[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M78AYPP2DVKRMW7SWTEV71TT":3,"skill-same:01M78AYPP2DVKRMW7SWTEV71TT":120},{"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":14,"repositoryHtmlUrl":35,"repositoryHomepageUrl":36,"ownerHtmlUrl":37,"ownerBlogUrl":38,"ownerDescription":39,"updatedAt":40,"install":41,"resources":54},"01M78AYPP2DVKRMW7SWTEV71TT","tao-train-rtdetr",null,"RT-DETR (Real-Time DEtection TRansformer) for 2D object detection. Designed for real-time inference with competitive accuracy and supports distillation and quantization for deployment optimization. Use when training, evaluating, distilling, quantizing, exporting, or running inference for a TAO RT-DETR model. Trigger phrases include \"train RT-DETR\", \"real-time DETR\", \"low-latency object detection\", \"RT-DETR distillation \u002F quantization\".","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\u002Fpublic\u002Fapi\u002Fskills\u002Fnvidia\u002Ftao-train-rtdetr\u002FSKILL.md",[],[],"2026-11-15 14:19:17",false,"---\nname: tao-train-rtdetr\ndescription: RT-DETR (Real-Time DEtection TRansformer) for 2D object detection. Designed for real-time inference with\n  competitive accuracy and supports distillation and quantization for deployment optimization. Use when training, evaluating,\n  distilling, quantizing, exporting, or running inference for a TAO RT-DETR model. Trigger phrases include \"train RT-DETR\",\n  \"real-time DETR\", \"low-latency object detection\", \"RT-DETR distillation \u002F quantization\".\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# RT-DETR\n\nRT-DETR (Real-Time DEtection TRansformer) for 2D object detection. Designed for real-time inference with competitive accuracy. Supports distillation and quantization for deployment optimization.\n\nSet model.pretrained_backbone_path for backbone weights or train.pretrained_model_path for full model.\n\nFor TAO Deploy TensorRT actions (`gen_trt_engine`, TensorRT `evaluate`, and TensorRT `inference`), read `references\u002Ftao-deploy-rtdetr.md` first. Deploy spec templates live in this skill's `references\u002F` folder with the `spec_template_deploy_*.yaml` prefix.\n\n## Dataclass Schemas\n\nGenerated TAO Core schemas are packaged in `schemas\u002F\u003Caction>.schema.json`, with `schemas\u002Fmanifest.json` listing available actions. Each generated schema also emits `references\u002Fspec_template_\u003Caction>.yaml` from the schema top-level `default` field. AutoML enablement is declared at the model layer in `references\u002Fskill_info.yaml` via `automl_enabled`. Runnable AutoML still requires `schemas\u002Ftrain.schema.json` and `references\u002Fspec_template_train.yaml` to exist and parse. Use the packaged train schema for `automl_default_parameters`, `automl_disabled_parameters`, defaults, min\u002Fmax bounds, enums, option weights, math conditions, dependencies, and popular parameters. Do not expect `~\u002Ftao-core` at runtime; maintainers regenerate schemas\u002Ftemplates before packaging the skill bank.\n\n## Train Action Policy\n\nThis model is AutoML-enabled at the model layer. Before handling any train-stage request, read `references\u002Fskill_info.yaml` and resolve the run override from either an explicit `automl_policy` value or the user's workflow request. Use `automl_policy: on` by default and only expose `on` \u002F `off` in new launch prompts. Treat phrases like \"turn off AutoML\", \"disable AutoML\", \"no HPO\", or \"plain training\" as `automl_policy: off` for this run only. When `automl_policy: on`, `automl_enabled: true`, and both `schemas\u002Ftrain.schema.json` and `references\u002Fspec_template_train.yaml` are packaged, route the train action through `tao-skill-bank:tao-run-automl` by default with this model's `skill_dir`. Preserve workflow\u002Fapplication overrides for datasets, specs, output directories, GPU\u002Fplatform settings, parent checkpoints, and `automl_policy`. Use direct model training only when `automl_policy: off` or the packaged train schema\u002Ftemplate is missing; in the missing-schema case, report that AutoML is enabled but not runnable for this model until schemas are generated.\n\nNon-train actions such as `evaluate`, `inference`, `export`, and deploy flows stay in this model skill. The per-run `automl_policy` override does not change model metadata.\n\n## Supported Actions\n\nThe packaged RT-DETR PyT CLI supports `train`, `distill`, `quantize`, `evaluate`, `export`, `inference`, and `default_specs`. This model skill exposes `train`, `distill`, `quantize`, `evaluate`, `export`, and `inference`; resume\u002Fretrain is performed through `train` with `train.resume_training_checkpoint_path`.\n\nThe parent PyT CLI does not expose `gen_trt_engine`. Use `models\u002Frtdetr\u002Fdeploy` for TensorRT engine generation, TensorRT evaluation, and TensorRT inference.\n\n## Training Requirements\n\n- **Dataset type:** object_detection\n- **Formats:** coco, coco_raw\n- **Monitoring metric:** mAP50 for quick operational checks; `val_mAP` for COCO\u002Fpaper-style benchmark comparisons.\n\n### Per-Action Dataset Requirements\n\n| Action | Spec Key | Source | Files | List? |\n|---|---|---|---|---|\n| distill | dataset.train_data_sources | train_datasets | image_dir: images.tar.gz, json_file: annotations.json | Yes |\n| distill | dataset.val_data_sources | eval_dataset | image_dir: images.tar.gz, json_file: annotations.json | No |\n| evaluate | dataset.test_data_sources | eval_dataset | image_dir: images.tar.gz, json_file: annotations.json | No |\n| inference | dataset.infer_data_sources | inference_dataset | image_dir: images.tar.gz, classmap: label_map.txt | Yes |\n| quantize | dataset.train_data_sources | train_datasets | image_dir: images.tar.gz, json_file: annotations.json | Yes |\n| quantize | dataset.val_data_sources | eval_dataset | image_dir: images.tar.gz, json_file: annotations.json | No |\n| quantize | dataset.quant_calibration_data_sources | train_datasets | image_dir: images.tar.gz, json_file: annotations.json | No |\n| train | dataset.train_data_sources | train_datasets | image_dir: images.tar.gz, json_file: annotations.json | Yes |\n| train | dataset.val_data_sources | eval_dataset | image_dir: images.tar.gz, json_file: annotations.json | No |\n\n### Typical Spec Overrides\n\nData source overrides are **mandatory for every action** — the agent MUST construct data source paths from the Per-Action Dataset Requirements table above and include them in `spec_overrides`.\n\n```python\nS3_TRAIN = \"s3:\u002F\u002Fbucket\u002Fdata\u002Ftrain\"\nS3_EVAL = \"s3:\u002F\u002Fbucket\u002Fdata\u002Feval\"\nCHECKPOINT = \"\u002Fresults\u002F{train_job_id}\u002Fresults_dir\u002Fmodel_epoch_000.pth\"\nONNX_FILE = \"\u002Fresults\u002F{export_job_id}\u002Fresults_dir\u002Frtdetr.onnx\"\n```\n\n**train (mandatory data sources):**\n```python\n{\n    \"train.num_epochs\": 10,\n    \"train.checkpoint_interval\": 10,\n    \"train.validation_interval\": 10,\n    \"train.num_gpus\": 1,\n    \"train.gpu_ids\": [0],\n    \"dataset.num_classes\": \"\u003Cnum_classes> + 1\",\n    \"dataset.train_data_sources\": [{\"image_dir\": f\"{S3_TRAIN}\u002Fimages.tar.gz\", \"json_file\": f\"{S3_TRAIN}\u002Fannotations.json\"}],\n    \"dataset.val_data_sources\": {\"image_dir\": f\"{S3_EVAL}\u002Fimages.tar.gz\", \"json_file\": f\"{S3_EVAL}\u002Fannotations.json\"},\n}\n```\n\n**resume train (mandatory checkpoint):**\n```python\n{\n    \"train.num_epochs\": 11,\n    \"train.resume_training_checkpoint_path\": CHECKPOINT,\n    \"dataset.num_classes\": \"\u003Cnum_classes> + 1\",\n    \"dataset.eval_class_ids\": [1, 2, 3, 4],\n    \"dataset.train_data_sources\": [{\"image_dir\": f\"{S3_TRAIN}\u002Fimages.tar.gz\", \"json_file\": f\"{S3_TRAIN}\u002Fannotations.json\"}],\n    \"dataset.val_data_sources\": {\"image_dir\": f\"{S3_EVAL}\u002Fimages.tar.gz\", \"json_file\": f\"{S3_EVAL}\u002Fannotations.json\"},\n}\n```\n\n**evaluate (mandatory data sources and checkpoint):**\n```python\n{\n    \"dataset.num_classes\": \"\u003Cnum_classes> + 1\",\n    \"dataset.eval_class_ids\": [1, 2, 3, 4],\n    \"dataset.test_data_sources\": {\"image_dir\": f\"{S3_EVAL}\u002Fimages.tar.gz\", \"json_file\": f\"{S3_EVAL}\u002Fannotations.json\"},\n    \"evaluate.checkpoint\": CHECKPOINT,\n}\n```\n\n**export (mandatory checkpoint and output):**\n```python\n{\n    \"dataset.num_classes\": \"\u003Cnum_classes> + 1\",\n    \"export.checkpoint\": CHECKPOINT,\n    \"export.onnx_file\": ONNX_FILE,\n    \"export.input_height\": 640,\n    \"export.input_width\": 640,\n}\n```\n\n**quantize (mandatory data sources):**\n```python\n{\n    \"dataset.num_classes\": \"\u003Cnum_classes> + 1\",\n    \"quantize.layers\": [\n        {\n            \"module_name\": \"*\",\n            \"weights\": {\n                \"dtype\": \"float8_e4m3fn\"\n            },\n            \"activations\": {\n                \"dtype\": \"float8_e4m3fn\"\n            }\n        }\n    ],\n    \"dataset.train_data_sources\": [{\"image_dir\": f\"{S3_TRAIN}\u002Fimages.tar.gz\", \"json_file\": f\"{S3_TRAIN}\u002Fannotations.json\"}],\n    \"dataset.val_data_sources\": {\"image_dir\": f\"{S3_EVAL}\u002Fimages.tar.gz\", \"json_file\": f\"{S3_EVAL}\u002Fannotations.json\"},\n    \"dataset.quant_calibration_data_sources\": {\"image_dir\": f\"{S3_TRAIN}\u002Fimages.tar.gz\", \"json_file\": f\"{S3_TRAIN}\u002Fannotations.json\"},\n    \"quantize.model_path\": CHECKPOINT,\n}\n```\n\n**inference (mandatory data sources and checkpoint):**\n```python\n{\n    \"dataset.num_classes\": \"\u003Cnum_classes> + 1\",\n    \"dataset.infer_data_sources\": {\"image_dir\": [f\"{S3_EVAL}\u002Fimages.tar.gz\"], \"classmap\": f\"{S3_EVAL}\u002Flabel_map.txt\"},\n    \"inference.checkpoint\": CHECKPOINT,\n}\n```\n\n**distill (mandatory data sources and teacher checkpoint):**\n```python\n{\n    \"dataset.train_data_sources\": [{\"image_dir\": f\"{S3_TRAIN}\u002Fimages.tar.gz\", \"json_file\": f\"{S3_TRAIN}\u002Fannotations.json\"}],\n    \"dataset.val_data_sources\": {\"image_dir\": f\"{S3_EVAL}\u002Fimages.tar.gz\", \"json_file\": f\"{S3_EVAL}\u002Fannotations.json\"},\n    \"distill.pretrained_teacher_model_path\": CHECKPOINT,\n}\n```\n## Eval Dataset\n\nOptional. Provides validation mAP at each checkpoint if supplied.\n\n## Important Parameters\n\n- **dataset.num_classes**: Number of classes. Default 80 (MSCOCO 80-class). Must match your dataset annotations.\n- **model.backbone**: Default resnet_50. Supported: ResNet variants, ConvNeXt, FAN, EfficientViT. RT-DETR is optimized for real-time with lighter backbones.\n- **train.optim.lr**: Learning rate. Default 1e-4 (lower than DINO's 2e-4). lr_backbone defaults to 1e-5.\n- **dataset.augmentation.train_spatial_size**: Training input size. Default [640, 640]. Smaller than DINO's multi-scale (up to 1333). Key to RT-DETR's speed.\n- **model.num_feature_levels**: Default 3 (vs DINO's 4). return_interm_indices is [1,2,3].\n- **train.enable_ema**: Exponential moving average. Default False. Enable for potentially smoother convergence.\n- **dataset.remap_mscoco_category**: Default False. Set True only for original MSCOCO dataset with 91-to-80 category ID remapping.\n\n## Multi-GPU \u002F Multi-Node\n\n**Launch method:** `torchrun` (LIGHTNING_EXCLUDED_NETWORK). The entrypoint runs `torchrun --nnodes=N --nproc-per-node=M train.py`, NOT plain `python`.\n\n| Spec Key | Description | Default |\n|----------|-------------|---------|\n| `train.num_gpus` | Number of GPUs per node | 1 |\n| `train.gpu_ids` | GPU device indices | [0] |\n| `train.num_nodes` | Number of nodes | 1 |\n| `train.distributed_strategy` | `ddp` or `fsdp` | `ddp` |\n\n- When increasing `train.num_gpus`, also set `train.gpu_ids` to the same\n  visible device range. For example, an 8-GPU single-node Slurm run must\n  include both `\"train.num_gpus\": 8` and\n  `\"train.gpu_ids\": [0, 1, 2, 3, 4, 5, 6, 7]`.\n- `CUDA_VISIBLE_DEVICES` is explicitly set (unlike Lightning-managed models which use `TAO_VISIBLE_DEVICES`)\n- `ddp` with activation checkpointing: `find_unused_parameters=False`\n- `ddp` without: `find_unused_parameters=True`\n- `fsdp` supported, forces FP16\n\n**Multi-node env vars** (set by orchestrator):\n\n| Variable | Purpose |\n|----------|---------|\n| `WORLD_SIZE` | Number of nodes (triggers multinode mode) |\n| `NODE_RANK` | This node's rank (0-indexed) |\n| `MASTER_ADDR` | Rank-0 node IP |\n| `MASTER_PORT` | Rank-0 port (default 29500) |\n| `NUM_GPU_PER_NODE` | GPUs per node (default: all visible) |\n\n**CRITICAL:** `NODE_RANK` is copied to `RANK` if `RANK` is unset. This is required for torchrun multinode.\n\n## Export \u002F TRT Defaults\n\n- Export input: 640x640, opset 17\n- TRT data types: FP32, FP16, INT8\n- TRT workspace: 1024 MB\n- TRT max_batch_size: 4\n\n## Distillation\n\nRT-DETR supports knowledge distillation with a teacher model. Requires `distill` action with `distill.pretrained_teacher_model_path` and a distillation binding configuration.\n\nUse the packaged `references\u002Fspec_template_distill.yaml` as the starting point. The validated default binding uses the RT-DETR distiller's explicit IOU feature path:\n\n```yaml\ndistill:\n  bindings:\n  - student_module_name: srcs\n    teacher_module_name: srcs\n    criterion: IOU\n    weight: 1.0\n```\n\nDo not substitute DINO-style output names such as `pred_logits` \u002F `pred_boxes`, and do not bind arbitrary decoder heads unless you have verified the module returns captured feature lists. The RT-DETR distiller asserts that IOU bindings must use `srcs` or `dsrcs`.\n\n## Hardware\n\nMinimum 1 GPU(s), recommended 2 GPU(s). 16GB+ (V100 or A100) VRAM per GPU. RT-DETR is more memory-efficient than DINO\u002FGDINO due to smaller input size (640x640) and fewer feature levels. Trains well on single GPU for small-medium datasets.\n\n## Error Patterns\n\n**CUDA out of memory**: Reduce batch_size. RT-DETR at 640x640 is lighter than DINO at 1333px, but batch_size > 8 may still OOM on 16GB GPUs.\n\n**num_classes mismatch**: RT-DETR defaults to 80 (not 91 like DINO). Ensure dataset.num_classes matches your annotation categories.\n\n**CUDA index assert from category IDs**: If COCO category IDs are one-based or otherwise not remapped to zero-based contiguous IDs, set `dataset.num_classes` to `max(category_id) + 1` and keep `dataset.eval_class_ids` aligned to the actual category IDs. For the packaged four-class S3 sample with IDs 1-4, use `dataset.num_classes: 5` and `dataset.eval_class_ids: [1, 2, 3, 4]`.\n\n**return_interm_indices vs num_feature_levels**: Default is [1,2,3] with num_feature_levels=3. Must be consistent if changed.\n\n**Export shape mismatch**: Keep RT-DETR export and deploy consumer input size at\nthe validated `640x640` default unless the model has been trained and checked\nfor a different shape. The older packaged `960x544` template shape can fail\nduring ONNX tracing with `The size of tensor a (...) must match the size of\ntensor b (...)` in `hybrid_encoder.py` positional embedding addition.\n\n**AutoML metric extraction**: RT-DETR emits detection metrics in structured training status and logs. For COCO\u002Fpaper-style benchmark comparisons, optimize `val_mAP` with `direction: maximize`; for explicit AP50 workflows, optimize `mAP50`. Prefer `results_dir\u002Ftrain\u002Fstatus.json` or AutoML result state before parsing raw logs. Do not optimize `val_loss` for default detection model invocations.\n\n**Checkpoint handoff**: For evaluate\u002Fexport\u002Finference\u002Fquantize\u002Fdistill\u002Fresume, use the checkpoint resolver on the best AutoML child job's `results_dir\u002Ftrain\u002F` folder and select the action-appropriate `model_epoch_*.pth` checkpoint. RT-DETR may also write a latest symlink, but that should only be used when a caller explicitly requests latest. Keep `dataset.num_classes`, `dataset.eval_class_ids`, `model.num_queries`, and `model.num_select` consistent with training.\n\n**Parent `rtdetr gen_trt_engine` rejected by the PyT CLI**: In the validated 7.0.0 PyT container, `rtdetr gen_trt_engine` is not a valid parent-model subtask. Use the RT-DETR deploy workflow (`references\u002Ftao-deploy-rtdetr.md`) for TensorRT engine generation, TensorRT evaluation, and TensorRT inference.\n\n## Spec Param \u002F Parent Model Inference\n\nModel-specific inference mappings belong in this MD file, not in `config.json`. Generated runners should read this section and apply the mappings with SDK helpers before `create_job()`. This mirrors the old microservices `infer_params.py` flow.\n\nInference mappings from TAO Core `rtdetr.config.json`:\n\n| Action | Spec Field | Inference Function | Meaning |\n|---|---|---|---|\n| distill | `distill.pretrained_teacher_model_path` | `parent_model` | model file inferred from the parent job results folder |\n| distill | `encryption_key` | `key` | encryption key |\n| distill | `results_dir` | `output_dir` | current job results directory |\n| evaluate | `encryption_key` | `key` | encryption key |\n| evaluate | `evaluate.checkpoint` | `parent_model` | model file inferred from the parent job results folder |\n| evaluate | `evaluate.trt_engine` | `parent_model` | model file inferred from the parent job results folder |\n| evaluate | `results_dir` | `output_dir` | current job results directory |\n| export | `encryption_key` | `key` | encryption key |\n| export | `export.checkpoint` | `parent_model` | model file inferred from the parent job results folder |\n| export | `export.onnx_file` | `create_onnx_file` | output ONNX path |\n| export | `results_dir` | `output_dir` | current job results directory |\n| inference | `encryption_key` | `key` | encryption key |\n| inference | `inference.checkpoint` | `parent_model` | model file inferred from the parent job results folder |\n| inference | `inference.trt_engine` | `parent_model` | model file inferred from the parent job results folder |\n| inference | `results_dir` | `output_dir` | current job results directory |\n| quantize | `encryption_key` | `key` | encryption key |\n| quantize | `quantize.model_path` | `parent_model` | model file inferred from the parent job results folder |\n| quantize | `results_dir` | `output_dir` | current job results directory |\n| train | `encryption_key` | `key` | encryption key |\n| train | `model.pretrained_backbone_path` | `ptm_if_no_resume_model` | PTM when no resume checkpoint exists |\n| train | `results_dir` | `output_dir` | current job results directory |\n| train | `train.pretrained_model_path` | `ptm_if_no_resume_model` | PTM when no resume checkpoint exists |\n| train | `train.resume_training_checkpoint_path` | `resume_model` | model file inferred from the current job results folder |\n\nFor `parent_model` or `parent_model_folder`, pass the upstream train\u002Fexport\u002FAutoML child job id as `parent_job_id`. The SDK lists the parent result folder, filters checkpoint artifacts, and returns the selected model file or folder. Do not add these mappings back to `config.json` and do not patch generated runner scripts to guess checkpoint paths.\n\n## Deployment\n\n- [tao-deploy-rtdetr](references\u002Ftao-deploy-rtdetr.md)\n","# RT-DETR\n\nRT-DETR (Real-Time DEtection TRansformer) for 2D object detection. Designed for real-time inference with competitive accuracy. Supports distillation and quantization for deployment optimization.\n\nSet model.pretrained_backbone_path for backbone weights or train.pretrained_model_path for full model.\n\nFor TAO Deploy TensorRT actions (`gen_trt_engine`, TensorRT `evaluate`, and TensorRT `inference`), read `references\u002Ftao-deploy-rtdetr.md` first. Deploy spec templates live in this skill's `references\u002F` folder with the `spec_template_deploy_*.yaml` prefix.\n\n## Dataclass Schemas\n\nGenerated TAO Core schemas are packaged in `schemas\u002F\u003Caction>.schema.json`, with `schemas\u002Fmanifest.json` listing available actions. Each generated schema also emits `references\u002Fspec_template_\u003Caction>.yaml` from the schema top-level `default` field. AutoML enablement is declared at the model layer in `references\u002Fskill_info.yaml` via `automl_enabled`. Runnable AutoML still requires `schemas\u002Ftrain.schema.json` and `references\u002Fspec_template_train.yaml` to exist and parse. Use the packaged train schema for `automl_default_parameters`, `automl_disabled_parameters`, defaults, min\u002Fmax bounds, enums, option weights, math conditions, dependencies, and popular parameters. Do not expect `~\u002Ftao-core` at runtime; maintainers regenerate schemas\u002Ftemplates before packaging the skill bank.\n\n## Train Action Policy\n\nThis model is AutoML-enabled at the model layer. Before handling any train-stage request, read `references\u002Fskill_info.yaml` and resolve the run override from either an explicit `automl_policy` value or the user's workflow request. Use `automl_policy: on` by default and only expose `on` \u002F `off` in new launch prompts. Treat phrases like \"turn off AutoML\", \"disable AutoML\", \"no HPO\", or \"plain training\" as `automl_policy: off` for this run only. When `automl_policy: on`, `automl_enabled: true`, and both `schemas\u002Ftrain.schema.json` and `references\u002Fspec_template_train.yaml` are packaged, route the train action through `tao-skill-bank:tao-run-automl` by default with this model's `skill_dir`. Preserve workflow\u002Fapplication overrides for datasets, specs, output directories, GPU\u002Fplatform settings, parent checkpoints, and `automl_policy`. Use direct model training only when `automl_policy: off` or the packaged train schema\u002Ftemplate is missing; in the missing-schema case, report that AutoML is enabled but not runnable for this model until schemas are generated.\n\nNon-train actions such as `evaluate`, `inference`, `export`, and deploy flows stay in this model skill. The per-run `automl_policy` override does not change model metadata.\n\n## Supported Actions\n\nThe packaged RT-DETR PyT CLI supports `train`, `distill`, `quantize`, `evaluate`, `export`, `inference`, and `default_specs`. This model skill exposes `train`, `distill`, `quantize`, `evaluate`, `export`, and `inference`; resume\u002Fretrain is performed through `train` with `train.resume_training_checkpoint_path`.\n\nThe parent PyT CLI does not expose `gen_trt_engine`. Use `models\u002Frtdetr\u002Fdeploy` for TensorRT engine generation, TensorRT evaluation, and TensorRT inference.\n\n## Training Requirements\n\n- **Dataset type:** object_detection\n- **Formats:** coco, coco_raw\n- **Monitoring metric:** mAP50 for quick operational checks; `val_mAP` for COCO\u002Fpaper-style benchmark comparisons.\n\n### Per-Action Dataset Requirements\n\n| Action | Spec Key | Source | Files | List? |\n|---|---|---|---|---|\n| distill | dataset.train_data_sources | train_datasets | image_dir: images.tar.gz, json_file: annotations.json | Yes |\n| distill | dataset.val_data_sources | eval_dataset | image_dir: images.tar.gz, json_file: annotations.json | No |\n| evaluate | dataset.test_data_sources | eval_dataset | image_dir: images.tar.gz, json_file: annotations.json | No |\n| inference | dataset.infer_data_sources | inference_dataset | image_dir: images.tar.gz, classmap: label_map.txt | Yes |\n| quantize | dataset.train_data_sources | train_datasets | image_dir: images.tar.gz, json_file: annotations.json | Yes |\n| quantize | dataset.val_data_sources | eval_dataset | image_dir: images.tar.gz, json_file: annotations.json | No |\n| quantize | dataset.quant_calibration_data_sources | train_datasets | image_dir: images.tar.gz, json_file: annotations.json | No |\n| train | dataset.train_data_sources | train_datasets | image_dir: images.tar.gz, json_file: annotations.json | Yes |\n| train | dataset.val_data_sources | eval_dataset | image_dir: images.tar.gz, json_file: annotations.json | No |\n\n### Typical Spec Overrides\n\nData source overrides are **mandatory for every action** — the agent MUST construct data source paths from the Per-Action Dataset Requirements table above and include them in `spec_overrides`.\n\n```python\nS3_TRAIN = \"s3:\u002F\u002Fbucket\u002Fdata\u002Ftrain\"\nS3_EVAL = \"s3:\u002F\u002Fbucket\u002Fdata\u002Feval\"\nCHECKPOINT = \"\u002Fresults\u002F{train_job_id}\u002Fresults_dir\u002Fmodel_epoch_000.pth\"\nONNX_FILE = \"\u002Fresults\u002F{export_job_id}\u002Fresults_dir\u002Frtdetr.onnx\"\n```\n\n**train (mandatory data sources):**\n```python\n{\n    \"train.num_epochs\": 10,\n    \"train.checkpoint_interval\": 10,\n    \"train.validation_interval\": 10,\n    \"train.num_gpus\": 1,\n    \"train.gpu_ids\": [0],\n    \"dataset.num_classes\": \"\u003Cnum_classes> + 1\",\n    \"dataset.train_data_sources\": [{\"image_dir\": f\"{S3_TRAIN}\u002Fimages.tar.gz\", \"json_file\": f\"{S3_TRAIN}\u002Fannotations.json\"}],\n    \"dataset.val_data_sources\": {\"image_dir\": f\"{S3_EVAL}\u002Fimages.tar.gz\", \"json_file\": f\"{S3_EVAL}\u002Fannotations.json\"},\n}\n```\n\n**resume train (mandatory checkpoint):**\n```python\n{\n    \"train.num_epochs\": 11,\n    \"train.resume_training_checkpoint_path\": CHECKPOINT,\n    \"dataset.num_classes\": \"\u003Cnum_classes> + 1\",\n    \"dataset.eval_class_ids\": [1, 2, 3, 4],\n    \"dataset.train_data_sources\": [{\"image_dir\": f\"{S3_TRAIN}\u002Fimages.tar.gz\", \"json_file\": f\"{S3_TRAIN}\u002Fannotations.json\"}],\n    \"dataset.val_data_sources\": {\"image_dir\": f\"{S3_EVAL}\u002Fimages.tar.gz\", \"json_file\": f\"{S3_EVAL}\u002Fannotations.json\"},\n}\n```\n\n**evaluate (mandatory data sources and checkpoint):**\n```python\n{\n    \"dataset.num_classes\": \"\u003Cnum_classes> + 1\",\n    \"dataset.eval_class_ids\": [1, 2, 3, 4],\n    \"dataset.test_data_sources\": {\"image_dir\": f\"{S3_EVAL}\u002Fimages.tar.gz\", \"json_file\": f\"{S3_EVAL}\u002Fannotations.json\"},\n    \"evaluate.checkpoint\": CHECKPOINT,\n}\n```\n\n**export (mandatory checkpoint and output):**\n```python\n{\n    \"dataset.num_classes\": \"\u003Cnum_classes> + 1\",\n    \"export.checkpoint\": CHECKPOINT,\n    \"export.onnx_file\": ONNX_FILE,\n    \"export.input_height\": 640,\n    \"export.input_width\": 640,\n}\n```\n\n**quantize (mandatory data sources):**\n```python\n{\n    \"dataset.num_classes\": \"\u003Cnum_classes> + 1\",\n    \"quantize.layers\": [\n        {\n            \"module_name\": \"*\",\n            \"weights\": {\n                \"dtype\": \"float8_e4m3fn\"\n            },\n            \"activations\": {\n                \"dtype\": \"float8_e4m3fn\"\n            }\n        }\n    ],\n    \"dataset.train_data_sources\": [{\"image_dir\": f\"{S3_TRAIN}\u002Fimages.tar.gz\", \"json_file\": f\"{S3_TRAIN}\u002Fannotations.json\"}],\n    \"dataset.val_data_sources\": {\"image_dir\": f\"{S3_EVAL}\u002Fimages.tar.gz\", \"json_file\": f\"{S3_EVAL}\u002Fannotations.json\"},\n    \"dataset.quant_calibration_data_sources\": {\"image_dir\": f\"{S3_TRAIN}\u002Fimages.tar.gz\", \"json_file\": f\"{S3_TRAIN}\u002Fannotations.json\"},\n    \"quantize.model_path\": CHECKPOINT,\n}\n```\n\n**inference (mandatory data sources and checkpoint):**\n```python\n{\n    \"dataset.num_classes\": \"\u003Cnum_classes> + 1\",\n    \"dataset.infer_data_sources\": {\"image_dir\": [f\"{S3_EVAL}\u002Fimages.tar.gz\"], \"classmap\": f\"{S3_EVAL}\u002Flabel_map.txt\"},\n    \"inference.checkpoint\": CHECKPOINT,\n}\n```\n\n**distill (mandatory data sources and teacher checkpoint):**\n```python\n{\n    \"dataset.train_data_sources\": [{\"image_dir\": f\"{S3_TRAIN}\u002Fimages.tar.gz\", \"json_file\": f\"{S3_TRAIN}\u002Fannotations.json\"}],\n    \"dataset.val_data_sources\": {\"image_dir\": f\"{S3_EVAL}\u002Fimages.tar.gz\", \"json_file\": f\"{S3_EVAL}\u002Fannotations.json\"},\n    \"distill.pretrained_teacher_model_path\": CHECKPOINT,\n}\n```\n## Eval Dataset\n\nOptional. Provides validation mAP at each checkpoint if supplied.\n\n## Important Parameters\n\n- **dataset.num_classes**: Number of classes. Default 80 (MSCOCO 80-class). Must match your dataset annotations.\n- **model.backbone**: Default resnet_50. Supported: ResNet variants, ConvNeXt, FAN, EfficientViT. RT-DETR is optimized for real-time with lighter backbones.\n- **train.optim.lr**: Learning rate. Default 1e-4 (lower than DINO's 2e-4). lr_backbone defaults to 1e-5.\n- **dataset.augmentation.train_spatial_size**: Training input size. Default [640, 640]. Smaller than DINO's multi-scale (up to 1333). Key to RT-DETR's speed.\n- **model.num_feature_levels**: Default 3 (vs DINO's 4). return_interm_indices is [1,2,3].\n- **train.enable_ema**: Exponential moving average. Default False. Enable for potentially smoother convergence.\n- **dataset.remap_mscoco_category**: Default False. Set True only for original MSCOCO dataset with 91-to-80 category ID remapping.\n\n## Multi-GPU \u002F Multi-Node\n\n**Launch method:** `torchrun` (LIGHTNING_EXCLUDED_NETWORK). The entrypoint runs `torchrun --nnodes=N --nproc-per-node=M train.py`, NOT plain `python`.\n\n| Spec Key | Description | Default |\n|----------|-------------|---------|\n| `train.num_gpus` | Number of GPUs per node | 1 |\n| `train.gpu_ids` | GPU device indices | [0] |\n| `train.num_nodes` | Number of nodes | 1 |\n| `train.distributed_strategy` | `ddp` or `fsdp` | `ddp` |\n\n- When increasing `train.num_gpus`, also set `train.gpu_ids` to the same\n  visible device range. For example, an 8-GPU single-node Slurm run must\n  include both `\"train.num_gpus\": 8` and\n  `\"train.gpu_ids\": [0, 1, 2, 3, 4, 5, 6, 7]`.\n- `CUDA_VISIBLE_DEVICES` is explicitly set (unlike Lightning-managed models which use `TAO_VISIBLE_DEVICES`)\n- `ddp` with activation checkpointing: `find_unused_parameters=False`\n- `ddp` without: `find_unused_parameters=True`\n- `fsdp` supported, forces FP16\n\n**Multi-node env vars** (set by orchestrator):\n\n| Variable | Purpose |\n|----------|---------|\n| `WORLD_SIZE` | Number of nodes (triggers multinode mode) |\n| `NODE_RANK` | This node's rank (0-indexed) |\n| `MASTER_ADDR` | Rank-0 node IP |\n| `MASTER_PORT` | Rank-0 port (default 29500) |\n| `NUM_GPU_PER_NODE` | GPUs per node (default: all visible) |\n\n**CRITICAL:** `NODE_RANK` is copied to `RANK` if `RANK` is unset. This is required for torchrun multinode.\n\n## Export \u002F TRT Defaults\n\n- Export input: 640x640, opset 17\n- TRT data types: FP32, FP16, INT8\n- TRT workspace: 1024 MB\n- TRT max_batch_size: 4\n\n## Distillation\n\nRT-DETR supports knowledge distillation with a teacher model. Requires `distill` action with `distill.pretrained_teacher_model_path` and a distillation binding configuration.\n\nUse the packaged `references\u002Fspec_template_distill.yaml` as the starting point. The validated default binding uses the RT-DETR distiller's explicit IOU feature path:\n\n```yaml\ndistill:\n  bindings:\n  - student_module_name: srcs\n    teacher_module_name: srcs\n    criterion: IOU\n    weight: 1.0\n```\n\nDo not substitute DINO-style output names such as `pred_logits` \u002F `pred_boxes`, and do not bind arbitrary decoder heads unless you have verified the module returns captured feature lists. The RT-DETR distiller asserts that IOU bindings must use `srcs` or `dsrcs`.\n\n## Hardware\n\nMinimum 1 GPU(s), recommended 2 GPU(s). 16GB+ (V100 or A100) VRAM per GPU. RT-DETR is more memory-efficient than DINO\u002FGDINO due to smaller input size (640x640) and fewer feature levels. Trains well on single GPU for small-medium datasets.\n\n## Error Patterns\n\n**CUDA out of memory**: Reduce batch_size. RT-DETR at 640x640 is lighter than DINO at 1333px, but batch_size > 8 may still OOM on 16GB GPUs.\n\n**num_classes mismatch**: RT-DETR defaults to 80 (not 91 like DINO). Ensure dataset.num_classes matches your annotation categories.\n\n**CUDA index assert from category IDs**: If COCO category IDs are one-based or otherwise not remapped to zero-based contiguous IDs, set `dataset.num_classes` to `max(category_id) + 1` and keep `dataset.eval_class_ids` aligned to the actual category IDs. For the packaged four-class S3 sample with IDs 1-4, use `dataset.num_classes: 5` and `dataset.eval_class_ids: [1, 2, 3, 4]`.\n\n**return_interm_indices vs num_feature_levels**: Default is [1,2,3] with num_feature_levels=3. Must be consistent if changed.\n\n**Export shape mismatch**: Keep RT-DETR export and deploy consumer input size at\nthe validated `640x640` default unless the model has been trained and checked\nfor a different shape. The older packaged `960x544` template shape can fail\nduring ONNX tracing with `The size of tensor a (...) must match the size of\ntensor b (...)` in `hybrid_encoder.py` positional embedding addition.\n\n**AutoML metric extraction**: RT-DETR emits detection metrics in structured training status and logs. For COCO\u002Fpaper-style benchmark comparisons, optimize `val_mAP` with `direction: maximize`; for explicit AP50 workflows, optimize `mAP50`. Prefer `results_dir\u002Ftrain\u002Fstatus.json` or AutoML result state before parsing raw logs. Do not optimize `val_loss` for default detection model invocations.\n\n**Checkpoint handoff**: For evaluate\u002Fexport\u002Finference\u002Fquantize\u002Fdistill\u002Fresume, use the checkpoint resolver on the best AutoML child job's `results_dir\u002Ftrain\u002F` folder and select the action-appropriate `model_epoch_*.pth` checkpoint. RT-DETR may also write a latest symlink, but that should only be used when a caller explicitly requests latest. Keep `dataset.num_classes`, `dataset.eval_class_ids`, `model.num_queries`, and `model.num_select` consistent with training.\n\n**Parent `rtdetr gen_trt_engine` rejected by the PyT CLI**: In the validated 7.0.0 PyT container, `rtdetr gen_trt_engine` is not a valid parent-model subtask. Use the RT-DETR deploy workflow (`references\u002Ftao-deploy-rtdetr.md`) for TensorRT engine generation, TensorRT evaluation, and TensorRT inference.\n\n## Spec Param \u002F Parent Model Inference\n\nModel-specific inference mappings belong in this MD file, not in `config.json`. Generated runners should read this section and apply the mappings with SDK helpers before `create_job()`. This mirrors the old microservices `infer_params.py` flow.\n\nInference mappings from TAO Core `rtdetr.config.json`:\n\n| Action | Spec Field | Inference Function | Meaning |\n|---|---|---|---|\n| distill | `distill.pretrained_teacher_model_path` | `parent_model` | model file inferred from the parent job results folder |\n| distill | `encryption_key` | `key` | encryption key |\n| distill | `results_dir` | `output_dir` | current job results directory |\n| evaluate | `encryption_key` | `key` | encryption key |\n| evaluate | `evaluate.checkpoint` | `parent_model` | model file inferred from the parent job results folder |\n| evaluate | `evaluate.trt_engine` | `parent_model` | model file inferred from the parent job results folder |\n| evaluate | `results_dir` | `output_dir` | current job results directory |\n| export | `encryption_key` | `key` | encryption key |\n| export | `export.checkpoint` | `parent_model` | model file inferred from the parent job results folder |\n| export | `export.onnx_file` | `create_onnx_file` | output ONNX path |\n| export | `results_dir` | `output_dir` | current job results directory |\n| inference | `encryption_key` | `key` | encryption key |\n| inference | `inference.checkpoint` | `parent_model` | model file inferred from the parent job results folder |\n| inference | `inference.trt_engine` | `parent_model` | model file inferred from the parent job results folder |\n| inference | `results_dir` | `output_dir` | current job results directory |\n| quantize | `encryption_key` | `key` | encryption key |\n| quantize | `quantize.model_path` | `parent_model` | model file inferred from the parent job results folder |\n| quantize | `results_dir` | `output_dir` | current job results directory |\n| train | `encryption_key` | `key` | encryption key |\n| train | `model.pretrained_backbone_path` | `ptm_if_no_resume_model` | PTM when no resume checkpoint exists |\n| train | `results_dir` | `output_dir` | current job results directory |\n| train | `train.pretrained_model_path` | `ptm_if_no_resume_model` | PTM when no resume checkpoint exists |\n| train | `train.resume_training_checkpoint_path` | `resume_model` | model file inferred from the current job results folder |\n\nFor `parent_model` or `parent_model_folder`, pass the upstream train\u002Fexport\u002FAutoML child job id as `parent_job_id`. The SDK lists the parent result folder, filters checkpoint artifacts, and returns the selected model file or folder. Do not add these mappings back to `config.json` and do not patch generated runner scripts to guess checkpoint paths.\n\n## Deployment\n\n- [tao-deploy-rtdetr](references\u002Ftao-deploy-rtdetr.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","be3de0244dc1a57e323d8160787ebb40fb64930e","public\u002Fapi\u002Fskills\u002Fnvidia\u002Ftao-train-rtdetr","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":42,"prompt":52,"zipAvailable":53},[43,46,49],{"tool":44,"command":45},"npx","npx skills add Solizardking\u002Fskillhub-main --skill tao-train-rtdetr",{"tool":47,"command":48},"bunx","bunx skills add Solizardking\u002Fskillhub-main --skill tao-train-rtdetr",{"tool":50,"command":51},"pnpm","pnpm dlx skills add Solizardking\u002Fskillhub-main --skill tao-train-rtdetr","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002FSolizardking\u002Fskillhub-main\u002Fblob\u002Fmain\u002Fpublic\u002Fapi\u002Fskills\u002Fnvidia\u002Ftao-train-rtdetr\u002FSKILL.md",true,[55,61,65,69,73,77,81,85,89,93,97,101,104,108,112,116],{"path":56,"resourceType":57,"resourceTypeDesc":58,"sizeBytes":59,"sourceHtmlUrl":60},"BENCHMARK.md","REFERENCE","参考文档",4050,"https:\u002F\u002Fgithub.com\u002FSolizardking\u002Fskillhub-main\u002Fblob\u002Fmain\u002Fpublic\u002Fapi\u002Fskills\u002Fnvidia\u002Ftao-train-rtdetr\u002FBENCHMARK.md",{"path":62,"resourceType":57,"resourceTypeDesc":58,"sizeBytes":63,"sourceHtmlUrl":64},"metadata.json",645,"https:\u002F\u002Fgithub.com\u002FSolizardking\u002Fskillhub-main\u002Fblob\u002Fmain\u002Fpublic\u002Fapi\u002Fskills\u002Fnvidia\u002Ftao-train-rtdetr\u002Fmetadata.json",{"path":66,"resourceType":57,"resourceTypeDesc":58,"sizeBytes":67,"sourceHtmlUrl":68},"references\u002Fskill_info.yaml",7187,"https:\u002F\u002Fgithub.com\u002FSolizardking\u002Fskillhub-main\u002Fblob\u002Fmain\u002Fpublic\u002Fapi\u002Fskills\u002Fnvidia\u002Ftao-train-rtdetr\u002Freferences\u002Fskill_info.yaml",{"path":70,"resourceType":57,"resourceTypeDesc":58,"sizeBytes":71,"sourceHtmlUrl":72},"references\u002Fspec_template_deploy_evaluate.yaml",508,"https:\u002F\u002Fgithub.com\u002FSolizardking\u002Fskillhub-main\u002Fblob\u002Fmain\u002Fpublic\u002Fapi\u002Fskills\u002Fnvidia\u002Ftao-train-rtdetr\u002Freferences\u002Fspec_template_deploy_evaluate.yaml",{"path":74,"resourceType":57,"resourceTypeDesc":58,"sizeBytes":75,"sourceHtmlUrl":76},"references\u002Fspec_template_deploy_gen_trt_engine.yaml",665,"https:\u002F\u002Fgithub.com\u002FSolizardking\u002Fskillhub-main\u002Fblob\u002Fmain\u002Fpublic\u002Fapi\u002Fskills\u002Fnvidia\u002Ftao-train-rtdetr\u002Freferences\u002Fspec_template_deploy_gen_trt_engine.yaml",{"path":78,"resourceType":57,"resourceTypeDesc":58,"sizeBytes":79,"sourceHtmlUrl":80},"references\u002Fspec_template_deploy_inference.yaml",562,"https:\u002F\u002Fgithub.com\u002FSolizardking\u002Fskillhub-main\u002Fblob\u002Fmain\u002Fpublic\u002Fapi\u002Fskills\u002Fnvidia\u002Ftao-train-rtdetr\u002Freferences\u002Fspec_template_deploy_inference.yaml",{"path":82,"resourceType":57,"resourceTypeDesc":58,"sizeBytes":83,"sourceHtmlUrl":84},"references\u002Fspec_template_distill.yaml",4212,"https:\u002F\u002Fgithub.com\u002FSolizardking\u002Fskillhub-main\u002Fblob\u002Fmain\u002Fpublic\u002Fapi\u002Fskills\u002Fnvidia\u002Ftao-train-rtdetr\u002Freferences\u002Fspec_template_distill.yaml",{"path":86,"resourceType":57,"resourceTypeDesc":58,"sizeBytes":87,"sourceHtmlUrl":88},"references\u002Fspec_template_evaluate.yaml",3136,"https:\u002F\u002Fgithub.com\u002FSolizardking\u002Fskillhub-main\u002Fblob\u002Fmain\u002Fpublic\u002Fapi\u002Fskills\u002Fnvidia\u002Ftao-train-rtdetr\u002Freferences\u002Fspec_template_evaluate.yaml",{"path":90,"resourceType":57,"resourceTypeDesc":58,"sizeBytes":91,"sourceHtmlUrl":92},"references\u002Fspec_template_export.yaml",3233,"https:\u002F\u002Fgithub.com\u002FSolizardking\u002Fskillhub-main\u002Fblob\u002Fmain\u002Fpublic\u002Fapi\u002Fskills\u002Fnvidia\u002Ftao-train-rtdetr\u002Freferences\u002Fspec_template_export.yaml",{"path":94,"resourceType":57,"resourceTypeDesc":58,"sizeBytes":95,"sourceHtmlUrl":96},"references\u002Fspec_template_inference.yaml",3216,"https:\u002F\u002Fgithub.com\u002FSolizardking\u002Fskillhub-main\u002Fblob\u002Fmain\u002Fpublic\u002Fapi\u002Fskills\u002Fnvidia\u002Ftao-train-rtdetr\u002Freferences\u002Fspec_template_inference.yaml",{"path":98,"resourceType":57,"resourceTypeDesc":58,"sizeBytes":99,"sourceHtmlUrl":100},"references\u002Fspec_template_quantize.yaml",2966,"https:\u002F\u002Fgithub.com\u002FSolizardking\u002Fskillhub-main\u002Fblob\u002Fmain\u002Fpublic\u002Fapi\u002Fskills\u002Fnvidia\u002Ftao-train-rtdetr\u002Freferences\u002Fspec_template_quantize.yaml",{"path":102,"resourceType":57,"resourceTypeDesc":58,"sizeBytes":99,"sourceHtmlUrl":103},"references\u002Fspec_template_train.yaml","https:\u002F\u002Fgithub.com\u002FSolizardking\u002Fskillhub-main\u002Fblob\u002Fmain\u002Fpublic\u002Fapi\u002Fskills\u002Fnvidia\u002Ftao-train-rtdetr\u002Freferences\u002Fspec_template_train.yaml",{"path":105,"resourceType":57,"resourceTypeDesc":58,"sizeBytes":106,"sourceHtmlUrl":107},"references\u002Ftao-deploy-rtdetr.md",6081,"https:\u002F\u002Fgithub.com\u002FSolizardking\u002Fskillhub-main\u002Fblob\u002Fmain\u002Fpublic\u002Fapi\u002Fskills\u002Fnvidia\u002Ftao-train-rtdetr\u002Freferences\u002Ftao-deploy-rtdetr.md",{"path":109,"resourceType":57,"resourceTypeDesc":58,"sizeBytes":110,"sourceHtmlUrl":111},"references\u002Ftao-deploy-rtdetr.skill_info.yaml",3322,"https:\u002F\u002Fgithub.com\u002FSolizardking\u002Fskillhub-main\u002Fblob\u002Fmain\u002Fpublic\u002Fapi\u002Fskills\u002Fnvidia\u002Ftao-train-rtdetr\u002Freferences\u002Ftao-deploy-rtdetr.skill_info.yaml",{"path":113,"resourceType":57,"resourceTypeDesc":58,"sizeBytes":114,"sourceHtmlUrl":115},"skill-card.md",4213,"https:\u002F\u002Fgithub.com\u002FSolizardking\u002Fskillhub-main\u002Fblob\u002Fmain\u002Fpublic\u002Fapi\u002Fskills\u002Fnvidia\u002Ftao-train-rtdetr\u002Fskill-card.md",{"path":117,"resourceType":57,"resourceTypeDesc":58,"sizeBytes":118,"sourceHtmlUrl":119},"verification.json",3797,"https:\u002F\u002Fgithub.com\u002FSolizardking\u002Fskillhub-main\u002Fblob\u002Fmain\u002Fpublic\u002Fapi\u002Fskills\u002Fnvidia\u002Ftao-train-rtdetr\u002Fverification.json",[121,129,136,143,150,158],{"skillNo":122,"name":123,"nameZh":6,"description":124,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":125,"licenseFileUrl":6,"sourceHtmlUrl":126,"categories":127,"keywords":128,"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":130,"name":131,"nameZh":6,"description":132,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":125,"licenseFileUrl":6,"sourceHtmlUrl":133,"categories":134,"keywords":135,"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":137,"name":138,"nameZh":6,"description":139,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":125,"licenseFileUrl":6,"sourceHtmlUrl":140,"categories":141,"keywords":142,"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":144,"name":145,"nameZh":6,"description":146,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":125,"licenseFileUrl":6,"sourceHtmlUrl":147,"categories":148,"keywords":149,"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":151,"name":152,"nameZh":6,"description":153,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":125,"licenseFileUrl":154,"sourceHtmlUrl":155,"categories":156,"keywords":157,"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":159,"name":160,"nameZh":6,"description":161,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":125,"licenseFileUrl":6,"sourceHtmlUrl":162,"categories":163,"keywords":164,"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",[],[]]