskills\u002Fzkrouter\u002FSKILL.md",[],[],{"skillNo":169,"name":170,"nameZh":6,"description":171,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":164,"licenseFileUrl":6,"sourceHtmlUrl":172,"categories":173,"keywords":174,"publishedAt":19},"01M78E0R1Z8BDT6EDTHY17MTNB","zk","For custom ZK Solana programs and privacy-preserving applications to prevent double spending. Guide to integrate rent-free nullifier PDAs for double-spend prevention.","https:\u002F\u002Fgithub.com\u002FSolizardking\u002Fskillhub-main\u002Fblob\u002Fmain\u002Fskills\u002Fzk\u002FSKILL.md",[],[],{"skillNo":176,"name":177,"nameZh":6,"description":178,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":164,"licenseFileUrl":6,"sourceHtmlUrl":179,"categories":180,"keywords":181,"publishedAt":19},"01M78E0P3ZTZ1YWF8MD5MM23QE","zk-omni-messaging","Zero-knowledge omnichain messaging between Robinhood Chain and Solana using CheshireZkOmniMessenger (msgType 4) with nullifier anti-replay and the zk-omni-relayer service.","https:\u002F\u002Fgithub.com\u002FSolizardking\u002Fskillhub-main\u002Fblob\u002Fmain\u002Fskills\u002Fzk-omni-messaging\u002FSKILL.md",[],[],{"skillNo":183,"name":184,"nameZh":6,"description":185,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":164,"licenseFileUrl":6,"sourceHtmlUrl":186,"categories":187,"keywords":188,"publishedAt":19},"01M78E0JXZ2PNB1896VXAGDBWD","zerion-wallet-data","Use when the user asks to check a wallet's portfolio, token balances, DeFi positions, PnL, transaction history, or NFTs. Provides interpreted on-chain data across 40+ EVM chains and Solana via x402 pay-per-request or API key — no raw RPC decoding needed.","https:\u002F\u002Fgithub.com\u002FSolizardking\u002Fskillhub-main\u002Fblob\u002Fmain\u002Fskills\u002Fzerion-wallet-data\u002FSKILL.md",[],[],{"skillNo":190,"name":191,"nameZh":6,"description":192,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":164,"licenseFileUrl":193,"sourceHtmlUrl":194,"categories":195,"keywords":196,"publishedAt":19},"01M78E0G14FH9GRRTQYMKD5E59","youtube-clipper","YouTube 视频智能剪辑工具。下载视频和字幕，AI 分析生成精细章节（几分钟级别）， 用户选择片段后自动剪辑、翻译字幕为中英双语、烧录字幕到视频，并生成总结文案。 使用场景：当用户需要剪辑 YouTube 视频、生成短视频片段、制作双语字幕版本时。 关键词：视频剪辑、YouTube、字幕翻译、双语字幕、视频下载、clip video","https:\u002F\u002Fgithub.com\u002FSolizardking\u002Fskillhub-main\u002Fblob\u002Fmain\u002Fskills\u002Fyoutube-clipper\u002FLICENSE","https:\u002F\u002Fgithub.com\u002FSolizardking\u002Fskillhub-main\u002Fblob\u002Fmain\u002Fskills\u002Fyoutube-clipper\u002FSKILL.md",[],[],{"skillNo":198,"name":199,"nameZh":6,"description":200,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":164,"licenseFileUrl":6,"sourceHtmlUrl":201,"categories":202,"keywords":203,"publishedAt":19},"01M78E0C2ZYA99CFFDNB7ZMK5S","yield-optimization","Multi-chain yield optimization via yield.xyz (StakeKit). Discover 2,988+ yield opportunities across 75+ blockchains, build deposit\u002Fwithdrawal transactions with shell scripts, and sign them with a MoonPay wallet. Use when the user wants to earn yield, find the best lending or staking rate, enter\u002Fexit a position, or check portfolio balances.","https:\u002F\u002Fgithub.com\u002FSolizardking\u002Fskillhub-main\u002Fblob\u002Fmain\u002Fskills\u002Fyield-optimization\u002FSKILL.md",[],[]]