[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M7CQPBKDP7QYD2ABEFSPWVZB":3,"skill-same:01M7CQPBKDP7QYD2ABEFSPWVZB":60},{"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":32,"commitSha":33,"skillDirectory":34,"forksCount":35,"repositoryHtmlUrl":36,"repositoryHomepageUrl":6,"ownerHtmlUrl":37,"ownerBlogUrl":6,"ownerDescription":38,"updatedAt":39,"install":40,"resources":53},"01M7CQPBKDP7QYD2ABEFSPWVZB","tao-train-dino",null,"DINO (DETR with Improved DeNoising Anchor Boxes) for 2D object detection. Transformer-based detector with denoising training, multi-scale features, and optional distillation support. Use when training, evaluating, exporting, distilling, quantizing, or running inference for a TAO DINO detector. Trigger phrases include \"train DINO\", \"DETR object detection\", \"TAO 2D detection\", \"DINO with distillation\".","om-scogo\u002Fskillsh-scraper","om-scogo","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F160466840?v=4","01M79AWB6DFVCQ7RK70CS34B96","Om Achrekar","01M79AWB6ZM9SKR8KDH1WSSM22",0,"Apache-2.0","https:\u002F\u002Fgithub.com\u002Fom-scogo\u002Fskillsh-scraper\u002Fblob\u002Fmain\u002Fdata\u002Fskills\u002Fnvidia-skills\u002Ftao-train-dino\u002FSkill.md",[],[],"2026-11-15 14:19:17",false,"---\nname: tao-train-dino\ndescription: DINO (DETR with Improved DeNoising Anchor Boxes) for 2D object detection. Transformer-based detector with\n  denoising training, multi-scale features, and optional distillation support. Use when training, evaluating, exporting,\n  distilling, quantizing, or running inference for a TAO DINO detector. Trigger phrases include \"train DINO\", \"DETR object\n  detection\", \"TAO 2D detection\", \"DINO with distillation\".\nlicense: Apache-2.0\ncompatibility: Requires docker + nvidia-container-toolkit.\nmetadata:\n  version: \"0.1.0\"\n  author: NVIDIA Corporation\nallowed-tools: Read Bash\ntags:\n- object\n- detection\n---\n\n# DINO\n\nDINO (DETR with Improved DeNoising Anchor Boxes) for 2D object detection. Transformer-based detector with denoising training, multi-scale features, and optional distillation support.\n\nUses pretrained backbone weights (e.g. ResNet-50 ImageNet). Set `model.pretrained_backbone_path` for backbone-only or `train.pretrained_model_path` for full model.\n\n## When To Use\n\nTrain, evaluate, export, distill, quantize, or run inference for a TAO DINO 2D object detector.\n\nFor TAO Deploy TensorRT actions (`gen_trt_engine`, TensorRT `evaluate`, and\nTensorRT `inference`), read `references\u002Ftao-deploy-dino.md` first. Deploy spec templates live\nin this skill's `references\u002F` folder with the `spec_template_deploy_*.yaml`\nprefix.\n\n## Reference Map\n\n- `references\u002Fdino-data-specs.md` — dataset contracts, per-action dataset requirements, per-action spec-override examples (train, evaluate, export, deploy\u002Fgen_trt_engine, inference, quantize, distill), data-source arrays, checkpoint inference, and dataset layout.\n- `references\u002Fdino-actions-errors.md` — important parameters, default values, evaluate\u002Fexport defaults, hardware, and the full error-pattern catalog.\n- `references\u002Fdino-tuning-multigpu.md` — full AutoML\u002FHPO notes (metrics, hyperparameters, extractor) and multi-GPU spec consistency.\n- `references\u002Fdino-automl-sdk.md` — AutoML metrics, SDK orchestration internals, data-source gap, and spec-param\u002Fparent-model inference.\n- `references\u002Ftao-deploy-dino.md` — TensorRT deploy workflow.\n- `references\u002Fdetailed-guide.md` — map to the detailed model guide.\n\n## Dataclass Schemas\n\nGenerated TAO Core schemas are packaged in `schemas\u002F\u003Caction>.schema.json`, with `schemas\u002Fmanifest.json` listing available actions. Each generated schema also emits `references\u002Fspec_template_\u003Caction>.yaml` from the schema top-level `default` field. AutoML enablement is declared at the model layer in `references\u002Fskill_info.yaml` via `automl_enabled`. Runnable AutoML still requires `schemas\u002Ftrain.schema.json` and `references\u002Fspec_template_train.yaml` to exist and parse. Use the packaged train schema for `automl_default_parameters`, `automl_disabled_parameters`, defaults, min\u002Fmax bounds, enums, option weights, math conditions, dependencies, and popular parameters. Do not expect `~\u002Ftao-core` at runtime; maintainers regenerate schemas\u002Ftemplates before packaging the skill bank.\n\n## Train Action Policy\n\nThis model is AutoML-enabled at the model layer. Before handling any train-stage request, read `references\u002Fskill_info.yaml` and resolve the run override from either an explicit `automl_policy` value or the user's workflow request. Use `automl_policy: on` by default and only expose `on` \u002F `off` in new launch prompts. Treat phrases like \"turn off AutoML\", \"disable AutoML\", \"no HPO\", or \"plain training\" as `automl_policy: off` for this run only. When `automl_policy: on`, `automl_enabled: true`, and both `schemas\u002Ftrain.schema.json` and `references\u002Fspec_template_train.yaml` are packaged, route the train action through `tao-skill-bank:tao-run-automl` by default with this model's `skill_dir`. Preserve workflow\u002Fapplication overrides for datasets, specs, output directories, GPU\u002Fplatform settings, parent checkpoints, and `automl_policy`. Use direct model training only when `automl_policy: off` or the packaged train schema\u002Ftemplate is missing; in the missing-schema case, report that AutoML is enabled but not runnable for this model until schemas are generated.\n\nNon-train actions such as `evaluate`, `inference`, `export`, and deploy flows stay in this model skill. The per-run `automl_policy` override does not change model metadata.\n\n## Training Requirements\n\nThe agent MUST read this section before generating any training or AutoML script for DINO.\n\n- **Dataset type:** object_detection\n- **Formats:** coco, coco_raw\n- **Accepted dataset intents:** training, evaluation, testing, calibration\n- **Monitoring metric:** mAP50 for quick operational checks; `val_mAP` for\n  COCO\u002Fpaper-style benchmark comparisons.\n\n**Required datasets — MUST resolve both:**\n\n| Dataset | Required | Why |\n|---|---|---|\n| Train dataset URI | Yes | Training data (COCO format) |\n| Validation dataset URI | **Yes — ALWAYS** | DINO unconditionally builds a val dataloader. Omitting `val_data_sources` causes `FileNotFoundError` at startup regardless of the metric or workflow. If the user has no separate eval split, reuse the train URI. |\n\n**Required inputs before generating any training spec:**\n\n1. **Train dataset URI** — S3 path to COCO-format training data\n2. **Validation dataset URI** — S3 path to COCO-format val data (can be same as train)\n3. **`num_classes`** — How many object classes? Default 91 (COCO). Must be >= `max(category_id) + 1`. Too low causes `CUDA error: device-side assert triggered`.\n\nResolve these from the user request or the default profile below. Prompt only\nfor values that are still missing after applying the profile rules.\n\n**Bankable local default profile for DINO AutoML smoke runs:**\n\nUse this profile only when the user asks to run DINO AutoML and does not provide\ndataset or class-count inputs. This profile is intentionally small and local to\nthis skill bank; it is for smoke\u002Fiteration runs, not a production benchmark.\nDo not search previous runners, logs, session state, shell history, or the home\ndirectory to recover these values.\n\n```python\nDINO_AUTOML_PROFILE = {\n    \"train_dataset_uri\": \"s3:\u002F\u002Fnvcf-storage-handling\u002Fdata\u002Ftao_od_synthetic_subset_train_no_convert\",\n    \"validation_dataset_uri\": \"s3:\u002F\u002Fnvcf-storage-handling\u002Fdata\u002Ftao_od_synthetic_subset_val_no_convert\",\n    \"object_classes\": 4,\n    \"dataset_num_classes\": 5,\n    \"image_archive\": \"images.tar.gz\",\n    \"annotation_file\": \"annotations.json\",\n    \"max_recommendations\": 10,\n    \"train_num_epochs\": 10,\n    \"train_checkpoint_interval\": 10,\n    \"train_validation_interval\": 1,\n    \"train_num_gpus\": 1,\n}\n```\n\nIf the user supplies any dataset URI or class-count value, prefer the user value\nand ask for any remaining required DINO value. Do not partially mix a user's\ncustom dataset with this profile's class count unless the user confirms it.\n\n**Do not prompt for image layout for the standard DINO dataset.** The standard\nTAO DINO dataset artifact is `images.tar.gz` plus `annotations.json`. Use\n`images.tar.gz` in the remote `image_dir` spec override. The SDK downloads the\narchive and rewrites the runtime spec to the extracted folder named after the\narchive stem (`images.tar.gz` -> `images`). Only deviate if the user explicitly\nprovides a different image artifact name.\n\n## Core Workflow\n\nDINO supports train, evaluate, export, distill, quantize, and inference. Data-source\noverrides are **mandatory for every action** — DINO's `config.json` has empty\n`data_sources` because the runner cannot auto-resolve array-of-objects spec keys.\nThe agent MUST construct data source paths and include them in `spec_overrides`.\n\nSee `references\u002Fdino-data-specs.md` for the per-action dataset requirements table,\nthe standard dataset artifact (`images.tar.gz` + `annotations.json`) and runtime\nfolder rewrite rules, and the complete per-action `spec_overrides` examples for\ntrain, evaluate, export, deploy\u002Fgen_trt_engine, inference, quantize, and distill —\nincluding checkpoint inference via `parent_model`, the `results_dir\u002Ftrain\u002F`\ncheckpoint location, and the distillation FAN-teacher \u002F student rules.\n\n## Important Parameters And Defaults\n\nKey defaults: `num_epochs=10`, `batch_size=4`, `learning_rate=2e-4`,\n`lr_backbone=2e-5`, `num_classes=91`, `backbone=resnet_50`.\n\n- **dataset.num_classes**: Default 91 (COCO). Must be >= `max(category_id) + 1`. Too low causes `CUDA error: device-side assert triggered`. Set as `\u003Cnum_classes> + 1` in spec overrides.\n- **num_epochs**: default 10 (quick iteration); real datasets typically need 30-50+ epochs for good mAP.\n\nSee `references\u002Fdino-actions-errors.md` for the full parameter list (backbone\noptions, `train.optim.lr`\u002F`lr_steps`, `model.num_queries`, `batch_size`),\ndefault values, evaluate defaults, export defaults (input 960x544, opset 17,\nTRT data types, workspace 1024 MB), and hardware requirements.\n\n## Multi-GPU And AutoML \u002F HPO\n\nWhen increasing `train.num_gpus`, also set `train.gpu_ids` to the same visible\ndevice range, or distributed startup can be inconsistent.\n\nAutoML runs training — all **Training Requirements** above apply. For no-input\nlocal smoke runs, use `DINO_AUTOML_PROFILE`. Recommended metric is `mAP50`\n(`val_mAP` for benchmark comparisons) with `direction=\"maximize\"` and a custom\n`metric_extractor`.\n\nSee `references\u002Fdino-tuning-multigpu.md` for the full multi-GPU spec-consistency\nrule (8-GPU example, NCCL timeout note) and the full AutoML\u002FHPO notes (metric\nselection, `metric_extractor`, recommended hyperparameters, `weight_decay`\nbehavior, dense-dataset resume guidance). See `references\u002Fdino-automl-sdk.md` for\nAutoML metric extractor code, SDK orchestration internals, and parent-model\ninference mappings.\n\n## Error Patterns\n\nCommon failures include CUDA OOM (reduce `batch_size`), missing `val_data_sources`\n(`FileNotFoundError` at startup — always supply val), `num_classes` too low (`CUDA\ndevice-side assert`), and the parent `dino gen_trt_engine` \u002F `dino convert` PyT-CLI\nrestrictions.\n\nSee `references\u002Fdino-actions-errors.md` for the complete error-pattern catalog\nwith diagnostics and fixes.\n\n## Spec Param \u002F Parent Model Inference\n\nModel-specific inference mappings belong in this MD file, not in `config.json`.\nGenerated runners read the mappings and apply them with SDK helpers before\n`create_job()`. For `parent_model`\u002F`parent_model_folder`, pass the upstream\ntrain\u002Fexport\u002FAutoML child job id as `parent_job_id`; the SDK lists the parent\nresult folder, filters checkpoint artifacts, and returns the selected model.\n\nSee `references\u002Fdino-automl-sdk.md` for the full inference-mapping table (per\naction: `parent_model`, `key`, `output_dir`, `ptm_if_no_resume_model`,\n`resume_model`, `create_onnx_file`) and the TensorRT-mapping note. TensorRT\nmappings live in the deploy workflow, not the PyT model skill.\n\n## Optional: running via the TAO SDK\n\nWhen running DINO through the TAO SDK (`script_runner` orchestration, S3 I\u002FO\nwrapping, AutoML), skills read `references\u002Fskill_info.yaml` for input and\nspec-param mappings. See `references\u002Fdino-automl-sdk.md` for SDK orchestration\ninternals, including the data-sources gap and the `[0]`-indexed `inputs`\ndeclarations. Skip this when running locally with `docker run`.\n\n## Deployment\n\n- [tao-deploy-dino](references\u002Ftao-deploy-dino.md)\n","# DINO\n\nDINO (DETR with Improved DeNoising Anchor Boxes) for 2D object detection. Transformer-based detector with denoising training, multi-scale features, and optional distillation support.\n\nUses pretrained backbone weights (e.g. ResNet-50 ImageNet). Set `model.pretrained_backbone_path` for backbone-only or `train.pretrained_model_path` for full model.\n\n## When To Use\n\nTrain, evaluate, export, distill, quantize, or run inference for a TAO DINO 2D object detector.\n\nFor TAO Deploy TensorRT actions (`gen_trt_engine`, TensorRT `evaluate`, and\nTensorRT `inference`), read `references\u002Ftao-deploy-dino.md` first. Deploy spec templates live\nin this skill's `references\u002F` folder with the `spec_template_deploy_*.yaml`\nprefix.\n\n## Reference Map\n\n- `references\u002Fdino-data-specs.md` — dataset contracts, per-action dataset requirements, per-action spec-override examples (train, evaluate, export, deploy\u002Fgen_trt_engine, inference, quantize, distill), data-source arrays, checkpoint inference, and dataset layout.\n- `references\u002Fdino-actions-errors.md` — important parameters, default values, evaluate\u002Fexport defaults, hardware, and the full error-pattern catalog.\n- `references\u002Fdino-tuning-multigpu.md` — full AutoML\u002FHPO notes (metrics, hyperparameters, extractor) and multi-GPU spec consistency.\n- `references\u002Fdino-automl-sdk.md` — AutoML metrics, SDK orchestration internals, data-source gap, and spec-param\u002Fparent-model inference.\n- `references\u002Ftao-deploy-dino.md` — TensorRT deploy workflow.\n- `references\u002Fdetailed-guide.md` — map to the detailed model guide.\n\n## Dataclass Schemas\n\nGenerated TAO Core schemas are packaged in `schemas\u002F\u003Caction>.schema.json`, with `schemas\u002Fmanifest.json` listing available actions. Each generated schema also emits `references\u002Fspec_template_\u003Caction>.yaml` from the schema top-level `default` field. AutoML enablement is declared at the model layer in `references\u002Fskill_info.yaml` via `automl_enabled`. Runnable AutoML still requires `schemas\u002Ftrain.schema.json` and `references\u002Fspec_template_train.yaml` to exist and parse. Use the packaged train schema for `automl_default_parameters`, `automl_disabled_parameters`, defaults, min\u002Fmax bounds, enums, option weights, math conditions, dependencies, and popular parameters. Do not expect `~\u002Ftao-core` at runtime; maintainers regenerate schemas\u002Ftemplates before packaging the skill bank.\n\n## Train Action Policy\n\nThis model is AutoML-enabled at the model layer. Before handling any train-stage request, read `references\u002Fskill_info.yaml` and resolve the run override from either an explicit `automl_policy` value or the user's workflow request. Use `automl_policy: on` by default and only expose `on` \u002F `off` in new launch prompts. Treat phrases like \"turn off AutoML\", \"disable AutoML\", \"no HPO\", or \"plain training\" as `automl_policy: off` for this run only. When `automl_policy: on`, `automl_enabled: true`, and both `schemas\u002Ftrain.schema.json` and `references\u002Fspec_template_train.yaml` are packaged, route the train action through `tao-skill-bank:tao-run-automl` by default with this model's `skill_dir`. Preserve workflow\u002Fapplication overrides for datasets, specs, output directories, GPU\u002Fplatform settings, parent checkpoints, and `automl_policy`. Use direct model training only when `automl_policy: off` or the packaged train schema\u002Ftemplate is missing; in the missing-schema case, report that AutoML is enabled but not runnable for this model until schemas are generated.\n\nNon-train actions such as `evaluate`, `inference`, `export`, and deploy flows stay in this model skill. The per-run `automl_policy` override does not change model metadata.\n\n## Training Requirements\n\nThe agent MUST read this section before generating any training or AutoML script for DINO.\n\n- **Dataset type:** object_detection\n- **Formats:** coco, coco_raw\n- **Accepted dataset intents:** training, evaluation, testing, calibration\n- **Monitoring metric:** mAP50 for quick operational checks; `val_mAP` for\n  COCO\u002Fpaper-style benchmark comparisons.\n\n**Required datasets — MUST resolve both:**\n\n| Dataset | Required | Why |\n|---|---|---|\n| Train dataset URI | Yes | Training data (COCO format) |\n| Validation dataset URI | **Yes — ALWAYS** | DINO unconditionally builds a val dataloader. Omitting `val_data_sources` causes `FileNotFoundError` at startup regardless of the metric or workflow. If the user has no separate eval split, reuse the train URI. |\n\n**Required inputs before generating any training spec:**\n\n1. **Train dataset URI** — S3 path to COCO-format training data\n2. **Validation dataset URI** — S3 path to COCO-format val data (can be same as train)\n3. **`num_classes`** — How many object classes? Default 91 (COCO). Must be >= `max(category_id) + 1`. Too low causes `CUDA error: device-side assert triggered`.\n\nResolve these from the user request or the default profile below. Prompt only\nfor values that are still missing after applying the profile rules.\n\n**Bankable local default profile for DINO AutoML smoke runs:**\n\nUse this profile only when the user asks to run DINO AutoML and does not provide\ndataset or class-count inputs. This profile is intentionally small and local to\nthis skill bank; it is for smoke\u002Fiteration runs, not a production benchmark.\nDo not search previous runners, logs, session state, shell history, or the home\ndirectory to recover these values.\n\n```python\nDINO_AUTOML_PROFILE = {\n    \"train_dataset_uri\": \"s3:\u002F\u002Fnvcf-storage-handling\u002Fdata\u002Ftao_od_synthetic_subset_train_no_convert\",\n    \"validation_dataset_uri\": \"s3:\u002F\u002Fnvcf-storage-handling\u002Fdata\u002Ftao_od_synthetic_subset_val_no_convert\",\n    \"object_classes\": 4,\n    \"dataset_num_classes\": 5,\n    \"image_archive\": \"images.tar.gz\",\n    \"annotation_file\": \"annotations.json\",\n    \"max_recommendations\": 10,\n    \"train_num_epochs\": 10,\n    \"train_checkpoint_interval\": 10,\n    \"train_validation_interval\": 1,\n    \"train_num_gpus\": 1,\n}\n```\n\nIf the user supplies any dataset URI or class-count value, prefer the user value\nand ask for any remaining required DINO value. Do not partially mix a user's\ncustom dataset with this profile's class count unless the user confirms it.\n\n**Do not prompt for image layout for the standard DINO dataset.** The standard\nTAO DINO dataset artifact is `images.tar.gz` plus `annotations.json`. Use\n`images.tar.gz` in the remote `image_dir` spec override. The SDK downloads the\narchive and rewrites the runtime spec to the extracted folder named after the\narchive stem (`images.tar.gz` -> `images`). Only deviate if the user explicitly\nprovides a different image artifact name.\n\n## Core Workflow\n\nDINO supports train, evaluate, export, distill, quantize, and inference. Data-source\noverrides are **mandatory for every action** — DINO's `config.json` has empty\n`data_sources` because the runner cannot auto-resolve array-of-objects spec keys.\nThe agent MUST construct data source paths and include them in `spec_overrides`.\n\nSee `references\u002Fdino-data-specs.md` for the per-action dataset requirements table,\nthe standard dataset artifact (`images.tar.gz` + `annotations.json`) and runtime\nfolder rewrite rules, and the complete per-action `spec_overrides` examples for\ntrain, evaluate, export, deploy\u002Fgen_trt_engine, inference, quantize, and distill —\nincluding checkpoint inference via `parent_model`, the `results_dir\u002Ftrain\u002F`\ncheckpoint location, and the distillation FAN-teacher \u002F student rules.\n\n## Important Parameters And Defaults\n\nKey defaults: `num_epochs=10`, `batch_size=4`, `learning_rate=2e-4`,\n`lr_backbone=2e-5`, `num_classes=91`, `backbone=resnet_50`.\n\n- **dataset.num_classes**: Default 91 (COCO). Must be >= `max(category_id) + 1`. Too low causes `CUDA error: device-side assert triggered`. Set as `\u003Cnum_classes> + 1` in spec overrides.\n- **num_epochs**: default 10 (quick iteration); real datasets typically need 30-50+ epochs for good mAP.\n\nSee `references\u002Fdino-actions-errors.md` for the full parameter list (backbone\noptions, `train.optim.lr`\u002F`lr_steps`, `model.num_queries`, `batch_size`),\ndefault values, evaluate defaults, export defaults (input 960x544, opset 17,\nTRT data types, workspace 1024 MB), and hardware requirements.\n\n## Multi-GPU And AutoML \u002F HPO\n\nWhen increasing `train.num_gpus`, also set `train.gpu_ids` to the same visible\ndevice range, or distributed startup can be inconsistent.\n\nAutoML runs training — all **Training Requirements** above apply. For no-input\nlocal smoke runs, use `DINO_AUTOML_PROFILE`. Recommended metric is `mAP50`\n(`val_mAP` for benchmark comparisons) with `direction=\"maximize\"` and a custom\n`metric_extractor`.\n\nSee `references\u002Fdino-tuning-multigpu.md` for the full multi-GPU spec-consistency\nrule (8-GPU example, NCCL timeout note) and the full AutoML\u002FHPO notes (metric\nselection, `metric_extractor`, recommended hyperparameters, `weight_decay`\nbehavior, dense-dataset resume guidance). See `references\u002Fdino-automl-sdk.md` for\nAutoML metric extractor code, SDK orchestration internals, and parent-model\ninference mappings.\n\n## Error Patterns\n\nCommon failures include CUDA OOM (reduce `batch_size`), missing `val_data_sources`\n(`FileNotFoundError` at startup — always supply val), `num_classes` too low (`CUDA\ndevice-side assert`), and the parent `dino gen_trt_engine` \u002F `dino convert` PyT-CLI\nrestrictions.\n\nSee `references\u002Fdino-actions-errors.md` for the complete error-pattern catalog\nwith diagnostics and fixes.\n\n## Spec Param \u002F Parent Model Inference\n\nModel-specific inference mappings belong in this MD file, not in `config.json`.\nGenerated runners read the mappings and apply them with SDK helpers before\n`create_job()`. For `parent_model`\u002F`parent_model_folder`, pass the upstream\ntrain\u002Fexport\u002FAutoML child job id as `parent_job_id`; the SDK lists the parent\nresult folder, filters checkpoint artifacts, and returns the selected model.\n\nSee `references\u002Fdino-automl-sdk.md` for the full inference-mapping table (per\naction: `parent_model`, `key`, `output_dir`, `ptm_if_no_resume_model`,\n`resume_model`, `create_onnx_file`) and the TensorRT-mapping note. TensorRT\nmappings live in the deploy workflow, not the PyT model skill.\n\n## Optional: running via the TAO SDK\n\nWhen running DINO through the TAO SDK (`script_runner` orchestration, S3 I\u002FO\nwrapping, AutoML), skills read `references\u002Fskill_info.yaml` for input and\nspec-param mappings. See `references\u002Fdino-automl-sdk.md` for SDK orchestration\ninternals, including the data-sources gap and the `[0]`-indexed `inputs`\ndeclarations. Skip this when running locally with `docker run`.\n\n## Deployment\n\n- [tao-deploy-dino](references\u002Ftao-deploy-dino.md)\n",{"name":5,"tags":24,"license":15,"metadata":27,"description":7,"allowed-tools":30,"compatibility":31},[25,26],"object","detection",{"author":28,"version":29},"NVIDIA Corporation","0.1.0","Read Bash","Requires docker + nvidia-container-toolkit.","main","8859218955d9504cd18b59e7892100718039d3b4","data\u002Fskills\u002Fnvidia-skills\u002Ftao-train-dino",1,"https:\u002F\u002Fgithub.com\u002Fom-scogo\u002Fskillsh-scraper","https:\u002F\u002Fgithub.com\u002Fom-scogo","SDE-1 @ Scogo ","2026-07-18 12:22:46",{"source":8,"skillName":5,"commands":41,"prompt":51,"zipAvailable":52},[42,45,48],{"tool":43,"command":44},"npx","npx skills add om-scogo\u002Fskillsh-scraper --skill tao-train-dino",{"tool":46,"command":47},"bunx","bunx skills add om-scogo\u002Fskillsh-scraper --skill tao-train-dino",{"tool":49,"command":50},"pnpm","pnpm dlx skills add om-scogo\u002Fskillsh-scraper --skill tao-train-dino","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002Fom-scogo\u002Fskillsh-scraper\u002Fblob\u002Fmain\u002Fdata\u002Fskills\u002Fnvidia-skills\u002Ftao-train-dino\u002FSkill.md",true,[54],{"path":55,"resourceType":56,"resourceTypeDesc":57,"sizeBytes":58,"sourceHtmlUrl":59},"metadata.json","REFERENCE","参考文档",406,"https:\u002F\u002Fgithub.com\u002Fom-scogo\u002Fskillsh-scraper\u002Fblob\u002Fmain\u002Fdata\u002Fskills\u002Fnvidia-skills\u002Ftao-train-dino\u002Fmetadata.json",[61,69,74,81,86,93],{"skillNo":62,"name":63,"nameZh":6,"description":64,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":65,"licenseFileUrl":6,"sourceHtmlUrl":66,"categories":67,"keywords":68,"publishedAt":19},"01M7D6MC9FAAEV4CMNYX8W7JTG","arxiv-search","Search arXiv physics, math, and computer science preprints using natural language queries. Powered by Valyu semantic search.","MIT","https:\u002F\u002Fgithub.com\u002Fom-scogo\u002Fskillsh-scraper\u002Fblob\u002Fmain\u002Fdata\u002Fskills\u002Fyorkeccak-scientific-skills\u002Farxiv-search\u002Fskills\u002Farxiv-search\u002FSKILL.md",[],[],{"skillNo":70,"name":63,"nameZh":6,"description":64,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":65,"licenseFileUrl":6,"sourceHtmlUrl":71,"categories":72,"keywords":73,"publishedAt":19},"01M7D6M85CKKN87TA98VCQV4H7","https:\u002F\u002Fgithub.com\u002Fom-scogo\u002Fskillsh-scraper\u002Fblob\u002Fmain\u002Fdata\u002Fskills\u002Fyorkeccak-scientific-skills\u002Farxiv-search\u002FSkill.md",[],[],{"skillNo":75,"name":76,"nameZh":6,"description":77,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":65,"licenseFileUrl":6,"sourceHtmlUrl":78,"categories":79,"keywords":80,"publishedAt":19},"01M7D4ZTCS4DR3CE4JY06X84T2","working-with-legacy-code","Safely change and test untested codebases using Feathers' \"Working Effectively with Legacy Code\". Use when the user mentions \"legacy code\", \"no tests\", \"untested codebase\", \"how do I test this\", \"seams\", \"characterization tests\", \"golden master\", \"sprout method\", \"afraid to change this code\", \"monster method\", \"dependency breaking\", or \"inherited a messy codebase\". Also trigger when changing code without tests safely, getting a class under test when constructors, statics, or singletons block it, adding features to tangled modules, or planning incremental test coverage for an old codebase. Covers the legacy-code change algorithm, seams, characterization tests, sprout\u002Fwrap, and dependency-breaking techniques. For refactoring code that already has tests, see refactoring-patterns. For day-to-day code quality, see clean-code.","https:\u002F\u002Fgithub.com\u002Fom-scogo\u002Fskillsh-scraper\u002Fblob\u002Fmain\u002Fdata\u002Fskills\u002Fwondelai-skills\u002Fworking-with-legacy-code\u002Fworking-with-legacy-code\u002FSKILL.md",[],[],{"skillNo":82,"name":76,"nameZh":6,"description":77,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":65,"licenseFileUrl":6,"sourceHtmlUrl":83,"categories":84,"keywords":85,"publishedAt":19},"01M7D4ZHZKJD2RSNGAF63V14P4","https:\u002F\u002Fgithub.com\u002Fom-scogo\u002Fskillsh-scraper\u002Fblob\u002Fmain\u002Fdata\u002Fskills\u002Fwondelai-skills\u002Fworking-with-legacy-code\u002FSkill.md",[],[],{"skillNo":87,"name":88,"nameZh":6,"description":89,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":65,"licenseFileUrl":6,"sourceHtmlUrl":90,"categories":91,"keywords":92,"publishedAt":19},"01M7D4ZB4CJBFW5JE67CPH2QRV","web-typography","Select, pair, and implement typefaces for web projects. Use when the user mentions \"font pairing\", \"which typeface\", \"line height\", \"responsive typography\", \"web font loading\", \"type hierarchy\", \"variable fonts\", \"FOUT\u002FFOIT\", \"typographic scale\", or \"the text is hard to read\". Also trigger when choosing between system fonts and web fonts, optimizing font-loading performance, or designing readable long-form content. Covers readability evaluation, CSS implementation, and performance optimization. For overall UI design systems, see refactoring-ui. For dramatic typographic experiences, see top-design.","https:\u002F\u002Fgithub.com\u002Fom-scogo\u002Fskillsh-scraper\u002Fblob\u002Fmain\u002Fdata\u002Fskills\u002Fwondelai-skills\u002Fweb-typography\u002Fweb-typography\u002FSKILL.md",[],[],{"skillNo":94,"name":88,"nameZh":6,"description":89,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":65,"licenseFileUrl":6,"sourceHtmlUrl":95,"categories":96,"keywords":97,"publishedAt":19},"01M7D4Z4MVYVY3NMJ18BNV6WNA","https:\u002F\u002Fgithub.com\u002Fom-scogo\u002Fskillsh-scraper\u002Fblob\u002Fmain\u002Fdata\u002Fskills\u002Fwondelai-skills\u002Fweb-typography\u002FSkill.md",[],[]]