[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M71Q2D7EHXG4SA5F5QZP4493":3,"skill-same:01M71Q2D7EHXG4SA5F5QZP4493":224},{"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":23,"publishedAt":44,"linkOnly":45,"markdownContent":46,"markdownBody":47,"frontMatter":48,"refName":59,"commitSha":60,"skillDirectory":61,"forksCount":14,"repositoryHtmlUrl":62,"repositoryHomepageUrl":63,"ownerHtmlUrl":64,"ownerBlogUrl":6,"ownerDescription":6,"updatedAt":65,"install":66,"resources":79},"01M71Q2D7EHXG4SA5F5QZP4493","tao-train-visual-changenet",null,"Visual ChangeNet for binary image classification and segmentation in AOI defect detection. Use when training, evaluating, exporting, or running inference for PCB defect detection or visual inspection, comparing image pairs for PASS\u002FNO_PASS classification, or producing change-segmentation masks. Trigger phrases include \"train Visual ChangeNet\", \"ChangeNet classify\", \"ChangeNet segment\", \"AOI defect detection\", \"PCB inspection model\".","pattanation\u002FNorvik","pattanation","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F235419268?v=4","01M71M5THHYYB15ME8R2E4FCZG","Patta","01M71M5TJM1NDM992M8PDKYR2K",0,"Apache-2.0","https:\u002F\u002Fgithub.com\u002Fpattanation\u002FNorvik\u002Fblob\u002Fclaude\u002Finstall-gsd-build-foQdV\u002F.agents\u002Fskills\u002Ftao-train-visual-changenet\u002FSKILL.md",[18],{"categoryNo":19,"categoryName":20,"taxonomyType":21,"taxonomyTypeDesc":22},"01JBSEEDCAT000000000000011","运维与部署","SCENARIO","使用场景",[24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43],"tao","visual","changenet","train","tao train visual changenet","NVIDIA","segment","running","pcb","read","defect","classify","change","Bash","image","when","inspection","training","model","segmentation","2026-11-10 13:02:02",false,"---\nname: tao-train-visual-changenet\ndescription: Visual ChangeNet for binary image classification and segmentation in AOI defect detection. Use when training,\n  evaluating, exporting, or running inference for PCB defect detection or visual inspection, comparing image pairs for\n  PASS\u002FNO_PASS classification, or producing change-segmentation masks. Trigger phrases include \"train Visual ChangeNet\",\n  \"ChangeNet classify\", \"ChangeNet segment\", \"AOI defect detection\", \"PCB inspection model\".\nlicense: Apache-2.0\ncompatibility: Requires docker + nvidia-container-toolkit.\nmetadata:\n  author: NVIDIA Corporation\n  version: \"0.1.0\"\nallowed-tools: Read Bash\ntags:\n- pcb\n- aoi\n- defect\n- classification\n- segmentation\n- siamese\n- visual-inspection\n---\n\n# Visual ChangeNet\n\nVisual ChangeNet is a TAO Toolkit model for visual inspection and defect detection. It supports two tasks:\n\n- **Classify** — Binary image classification using a siamese-style architecture with a shared backbone (C-RADIO ViT) and a learnable difference module. Compares image pairs to classify defects as PASS\u002FNO_PASS.\n- **Segment** — Pixel-level change segmentation using a ViT-Large NVDINOv2 backbone. Compares before\u002Fafter image pairs to produce a binary change mask.\n\nThe backbone weight (`c_radio_v2_vit_base_patch16_224`) is the `nvidia\u002FC-RADIOv2-B` model from HuggingFace, distributed as `model.safetensors` (~393 MB). **The TAO 7.0.0-rc container does not auto-fetch from HF URLs** — `ptm_utils.load_pretrained_weights()` hands the `pretrained_backbone_path` value to `torch.load(path)` \u002F `safetensors.torch.load_file(path)` directly. Passing an `https:\u002F\u002Fhuggingface.co\u002F...` URL or a repo id produces `FileNotFoundError` and the run fails with `Execution status: FAIL` within a few seconds. Stage the file locally before launch:\n\n```bash\npython3 -c \"from huggingface_hub import hf_hub_download; import shutil; \\\nshutil.copy(hf_hub_download('nvidia\u002FC-RADIOv2-B', 'model.safetensors'), '\u003Cworkspace>\u002Fbackbone\u002Fc_radio_v2_b.safetensors')\"\n```\n\nMount it into the container (`-v \u003Cworkspace>\u002Fbackbone\u002Fc_radio_v2_b.safetensors:\u002Fdata\u002Fpretrained_models\u002FC-RADIOv2_B.safetensors`) and set the spec `model.backbone.pretrained_backbone_path` to the container path. `HF_TOKEN` is only needed at staging time, not at training time.\n\nSegment specs use `model.backbone.type: vit_large_nvdinov2` and the NVDINOv2\ncheckpoint family. Keep the checkpoint architecture aligned with the backbone\ntype: `NV_DINOV2_518_16_256.ckpt` is compatible with the packaged segment\ntemplates, but it must not be used with `fan_small_12_p4_hybrid`. If you switch\nto a different segment backbone, use a matching checkpoint or leave\n`model.backbone.pretrained_backbone_path` empty for default initialization.\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 declared by this model skill (`evaluate`, `inference`,\n`export`, `quantize`, `segment_evaluate`, and `segment_inference`) stay in this\nmodel skill. Do not present `segment_export` or `segment_quantize` as runnable\nparent-skill actions until matching entries are packaged in\n`schemas\u002Fmanifest.json`. Prune and retrain are not declared in the current\nparent `references\u002Fskill_info.yaml`; do not present them as runnable parent-skill\nactions unless the metadata is extended with matching action wiring and schemas.\nThe per-run `automl_policy` override does not change model metadata.\n\nFor TAO Deploy TensorRT actions (`gen_trt_engine`, TensorRT `evaluate`, and TensorRT `inference` for classify and segment variants), read `references\u002Ftao-deploy-visual-changenet.md` first. Deploy spec templates live in this skill's `references\u002F` folder with the `spec_template_deploy_*.yaml` prefix.\nDeploy requires an exported ONNX artifact as `parent_model`. If no ONNX artifact exists and the main skill does not expose an export action, report deploy as blocked instead of inventing an artifact.\n\n## Training Requirements\n\nVisual ChangeNet has two separate task modes with different dataset types and data source structures.\n\n### Classify\n\n- **Dataset type:** visual_changenet_classify\n- **Formats:** default\n- **Accepted dataset intents:** training, evaluation, testing, calibration\n- **Monitoring metric:** val_loss\n\n#### Per-Action Dataset Requirements (Classify)\n\nThe `quantize` and `gen_trt_engine` rows below describe TAO spec data requirements only. They are not parent-skill actions unless the corresponding action is declared in `references\u002Fskill_info.yaml` or `deploy\u002Fskill_info.yaml`.\n\n| Action | Spec Key | Source | Files | List? |\n|---|---|---|---|---|\n| train | dataset.classify.train_dataset.images_dir | train_datasets | images.tar.gz | No |\n| train | dataset.classify.train_dataset.csv_path | train_datasets | dataset.csv | No |\n| train | dataset.classify.validation_dataset.images_dir | eval_dataset | images.tar.gz | No |\n| train | dataset.classify.validation_dataset.csv_path | eval_dataset | dataset.csv | No |\n| quantize | dataset.classify.train_dataset.images_dir | train_datasets | images.tar.gz | No |\n| quantize | dataset.classify.train_dataset.csv_path | train_datasets | dataset.csv | No |\n| quantize | dataset.classify.validation_dataset.images_dir | eval_dataset | images.tar.gz | No |\n| quantize | dataset.classify.validation_dataset.csv_path | eval_dataset | dataset.csv | No |\n| quantize | dataset.classify.quant_calibration_dataset.images_dir | train_datasets | images.tar.gz | No |\n| evaluate | dataset.classify.validation_dataset.images_dir | eval_dataset | images.tar.gz | No |\n| evaluate | dataset.classify.validation_dataset.csv_path | eval_dataset | dataset.csv | No |\n| evaluate | dataset.classify.test_dataset.images_dir | eval_dataset | images.tar.gz | No |\n| evaluate | dataset.classify.test_dataset.csv_path | eval_dataset | dataset.csv | No |\n| inference | dataset.classify.infer_dataset.images_dir | inference_dataset | images.tar.gz | No |\n| inference | dataset.classify.infer_dataset.csv_path | inference_dataset | dataset.csv | No |\n| gen_trt_engine | gen_trt_engine.tensorrt.calibration.cal_image_dir | calibration_dataset | images.tar.gz | Yes |\n\n### Segment\n\n- **Dataset type:** visual_changenet_segment\n- **Formats:** default\n- **Accepted dataset intents:** training, calibration\n- **Monitoring metric:** val_loss\n\nSegment uses a paired directory structure (`A\u002F`, `B\u002F`, `list\u002F`, `label\u002F`) instead of CSV + images. The `root_dir` spec key points to the top-level directory containing all four subdirectories.\n\n**Required files per dataset:** `A.tar.gz`, `B.tar.gz`, `list.tar.gz`, `label.tar.gz`\n\n#### Per-Action Dataset Requirements (Segment)\n\nThe `quantize` and `gen_trt_engine` rows below describe TAO spec data requirements only. They are not parent-skill actions unless the corresponding action is declared in `references\u002Fskill_info.yaml` or `deploy\u002Fskill_info.yaml`.\n\n| Action | Spec Key | Source | Files | List? |\n|---|---|---|---|---|\n| train | dataset.segment.root_dir | train_datasets | (root directory) | No |\n| quantize | dataset.segment.root_dir | train_datasets | (root directory) | No |\n| quantize | dataset.segment.quant_calibration_dataset.images_dir | train_datasets | (root directory) | No |\n| evaluate | dataset.segment.root_dir | train_datasets | (root directory) | No |\n| inference | dataset.segment.root_dir | train_datasets | (root directory) | No |\n| gen_trt_engine | dataset.segment.root_dir | train_datasets | (root directory) | No |\n| gen_trt_engine | gen_trt_engine.tensorrt.calibration.cal_image_dir | calibration_dataset | images.tar.gz | Yes |\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\"\n```\n\n**train (classify, mandatory data sources):**\n```python\n{\n    \"train.num_epochs\": 30,\n    \"train.checkpoint_interval\": 10,\n    \"train.validation_interval\": 10,\n    \"train.num_gpus\": 1,\n    \"train.use_distributed_sampler\": False,\n    \"train.sync_batchnorm\": False,\n    \"dataset.classify.train_dataset.images_dir\": f\"{S3_TRAIN}\u002Fimages.tar.gz\",\n    \"dataset.classify.train_dataset.csv_path\": f\"{S3_TRAIN}\u002Fdataset.csv\",\n    \"dataset.classify.validation_dataset.images_dir\": f\"{S3_EVAL}\u002Fimages.tar.gz\",\n    \"dataset.classify.validation_dataset.csv_path\": f\"{S3_EVAL}\u002Fdataset.csv\",\n}\n```\n\n**train (segment, mandatory data sources):**\n```python\n{\n    \"train.num_epochs\": 30,\n    \"train.checkpoint_interval\": 10,\n    \"train.validation_interval\": 10,\n    \"train.num_gpus\": 1,\n    \"train.use_distributed_sampler\": False,\n    \"train.sync_batchnorm\": False,\n    \"dataset.segment.root_dir\": f\"{S3_TRAIN}\",\n}\n```\n\n**export (classify):**\n```python\n{\n    \"export.input_height\": 896,\n    \"export.input_width\": 224,\n}\n```\n\n**export (segment):**\n```python\n{\n    \"export.input_height\": 224,\n    \"export.input_width\": 224,\n}\n```\n\n**quantize (classify, mandatory data sources):**\n```python\n{\n    \"dataset.classify.train_dataset.images_dir\": f\"{S3_TRAIN}\u002Fimages.tar.gz\",\n    \"dataset.classify.train_dataset.csv_path\": f\"{S3_TRAIN}\u002Fdataset.csv\",\n    \"dataset.classify.validation_dataset.images_dir\": f\"{S3_EVAL}\u002Fimages.tar.gz\",\n    \"dataset.classify.validation_dataset.csv_path\": f\"{S3_EVAL}\u002Fdataset.csv\",\n    \"dataset.classify.quant_calibration_dataset.images_dir\": f\"{S3_TRAIN}\u002Fimages.tar.gz\",\n}\n```\n\n**evaluate (classify, mandatory data sources):**\n```python\n{\n    \"dataset.classify.validation_dataset.images_dir\": f\"{S3_EVAL}\u002Fimages.tar.gz\",\n    \"dataset.classify.validation_dataset.csv_path\": f\"{S3_EVAL}\u002Fdataset.csv\",\n    \"dataset.classify.test_dataset.images_dir\": f\"{S3_EVAL}\u002Fimages.tar.gz\",\n    \"dataset.classify.test_dataset.csv_path\": f\"{S3_EVAL}\u002Fdataset.csv\",\n}\n```\n\n**inference (classify, mandatory data sources):**\n```python\n{\n    \"dataset.classify.infer_dataset.images_dir\": f\"{S3_EVAL}\u002Fimages.tar.gz\",\n    \"dataset.classify.infer_dataset.csv_path\": f\"{S3_EVAL}\u002Fdataset.csv\",\n}\n```\n\n**gen_trt_engine (classify, mandatory data sources):**\n```python\n{\n    \"gen_trt_engine.tensorrt.calibration.cal_image_dir\": [f\"{S3_TRAIN}\u002Fimages.tar.gz\"],\n}\n```\n\n**quantize (segment, mandatory data sources):**\n```python\n{\n    \"dataset.segment.root_dir\": f\"{S3_TRAIN}\",\n    \"dataset.segment.quant_calibration_dataset.images_dir\": f\"{S3_TRAIN}\",\n}\n```\n\n**evaluate (segment, mandatory data sources):**\n```python\n{\n    \"dataset.segment.root_dir\": f\"{S3_TRAIN}\",\n}\n```\n\n**inference (segment, mandatory data sources):**\n```python\n{\n    \"dataset.segment.root_dir\": f\"{S3_TRAIN}\",\n}\n```\n\n**gen_trt_engine (segment, mandatory data sources):**\n```python\n{\n    \"dataset.segment.root_dir\": f\"{S3_TRAIN}\",\n    \"gen_trt_engine.tensorrt.calibration.cal_image_dir\": [f\"{S3_TRAIN}\u002Fimages.tar.gz\"],\n}\n```\n## Optional: running via the TAO SDK\n\nWhen running without the TAO SDK (local docker), resolve the TAO pyt image from `versions.yaml` and invoke `visual_changenet \u003Ctrain|evaluate|inference|export|quantize>` directly. `--shm-size=8g` is required, the C-RADIO `.safetensors` must be mounted to `\u002Fdata\u002Fpretrained_models\u002FC-RADIOv2_B.safetensors`, and checkpoint\u002Fresults_dir can be overridden on the command line. See `references\u002Flocal-docker.md` for the full `docker run` command, mounts, and overrides.\n\n## Tasks\n\n### Classify (default)\n\nUses actions: `train`, `evaluate`, `inference`. Defaults template: `references\u002Fspec_template_train.yaml`.\n\n### Segment\n\nUses skill action names `segment_train`, `segment_evaluate`, and\n`segment_inference`. When invoking local Docker directly, run TAO CLI subcommands\n`train`, `evaluate`, and `inference` with `task: segment` in the spec. The\nschema-driven action templates are `references\u002Fspec_template_segment_train.yaml`,\n`references\u002Fspec_template_segment_evaluate.yaml`, and\n`references\u002Fspec_template_segment_inference.yaml`; the compact direct-Docker\nexample template is `references\u002Fspec_template_segment.yaml`.\n\nSegmentation requires compiling custom CUDA ops (`MultiScaleDeformableAttention`) on first run, which takes ~5 minutes. The ViT adapter backbone uses these for multi-scale feature extraction.\n\nDataset structure for segmentation differs from classify — uses paired directories (`A\u002F`, `B\u002F`, `list\u002F`, `label\u002F`) instead of CSV files. See `dataset.segment.root_dir` in the defaults.\n\n## Data Format\n\nClassify needs a 4-column CSV (`input_path,golden_path,label,object_name`) plus an images directory; segment uses a paired directory structure (`A\u002F`, `B\u002F`, `list\u002F`, `label\u002F`) under `dataset.segment.root_dir` instead of CSV. The `image_ext` field (default `.jpg`) must match the actual file extensions; if images are `.png`, set `dataset.classify.image_ext: .png`. Multi-lighting input is configured via `dataset.classify.input_map` (each lighting name maps to a channel index) with `dataset.classify.num_input` set to match. See `references\u002Fdata-formats.md` for the per-field input tables (classify train\u002Feval\u002Finference, segment), CSV column semantics, lighting\u002Fpath-concatenation conventions, the segment directory layout, and `input_map`\u002F`grid_map` examples.\n\n## Important Parameters\n\nKey knobs include `train.validation_interval` (default 50, must be ≤ num_epochs), `train.checkpoint_interval` (default 200, must be ≤ num_epochs), `train.num_epochs` (default 100), `model.classify.eval_margin` (default 0.3, the precision\u002Frecall threshold), `model.classify.train_margin_euclid` (default 2.0), `model.classify.embedding_vectors` (default 5), `dataset.classify.batch_size` (default 16, must be > 1), `dataset.classify.fpratio_sampling` (default 0.25), and `train.classify.cls_weight` (default [1.0, 10.0]). Hardware: minimum 1 GPU with 16GB+ VRAM, recommended 8 GPUs (DDP); do not set `gpu_spec_key` (GPU count is managed internally by TAO), `num_nodes` (default 1) controls multi-node. See `references\u002Ftuning-parameters.md` for the full per-parameter guidance and hardware detail.\n\n## Error Patterns\n\nFor checkpoint-not-found, CSV format mismatch, image extension mismatch, OOM, low evaluation accuracy, the contrastive-loss `AssertionError`, checkpoint load key mismatch at evaluate\u002Finference, non-convergence, segment-only backbone dimension mismatch, the `MultiScaleDeformableAttention` `OSError`, the Lightning `MisconfigurationException`, `ModuleNotFoundError: nvidia_tao_pytorch`, and epoch defaults, see `references\u002Ftroubleshooting.md` for the full symptom-and-fix list.\n\n## Spec Param \u002F Parent Model Inference\n\nModel-specific parent-model mappings are declared in `references\u002Fskill_info.yaml` under `spec_params`, so generated runners and agents resolve checkpoints before `create_job()` instead of guessing file names. For `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. See `references\u002Fparent-model-inference.md` for the full per-action spec-field-to-inference-function mapping table.\n\n## Deployment\n\n- [tao-deploy-visual-changenet](references\u002Ftao-deploy-visual-changenet.md)\n","# Visual ChangeNet\n\nVisual ChangeNet is a TAO Toolkit model for visual inspection and defect detection. It supports two tasks:\n\n- **Classify** — Binary image classification using a siamese-style architecture with a shared backbone (C-RADIO ViT) and a learnable difference module. Compares image pairs to classify defects as PASS\u002FNO_PASS.\n- **Segment** — Pixel-level change segmentation using a ViT-Large NVDINOv2 backbone. Compares before\u002Fafter image pairs to produce a binary change mask.\n\nThe backbone weight (`c_radio_v2_vit_base_patch16_224`) is the `nvidia\u002FC-RADIOv2-B` model from HuggingFace, distributed as `model.safetensors` (~393 MB). **The TAO 7.0.0-rc container does not auto-fetch from HF URLs** — `ptm_utils.load_pretrained_weights()` hands the `pretrained_backbone_path` value to `torch.load(path)` \u002F `safetensors.torch.load_file(path)` directly. Passing an `https:\u002F\u002Fhuggingface.co\u002F...` URL or a repo id produces `FileNotFoundError` and the run fails with `Execution status: FAIL` within a few seconds. Stage the file locally before launch:\n\n```bash\npython3 -c \"from huggingface_hub import hf_hub_download; import shutil; \\\nshutil.copy(hf_hub_download('nvidia\u002FC-RADIOv2-B', 'model.safetensors'), '\u003Cworkspace>\u002Fbackbone\u002Fc_radio_v2_b.safetensors')\"\n```\n\nMount it into the container (`-v \u003Cworkspace>\u002Fbackbone\u002Fc_radio_v2_b.safetensors:\u002Fdata\u002Fpretrained_models\u002FC-RADIOv2_B.safetensors`) and set the spec `model.backbone.pretrained_backbone_path` to the container path. `HF_TOKEN` is only needed at staging time, not at training time.\n\nSegment specs use `model.backbone.type: vit_large_nvdinov2` and the NVDINOv2\ncheckpoint family. Keep the checkpoint architecture aligned with the backbone\ntype: `NV_DINOV2_518_16_256.ckpt` is compatible with the packaged segment\ntemplates, but it must not be used with `fan_small_12_p4_hybrid`. If you switch\nto a different segment backbone, use a matching checkpoint or leave\n`model.backbone.pretrained_backbone_path` empty for default initialization.\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 declared by this model skill (`evaluate`, `inference`,\n`export`, `quantize`, `segment_evaluate`, and `segment_inference`) stay in this\nmodel skill. Do not present `segment_export` or `segment_quantize` as runnable\nparent-skill actions until matching entries are packaged in\n`schemas\u002Fmanifest.json`. Prune and retrain are not declared in the current\nparent `references\u002Fskill_info.yaml`; do not present them as runnable parent-skill\nactions unless the metadata is extended with matching action wiring and schemas.\nThe per-run `automl_policy` override does not change model metadata.\n\nFor TAO Deploy TensorRT actions (`gen_trt_engine`, TensorRT `evaluate`, and TensorRT `inference` for classify and segment variants), read `references\u002Ftao-deploy-visual-changenet.md` first. Deploy spec templates live in this skill's `references\u002F` folder with the `spec_template_deploy_*.yaml` prefix.\nDeploy requires an exported ONNX artifact as `parent_model`. If no ONNX artifact exists and the main skill does not expose an export action, report deploy as blocked instead of inventing an artifact.\n\n## Training Requirements\n\nVisual ChangeNet has two separate task modes with different dataset types and data source structures.\n\n### Classify\n\n- **Dataset type:** visual_changenet_classify\n- **Formats:** default\n- **Accepted dataset intents:** training, evaluation, testing, calibration\n- **Monitoring metric:** val_loss\n\n#### Per-Action Dataset Requirements (Classify)\n\nThe `quantize` and `gen_trt_engine` rows below describe TAO spec data requirements only. They are not parent-skill actions unless the corresponding action is declared in `references\u002Fskill_info.yaml` or `deploy\u002Fskill_info.yaml`.\n\n| Action | Spec Key | Source | Files | List? |\n|---|---|---|---|---|\n| train | dataset.classify.train_dataset.images_dir | train_datasets | images.tar.gz | No |\n| train | dataset.classify.train_dataset.csv_path | train_datasets | dataset.csv | No |\n| train | dataset.classify.validation_dataset.images_dir | eval_dataset | images.tar.gz | No |\n| train | dataset.classify.validation_dataset.csv_path | eval_dataset | dataset.csv | No |\n| quantize | dataset.classify.train_dataset.images_dir | train_datasets | images.tar.gz | No |\n| quantize | dataset.classify.train_dataset.csv_path | train_datasets | dataset.csv | No |\n| quantize | dataset.classify.validation_dataset.images_dir | eval_dataset | images.tar.gz | No |\n| quantize | dataset.classify.validation_dataset.csv_path | eval_dataset | dataset.csv | No |\n| quantize | dataset.classify.quant_calibration_dataset.images_dir | train_datasets | images.tar.gz | No |\n| evaluate | dataset.classify.validation_dataset.images_dir | eval_dataset | images.tar.gz | No |\n| evaluate | dataset.classify.validation_dataset.csv_path | eval_dataset | dataset.csv | No |\n| evaluate | dataset.classify.test_dataset.images_dir | eval_dataset | images.tar.gz | No |\n| evaluate | dataset.classify.test_dataset.csv_path | eval_dataset | dataset.csv | No |\n| inference | dataset.classify.infer_dataset.images_dir | inference_dataset | images.tar.gz | No |\n| inference | dataset.classify.infer_dataset.csv_path | inference_dataset | dataset.csv | No |\n| gen_trt_engine | gen_trt_engine.tensorrt.calibration.cal_image_dir | calibration_dataset | images.tar.gz | Yes |\n\n### Segment\n\n- **Dataset type:** visual_changenet_segment\n- **Formats:** default\n- **Accepted dataset intents:** training, calibration\n- **Monitoring metric:** val_loss\n\nSegment uses a paired directory structure (`A\u002F`, `B\u002F`, `list\u002F`, `label\u002F`) instead of CSV + images. The `root_dir` spec key points to the top-level directory containing all four subdirectories.\n\n**Required files per dataset:** `A.tar.gz`, `B.tar.gz`, `list.tar.gz`, `label.tar.gz`\n\n#### Per-Action Dataset Requirements (Segment)\n\nThe `quantize` and `gen_trt_engine` rows below describe TAO spec data requirements only. They are not parent-skill actions unless the corresponding action is declared in `references\u002Fskill_info.yaml` or `deploy\u002Fskill_info.yaml`.\n\n| Action | Spec Key | Source | Files | List? |\n|---|---|---|---|---|\n| train | dataset.segment.root_dir | train_datasets | (root directory) | No |\n| quantize | dataset.segment.root_dir | train_datasets | (root directory) | No |\n| quantize | dataset.segment.quant_calibration_dataset.images_dir | train_datasets | (root directory) | No |\n| evaluate | dataset.segment.root_dir | train_datasets | (root directory) | No |\n| inference | dataset.segment.root_dir | train_datasets | (root directory) | No |\n| gen_trt_engine | dataset.segment.root_dir | train_datasets | (root directory) | No |\n| gen_trt_engine | gen_trt_engine.tensorrt.calibration.cal_image_dir | calibration_dataset | images.tar.gz | Yes |\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\"\n```\n\n**train (classify, mandatory data sources):**\n```python\n{\n    \"train.num_epochs\": 30,\n    \"train.checkpoint_interval\": 10,\n    \"train.validation_interval\": 10,\n    \"train.num_gpus\": 1,\n    \"train.use_distributed_sampler\": False,\n    \"train.sync_batchnorm\": False,\n    \"dataset.classify.train_dataset.images_dir\": f\"{S3_TRAIN}\u002Fimages.tar.gz\",\n    \"dataset.classify.train_dataset.csv_path\": f\"{S3_TRAIN}\u002Fdataset.csv\",\n    \"dataset.classify.validation_dataset.images_dir\": f\"{S3_EVAL}\u002Fimages.tar.gz\",\n    \"dataset.classify.validation_dataset.csv_path\": f\"{S3_EVAL}\u002Fdataset.csv\",\n}\n```\n\n**train (segment, mandatory data sources):**\n```python\n{\n    \"train.num_epochs\": 30,\n    \"train.checkpoint_interval\": 10,\n    \"train.validation_interval\": 10,\n    \"train.num_gpus\": 1,\n    \"train.use_distributed_sampler\": False,\n    \"train.sync_batchnorm\": False,\n    \"dataset.segment.root_dir\": f\"{S3_TRAIN}\",\n}\n```\n\n**export (classify):**\n```python\n{\n    \"export.input_height\": 896,\n    \"export.input_width\": 224,\n}\n```\n\n**export (segment):**\n```python\n{\n    \"export.input_height\": 224,\n    \"export.input_width\": 224,\n}\n```\n\n**quantize (classify, mandatory data sources):**\n```python\n{\n    \"dataset.classify.train_dataset.images_dir\": f\"{S3_TRAIN}\u002Fimages.tar.gz\",\n    \"dataset.classify.train_dataset.csv_path\": f\"{S3_TRAIN}\u002Fdataset.csv\",\n    \"dataset.classify.validation_dataset.images_dir\": f\"{S3_EVAL}\u002Fimages.tar.gz\",\n    \"dataset.classify.validation_dataset.csv_path\": f\"{S3_EVAL}\u002Fdataset.csv\",\n    \"dataset.classify.quant_calibration_dataset.images_dir\": f\"{S3_TRAIN}\u002Fimages.tar.gz\",\n}\n```\n\n**evaluate (classify, mandatory data sources):**\n```python\n{\n    \"dataset.classify.validation_dataset.images_dir\": f\"{S3_EVAL}\u002Fimages.tar.gz\",\n    \"dataset.classify.validation_dataset.csv_path\": f\"{S3_EVAL}\u002Fdataset.csv\",\n    \"dataset.classify.test_dataset.images_dir\": f\"{S3_EVAL}\u002Fimages.tar.gz\",\n    \"dataset.classify.test_dataset.csv_path\": f\"{S3_EVAL}\u002Fdataset.csv\",\n}\n```\n\n**inference (classify, mandatory data sources):**\n```python\n{\n    \"dataset.classify.infer_dataset.images_dir\": f\"{S3_EVAL}\u002Fimages.tar.gz\",\n    \"dataset.classify.infer_dataset.csv_path\": f\"{S3_EVAL}\u002Fdataset.csv\",\n}\n```\n\n**gen_trt_engine (classify, mandatory data sources):**\n```python\n{\n    \"gen_trt_engine.tensorrt.calibration.cal_image_dir\": [f\"{S3_TRAIN}\u002Fimages.tar.gz\"],\n}\n```\n\n**quantize (segment, mandatory data sources):**\n```python\n{\n    \"dataset.segment.root_dir\": f\"{S3_TRAIN}\",\n    \"dataset.segment.quant_calibration_dataset.images_dir\": f\"{S3_TRAIN}\",\n}\n```\n\n**evaluate (segment, mandatory data sources):**\n```python\n{\n    \"dataset.segment.root_dir\": f\"{S3_TRAIN}\",\n}\n```\n\n**inference (segment, mandatory data sources):**\n```python\n{\n    \"dataset.segment.root_dir\": f\"{S3_TRAIN}\",\n}\n```\n\n**gen_trt_engine (segment, mandatory data sources):**\n```python\n{\n    \"dataset.segment.root_dir\": f\"{S3_TRAIN}\",\n    \"gen_trt_engine.tensorrt.calibration.cal_image_dir\": [f\"{S3_TRAIN}\u002Fimages.tar.gz\"],\n}\n```\n## Optional: running via the TAO SDK\n\nWhen running without the TAO SDK (local docker), resolve the TAO pyt image from `versions.yaml` and invoke `visual_changenet \u003Ctrain|evaluate|inference|export|quantize>` directly. `--shm-size=8g` is required, the C-RADIO `.safetensors` must be mounted to `\u002Fdata\u002Fpretrained_models\u002FC-RADIOv2_B.safetensors`, and checkpoint\u002Fresults_dir can be overridden on the command line. See `references\u002Flocal-docker.md` for the full `docker run` command, mounts, and overrides.\n\n## Tasks\n\n### Classify (default)\n\nUses actions: `train`, `evaluate`, `inference`. Defaults template: `references\u002Fspec_template_train.yaml`.\n\n### Segment\n\nUses skill action names `segment_train`, `segment_evaluate`, and\n`segment_inference`. When invoking local Docker directly, run TAO CLI subcommands\n`train`, `evaluate`, and `inference` with `task: segment` in the spec. The\nschema-driven action templates are `references\u002Fspec_template_segment_train.yaml`,\n`references\u002Fspec_template_segment_evaluate.yaml`, and\n`references\u002Fspec_template_segment_inference.yaml`; the compact direct-Docker\nexample template is `references\u002Fspec_template_segment.yaml`.\n\nSegmentation requires compiling custom CUDA ops (`MultiScaleDeformableAttention`) on first run, which takes ~5 minutes. The ViT adapter backbone uses these for multi-scale feature extraction.\n\nDataset structure for segmentation differs from classify — uses paired directories (`A\u002F`, `B\u002F`, `list\u002F`, `label\u002F`) instead of CSV files. See `dataset.segment.root_dir` in the defaults.\n\n## Data Format\n\nClassify needs a 4-column CSV (`input_path,golden_path,label,object_name`) plus an images directory; segment uses a paired directory structure (`A\u002F`, `B\u002F`, `list\u002F`, `label\u002F`) under `dataset.segment.root_dir` instead of CSV. The `image_ext` field (default `.jpg`) must match the actual file extensions; if images are `.png`, set `dataset.classify.image_ext: .png`. Multi-lighting input is configured via `dataset.classify.input_map` (each lighting name maps to a channel index) with `dataset.classify.num_input` set to match. See `references\u002Fdata-formats.md` for the per-field input tables (classify train\u002Feval\u002Finference, segment), CSV column semantics, lighting\u002Fpath-concatenation conventions, the segment directory layout, and `input_map`\u002F`grid_map` examples.\n\n## Important Parameters\n\nKey knobs include `train.validation_interval` (default 50, must be ≤ num_epochs), `train.checkpoint_interval` (default 200, must be ≤ num_epochs), `train.num_epochs` (default 100), `model.classify.eval_margin` (default 0.3, the precision\u002Frecall threshold), `model.classify.train_margin_euclid` (default 2.0), `model.classify.embedding_vectors` (default 5), `dataset.classify.batch_size` (default 16, must be > 1), `dataset.classify.fpratio_sampling` (default 0.25), and `train.classify.cls_weight` (default [1.0, 10.0]). Hardware: minimum 1 GPU with 16GB+ VRAM, recommended 8 GPUs (DDP); do not set `gpu_spec_key` (GPU count is managed internally by TAO), `num_nodes` (default 1) controls multi-node. See `references\u002Ftuning-parameters.md` for the full per-parameter guidance and hardware detail.\n\n## Error Patterns\n\nFor checkpoint-not-found, CSV format mismatch, image extension mismatch, OOM, low evaluation accuracy, the contrastive-loss `AssertionError`, checkpoint load key mismatch at evaluate\u002Finference, non-convergence, segment-only backbone dimension mismatch, the `MultiScaleDeformableAttention` `OSError`, the Lightning `MisconfigurationException`, `ModuleNotFoundError: nvidia_tao_pytorch`, and epoch defaults, see `references\u002Ftroubleshooting.md` for the full symptom-and-fix list.\n\n## Spec Param \u002F Parent Model Inference\n\nModel-specific parent-model mappings are declared in `references\u002Fskill_info.yaml` under `spec_params`, so generated runners and agents resolve checkpoints before `create_job()` instead of guessing file names. For `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. See `references\u002Fparent-model-inference.md` for the full per-action spec-field-to-inference-function mapping table.\n\n## Deployment\n\n- [tao-deploy-visual-changenet](references\u002Ftao-deploy-visual-changenet.md)\n",{"name":5,"tags":49,"license":15,"metadata":54,"description":7,"allowed-tools":57,"compatibility":58},[32,50,34,51,43,52,53],"aoi","classification","siamese","visual-inspection",{"author":55,"version":56},"NVIDIA Corporation","0.1.0","Read Bash","Requires docker + nvidia-container-toolkit.","claude\u002Finstall-gsd-build-foQdV","d86cd127f527124b5fd0fd0bc679a4d79b7d01d7",".agents\u002Fskills\u002Ftao-train-visual-changenet","https:\u002F\u002Fgithub.com\u002Fpattanation\u002FNorvik","https:\u002F\u002Fnorvik-orpin.vercel.app","https:\u002F\u002Fgithub.com\u002Fpattanation","2026-07-17 11:17:04",{"source":8,"skillName":5,"commands":67,"prompt":77,"zipAvailable":78},[68,71,74],{"tool":69,"command":70},"npx","npx skills add pattanation\u002FNorvik --skill tao-train-visual-changenet",{"tool":72,"command":73},"bunx","bunx skills add pattanation\u002FNorvik --skill tao-train-visual-changenet",{"tool":75,"command":76},"pnpm","pnpm dlx skills add pattanation\u002FNorvik --skill tao-train-visual-changenet","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002Fpattanation\u002FNorvik\u002Fblob\u002Fclaude\u002Finstall-gsd-build-foQdV\u002F.agents\u002Fskills\u002Ftao-train-visual-changenet\u002FSKILL.md",true,[80,86,92,96,100,104,108,112,116,120,124,128,132,136,140,144,148,152,156,160,164,168,172,176,180,184,188,192,196,200,204,208,212,216,220],{"path":81,"resourceType":82,"resourceTypeDesc":83,"sizeBytes":84,"sourceHtmlUrl":85},"BENCHMARK.md","REFERENCE","参考文档",4117,"https:\u002F\u002Fgithub.com\u002Fpattanation\u002FNorvik\u002Fblob\u002Fclaude\u002Finstall-gsd-build-foQdV\u002F.agents\u002Fskills\u002Ftao-train-visual-changenet\u002FBENCHMARK.md",{"path":87,"resourceType":88,"resourceTypeDesc":89,"sizeBytes":90,"sourceHtmlUrl":91},"eval.config","OTHER","其他",5757,"https:\u002F\u002Fgithub.com\u002Fpattanation\u002FNorvik\u002Fblob\u002Fclaude\u002Finstall-gsd-build-foQdV\u002F.agents\u002Fskills\u002Ftao-train-visual-changenet\u002Feval.config",{"path":93,"resourceType":82,"resourceTypeDesc":83,"sizeBytes":94,"sourceHtmlUrl":95},"evals\u002Fevals.json",858,"https:\u002F\u002Fgithub.com\u002Fpattanation\u002FNorvik\u002Fblob\u002Fclaude\u002Finstall-gsd-build-foQdV\u002F.agents\u002Fskills\u002Ftao-train-visual-changenet\u002Fevals\u002Fevals.json",{"path":97,"resourceType":82,"resourceTypeDesc":83,"sizeBytes":98,"sourceHtmlUrl":99},"references\u002Fdata-formats.md",4046,"https:\u002F\u002Fgithub.com\u002Fpattanation\u002FNorvik\u002Fblob\u002Fclaude\u002Finstall-gsd-build-foQdV\u002F.agents\u002Fskills\u002Ftao-train-visual-changenet\u002Freferences\u002Fdata-formats.md",{"path":101,"resourceType":82,"resourceTypeDesc":83,"sizeBytes":102,"sourceHtmlUrl":103},"references\u002Flocal-docker.md",1886,"https:\u002F\u002Fgithub.com\u002Fpattanation\u002FNorvik\u002Fblob\u002Fclaude\u002Finstall-gsd-build-foQdV\u002F.agents\u002Fskills\u002Ftao-train-visual-changenet\u002Freferences\u002Flocal-docker.md",{"path":105,"resourceType":82,"resourceTypeDesc":83,"sizeBytes":106,"sourceHtmlUrl":107},"references\u002Fparent-model-inference.md",2571,"https:\u002F\u002Fgithub.com\u002Fpattanation\u002FNorvik\u002Fblob\u002Fclaude\u002Finstall-gsd-build-foQdV\u002F.agents\u002Fskills\u002Ftao-train-visual-changenet\u002Freferences\u002Fparent-model-inference.md",{"path":109,"resourceType":82,"resourceTypeDesc":83,"sizeBytes":110,"sourceHtmlUrl":111},"references\u002Fskill_info.yaml",9944,"https:\u002F\u002Fgithub.com\u002Fpattanation\u002FNorvik\u002Fblob\u002Fclaude\u002Finstall-gsd-build-foQdV\u002F.agents\u002Fskills\u002Ftao-train-visual-changenet\u002Freferences\u002Fskill_info.yaml",{"path":113,"resourceType":82,"resourceTypeDesc":83,"sizeBytes":114,"sourceHtmlUrl":115},"references\u002Fspec_template_deploy_classify_evaluate.yaml",1034,"https:\u002F\u002Fgithub.com\u002Fpattanation\u002FNorvik\u002Fblob\u002Fclaude\u002Finstall-gsd-build-foQdV\u002F.agents\u002Fskills\u002Ftao-train-visual-changenet\u002Freferences\u002Fspec_template_deploy_classify_evaluate.yaml",{"path":117,"resourceType":82,"resourceTypeDesc":83,"sizeBytes":118,"sourceHtmlUrl":119},"references\u002Fspec_template_deploy_classify_gen_trt_engine.yaml",430,"https:\u002F\u002Fgithub.com\u002Fpattanation\u002FNorvik\u002Fblob\u002Fclaude\u002Finstall-gsd-build-foQdV\u002F.agents\u002Fskills\u002Ftao-train-visual-changenet\u002Freferences\u002Fspec_template_deploy_classify_gen_trt_engine.yaml",{"path":121,"resourceType":82,"resourceTypeDesc":83,"sizeBytes":122,"sourceHtmlUrl":123},"references\u002Fspec_template_deploy_classify_inference.yaml",1035,"https:\u002F\u002Fgithub.com\u002Fpattanation\u002FNorvik\u002Fblob\u002Fclaude\u002Finstall-gsd-build-foQdV\u002F.agents\u002Fskills\u002Ftao-train-visual-changenet\u002Freferences\u002Fspec_template_deploy_classify_inference.yaml",{"path":125,"resourceType":82,"resourceTypeDesc":83,"sizeBytes":126,"sourceHtmlUrl":127},"references\u002Fspec_template_deploy_segment_evaluate.yaml",533,"https:\u002F\u002Fgithub.com\u002Fpattanation\u002FNorvik\u002Fblob\u002Fclaude\u002Finstall-gsd-build-foQdV\u002F.agents\u002Fskills\u002Ftao-train-visual-changenet\u002Freferences\u002Fspec_template_deploy_segment_evaluate.yaml",{"path":129,"resourceType":82,"resourceTypeDesc":83,"sizeBytes":130,"sourceHtmlUrl":131},"references\u002Fspec_template_deploy_segment_gen_trt_engine.yaml",419,"https:\u002F\u002Fgithub.com\u002Fpattanation\u002FNorvik\u002Fblob\u002Fclaude\u002Finstall-gsd-build-foQdV\u002F.agents\u002Fskills\u002Ftao-train-visual-changenet\u002Freferences\u002Fspec_template_deploy_segment_gen_trt_engine.yaml",{"path":133,"resourceType":82,"resourceTypeDesc":83,"sizeBytes":134,"sourceHtmlUrl":135},"references\u002Fspec_template_deploy_segment_inference.yaml",537,"https:\u002F\u002Fgithub.com\u002Fpattanation\u002FNorvik\u002Fblob\u002Fclaude\u002Finstall-gsd-build-foQdV\u002F.agents\u002Fskills\u002Ftao-train-visual-changenet\u002Freferences\u002Fspec_template_deploy_segment_inference.yaml",{"path":137,"resourceType":82,"resourceTypeDesc":83,"sizeBytes":138,"sourceHtmlUrl":139},"references\u002Fspec_template_evaluate.yaml",4214,"https:\u002F\u002Fgithub.com\u002Fpattanation\u002FNorvik\u002Fblob\u002Fclaude\u002Finstall-gsd-build-foQdV\u002F.agents\u002Fskills\u002Ftao-train-visual-changenet\u002Freferences\u002Fspec_template_evaluate.yaml",{"path":141,"resourceType":82,"resourceTypeDesc":83,"sizeBytes":142,"sourceHtmlUrl":143},"references\u002Fspec_template_export.yaml",3171,"https:\u002F\u002Fgithub.com\u002Fpattanation\u002FNorvik\u002Fblob\u002Fclaude\u002Finstall-gsd-build-foQdV\u002F.agents\u002Fskills\u002Ftao-train-visual-changenet\u002Freferences\u002Fspec_template_export.yaml",{"path":145,"resourceType":82,"resourceTypeDesc":83,"sizeBytes":146,"sourceHtmlUrl":147},"references\u002Fspec_template_inference.yaml",4215,"https:\u002F\u002Fgithub.com\u002Fpattanation\u002FNorvik\u002Fblob\u002Fclaude\u002Finstall-gsd-build-foQdV\u002F.agents\u002Fskills\u002Ftao-train-visual-changenet\u002Freferences\u002Fspec_template_inference.yaml",{"path":149,"resourceType":82,"resourceTypeDesc":83,"sizeBytes":150,"sourceHtmlUrl":151},"references\u002Fspec_template_quantize.yaml",2850,"https:\u002F\u002Fgithub.com\u002Fpattanation\u002FNorvik\u002Fblob\u002Fclaude\u002Finstall-gsd-build-foQdV\u002F.agents\u002Fskills\u002Ftao-train-visual-changenet\u002Freferences\u002Fspec_template_quantize.yaml",{"path":153,"resourceType":82,"resourceTypeDesc":83,"sizeBytes":154,"sourceHtmlUrl":155},"references\u002Fspec_template_segment_evaluate.yaml",4471,"https:\u002F\u002Fgithub.com\u002Fpattanation\u002FNorvik\u002Fblob\u002Fclaude\u002Finstall-gsd-build-foQdV\u002F.agents\u002Fskills\u002Ftao-train-visual-changenet\u002Freferences\u002Fspec_template_segment_evaluate.yaml",{"path":157,"resourceType":82,"resourceTypeDesc":83,"sizeBytes":158,"sourceHtmlUrl":159},"references\u002Fspec_template_segment_inference.yaml",4472,"https:\u002F\u002Fgithub.com\u002Fpattanation\u002FNorvik\u002Fblob\u002Fclaude\u002Finstall-gsd-build-foQdV\u002F.agents\u002Fskills\u002Ftao-train-visual-changenet\u002Freferences\u002Fspec_template_segment_inference.yaml",{"path":161,"resourceType":82,"resourceTypeDesc":83,"sizeBytes":162,"sourceHtmlUrl":163},"references\u002Fspec_template_segment_train.yaml",4321,"https:\u002F\u002Fgithub.com\u002Fpattanation\u002FNorvik\u002Fblob\u002Fclaude\u002Finstall-gsd-build-foQdV\u002F.agents\u002Fskills\u002Ftao-train-visual-changenet\u002Freferences\u002Fspec_template_segment_train.yaml",{"path":165,"resourceType":82,"resourceTypeDesc":83,"sizeBytes":166,"sourceHtmlUrl":167},"references\u002Fspec_template_segment.yaml",2163,"https:\u002F\u002Fgithub.com\u002Fpattanation\u002FNorvik\u002Fblob\u002Fclaude\u002Finstall-gsd-build-foQdV\u002F.agents\u002Fskills\u002Ftao-train-visual-changenet\u002Freferences\u002Fspec_template_segment.yaml",{"path":169,"resourceType":82,"resourceTypeDesc":83,"sizeBytes":170,"sourceHtmlUrl":171},"references\u002Fspec_template_train.yaml",2881,"https:\u002F\u002Fgithub.com\u002Fpattanation\u002FNorvik\u002Fblob\u002Fclaude\u002Finstall-gsd-build-foQdV\u002F.agents\u002Fskills\u002Ftao-train-visual-changenet\u002Freferences\u002Fspec_template_train.yaml",{"path":173,"resourceType":82,"resourceTypeDesc":83,"sizeBytes":174,"sourceHtmlUrl":175},"references\u002Ftao-deploy-visual-changenet.md",6921,"https:\u002F\u002Fgithub.com\u002Fpattanation\u002FNorvik\u002Fblob\u002Fclaude\u002Finstall-gsd-build-foQdV\u002F.agents\u002Fskills\u002Ftao-train-visual-changenet\u002Freferences\u002Ftao-deploy-visual-changenet.md",{"path":177,"resourceType":82,"resourceTypeDesc":83,"sizeBytes":178,"sourceHtmlUrl":179},"references\u002Ftao-deploy-visual-changenet.skill_info.yaml",4384,"https:\u002F\u002Fgithub.com\u002Fpattanation\u002FNorvik\u002Fblob\u002Fclaude\u002Finstall-gsd-build-foQdV\u002F.agents\u002Fskills\u002Ftao-train-visual-changenet\u002Freferences\u002Ftao-deploy-visual-changenet.skill_info.yaml",{"path":181,"resourceType":82,"resourceTypeDesc":83,"sizeBytes":182,"sourceHtmlUrl":183},"references\u002Ftroubleshooting.md",4617,"https:\u002F\u002Fgithub.com\u002Fpattanation\u002FNorvik\u002Fblob\u002Fclaude\u002Finstall-gsd-build-foQdV\u002F.agents\u002Fskills\u002Ftao-train-visual-changenet\u002Freferences\u002Ftroubleshooting.md",{"path":185,"resourceType":82,"resourceTypeDesc":83,"sizeBytes":186,"sourceHtmlUrl":187},"references\u002Ftuning-parameters.md",2495,"https:\u002F\u002Fgithub.com\u002Fpattanation\u002FNorvik\u002Fblob\u002Fclaude\u002Finstall-gsd-build-foQdV\u002F.agents\u002Fskills\u002Ftao-train-visual-changenet\u002Freferences\u002Ftuning-parameters.md",{"path":189,"resourceType":82,"resourceTypeDesc":83,"sizeBytes":190,"sourceHtmlUrl":191},"schemas\u002Fevaluate.schema.json",80479,"https:\u002F\u002Fgithub.com\u002Fpattanation\u002FNorvik\u002Fblob\u002Fclaude\u002Finstall-gsd-build-foQdV\u002F.agents\u002Fskills\u002Ftao-train-visual-changenet\u002Fschemas\u002Fevaluate.schema.json",{"path":193,"resourceType":82,"resourceTypeDesc":83,"sizeBytes":194,"sourceHtmlUrl":195},"schemas\u002Finference.schema.json",80480,"https:\u002F\u002Fgithub.com\u002Fpattanation\u002FNorvik\u002Fblob\u002Fclaude\u002Finstall-gsd-build-foQdV\u002F.agents\u002Fskills\u002Ftao-train-visual-changenet\u002Fschemas\u002Finference.schema.json",{"path":197,"resourceType":82,"resourceTypeDesc":83,"sizeBytes":198,"sourceHtmlUrl":199},"schemas\u002Fmanifest.json",34390,"https:\u002F\u002Fgithub.com\u002Fpattanation\u002FNorvik\u002Fblob\u002Fclaude\u002Finstall-gsd-build-foQdV\u002F.agents\u002Fskills\u002Ftao-train-visual-changenet\u002Fschemas\u002Fmanifest.json",{"path":201,"resourceType":82,"resourceTypeDesc":83,"sizeBytes":202,"sourceHtmlUrl":203},"schemas\u002Fsegment_evaluate.schema.json",80485,"https:\u002F\u002Fgithub.com\u002Fpattanation\u002FNorvik\u002Fblob\u002Fclaude\u002Finstall-gsd-build-foQdV\u002F.agents\u002Fskills\u002Ftao-train-visual-changenet\u002Fschemas\u002Fsegment_evaluate.schema.json",{"path":205,"resourceType":82,"resourceTypeDesc":83,"sizeBytes":206,"sourceHtmlUrl":207},"schemas\u002Fsegment_inference.schema.json",80486,"https:\u002F\u002Fgithub.com\u002Fpattanation\u002FNorvik\u002Fblob\u002Fclaude\u002Finstall-gsd-build-foQdV\u002F.agents\u002Fskills\u002Ftao-train-visual-changenet\u002Fschemas\u002Fsegment_inference.schema.json",{"path":209,"resourceType":82,"resourceTypeDesc":83,"sizeBytes":210,"sourceHtmlUrl":211},"schemas\u002Fsegment_train.schema.json",77668,"https:\u002F\u002Fgithub.com\u002Fpattanation\u002FNorvik\u002Fblob\u002Fclaude\u002Finstall-gsd-build-foQdV\u002F.agents\u002Fskills\u002Ftao-train-visual-changenet\u002Fschemas\u002Fsegment_train.schema.json",{"path":213,"resourceType":82,"resourceTypeDesc":83,"sizeBytes":214,"sourceHtmlUrl":215},"schemas\u002Ftrain.schema.json",77660,"https:\u002F\u002Fgithub.com\u002Fpattanation\u002FNorvik\u002Fblob\u002Fclaude\u002Finstall-gsd-build-foQdV\u002F.agents\u002Fskills\u002Ftao-train-visual-changenet\u002Fschemas\u002Ftrain.schema.json",{"path":217,"resourceType":82,"resourceTypeDesc":83,"sizeBytes":218,"sourceHtmlUrl":219},"skill-card.md",3866,"https:\u002F\u002Fgithub.com\u002Fpattanation\u002FNorvik\u002Fblob\u002Fclaude\u002Finstall-gsd-build-foQdV\u002F.agents\u002Fskills\u002Ftao-train-visual-changenet\u002Fskill-card.md",{"path":221,"resourceType":88,"resourceTypeDesc":89,"sizeBytes":222,"sourceHtmlUrl":223},"skill.oms.sig",12505,"https:\u002F\u002Fgithub.com\u002Fpattanation\u002FNorvik\u002Fblob\u002Fclaude\u002Finstall-gsd-build-foQdV\u002F.agents\u002Fskills\u002Ftao-train-visual-changenet\u002Fskill.oms.sig",[225,256,282,306,328,343],{"skillNo":226,"name":227,"nameZh":6,"description":228,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":229,"sourceHtmlUrl":230,"categories":231,"keywords":238,"publishedAt":44},"01M71QDV4DM8T19HEHWATEF1K4","webapp-testing","Toolkit for interacting with and testing local web applications using Playwright. Supports verifying frontend functionality, debugging UI behavior, capturing browser screenshots, and viewing browser logs.","https:\u002F\u002Fgithub.com\u002Fpattanation\u002FNorvik\u002Fblob\u002Fclaude\u002Finstall-gsd-build-foQdV\u002F.agents\u002Fskills\u002Fwebapp-testing\u002FLICENSE.txt","https:\u002F\u002Fgithub.com\u002Fpattanation\u002FNorvik\u002Fblob\u002Fclaude\u002Finstall-gsd-build-foQdV\u002F.agents\u002Fskills\u002Fwebapp-testing\u002FSKILL.md",[232,235],{"categoryNo":233,"categoryName":234,"taxonomyType":21,"taxonomyTypeDesc":22},"01JBSEEDCAT000000000000008","测试与质量",{"categoryNo":236,"categoryName":237,"taxonomyType":21,"taxonomyTypeDesc":22},"01JBSEEDCAT000000000000009","前端开发",[239,240,241,242,243,244,245,246,247,248,249,250,251,252,253,254,255],"webapp","testing","webapp testing","for","applications","local","frontend","debugging","toolkit","logs","browser","with","playwright","web","behavior","ui","screenshots",{"skillNo":257,"name":258,"nameZh":6,"description":259,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":260,"sourceHtmlUrl":261,"categories":262,"keywords":264,"publishedAt":44},"01M71QDH2GGMXGA9105C9W1N7B","web-artifacts-builder","Suite of tools for creating elaborate, multi-component claude.ai HTML artifacts using modern frontend web technologies (React, Tailwind CSS, shadcn\u002Fui). Use for complex artifacts requiring state management, routing, or shadcn\u002Fui components - not for simple single-file HTML\u002FJSX artifacts.","https:\u002F\u002Fgithub.com\u002Fpattanation\u002FNorvik\u002Fblob\u002Fclaude\u002Finstall-gsd-build-foQdV\u002F.agents\u002Fskills\u002Fweb-artifacts-builder\u002FLICENSE.txt","https:\u002F\u002Fgithub.com\u002Fpattanation\u002FNorvik\u002Fblob\u002Fclaude\u002Finstall-gsd-build-foQdV\u002F.agents\u002Fskills\u002Fweb-artifacts-builder\u002FSKILL.md",[263],{"categoryNo":236,"categoryName":237,"taxonomyType":21,"taxonomyTypeDesc":22},[265,266,267,252,268,269,270,242,271,272,245,273,274,275,276,277,278,279,280,281],"builder","artifacts builder","artifacts","web artifacts builder","single-file","css","Tailwind","ui components","single","use","file","components","react","multi","state","routing","state management",{"skillNo":283,"name":284,"nameZh":6,"description":285,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":286,"licenseFileUrl":287,"sourceHtmlUrl":288,"categories":289,"keywords":291,"publishedAt":44},"01M71QD9T2XC3EWMQNN9ACNWQ1","vueuse-functions","Apply VueUse composables where appropriate to build concise, maintainable Vue.js \u002F Nuxt features.","MIT","https:\u002F\u002Fgithub.com\u002Fpattanation\u002FNorvik\u002Fblob\u002Fclaude\u002Finstall-gsd-build-foQdV\u002F.agents\u002Fskills\u002Fvueuse-functions\u002FLICENSE.md","https:\u002F\u002Fgithub.com\u002Fpattanation\u002FNorvik\u002Fblob\u002Fclaude\u002Finstall-gsd-build-foQdV\u002F.agents\u002Fskills\u002Fvueuse-functions\u002FSKILL.md",[290],{"categoryNo":236,"categoryName":237,"taxonomyType":21,"taxonomyTypeDesc":22},[292,293,294,295,296,297,298,299,300,301,302,303,304,305],"functions","vueuse","vueuse functions","project","nuxt","apply","github","concise","build","https","features","vue","js","composables",{"skillNo":307,"name":308,"nameZh":6,"description":309,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":286,"licenseFileUrl":310,"sourceHtmlUrl":311,"categories":312,"keywords":315,"publishedAt":44},"01M71QD51HWSGYMCMB7RW7PGK8","vue-testing-best-practices","Use for Vue.js testing. Covers Vitest, Vue Test Utils, component testing, mocking, testing patterns, and Playwright for E2E testing.","https:\u002F\u002Fgithub.com\u002Fpattanation\u002FNorvik\u002Fblob\u002Fclaude\u002Finstall-gsd-build-foQdV\u002F.agents\u002Fskills\u002Fvue-testing-best-practices\u002FLICENSE.md","https:\u002F\u002Fgithub.com\u002Fpattanation\u002FNorvik\u002Fblob\u002Fclaude\u002Finstall-gsd-build-foQdV\u002F.agents\u002Fskills\u002Fvue-testing-best-practices\u002FSKILL.md",[313,314],{"categoryNo":233,"categoryName":234,"taxonomyType":21,"taxonomyTypeDesc":22},{"categoryNo":236,"categoryName":237,"taxonomyType":21,"taxonomyTypeDesc":22},[316,240,317,303,318,251,319,320,321,298,322,242,323,324,325,274,326,304,327],"vue testing best practices","testing best practices","best practices","test","testing patterns","component","vitest","mocking","e2e","utils","patterns","e2e testing",{"skillNo":329,"name":330,"nameZh":6,"description":331,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":286,"licenseFileUrl":332,"sourceHtmlUrl":333,"categories":334,"keywords":336,"publishedAt":44},"01M71QD358H2DJ9THJRBP77T3T","vue-router-best-practices","Vue Router 4 patterns, navigation guards, route params, and route-component lifecycle interactions.","https:\u002F\u002Fgithub.com\u002Fpattanation\u002FNorvik\u002Fblob\u002Fclaude\u002Finstall-gsd-build-foQdV\u002F.agents\u002Fskills\u002Fvue-router-best-practices\u002FLICENSE.md","https:\u002F\u002Fgithub.com\u002Fpattanation\u002FNorvik\u002Fblob\u002Fclaude\u002Finstall-gsd-build-foQdV\u002F.agents\u002Fskills\u002Fvue-router-best-practices\u002FSKILL.md",[335],{"categoryNo":236,"categoryName":237,"taxonomyType":21,"taxonomyTypeDesc":22},[337,303,338,318,339,321,340,298,326,341,342],"router","vue router best practices","route","guards","navigation","lifecycle",{"skillNo":344,"name":345,"nameZh":6,"description":346,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":286,"licenseFileUrl":347,"sourceHtmlUrl":348,"categories":349,"keywords":351,"publishedAt":44},"01M71QCYPNWDE2MQZ1TF3ED5S4","vue-best-practices","MUST be used for Vue.js tasks. Strongly recommends Composition API with `\u003Cscript setup>` and TypeScript as the standard approach. Covers Vue 3, SSR, Volar, vue-tsc. Load for any Vue, .vue files, Vue Router, Pinia, or Vite with Vue work. ALWAYS use Composition API unless the project explicitly requires Options API.","https:\u002F\u002Fgithub.com\u002Fpattanation\u002FNorvik\u002Fblob\u002Fclaude\u002Finstall-gsd-build-foQdV\u002F.agents\u002Fskills\u002Fvue-best-practices\u002FLICENSE.md","https:\u002F\u002Fgithub.com\u002Fpattanation\u002FNorvik\u002Fblob\u002Fclaude\u002Finstall-gsd-build-foQdV\u002F.agents\u002Fskills\u002Fvue-best-practices\u002FSKILL.md",[350],{"categoryNo":236,"categoryName":237,"taxonomyType":21,"taxonomyTypeDesc":22},[352,303,318,353,295,354,355,356,337,357,358,359,360,361,362,242,363,274,364,365],"vue best practices","Composition","files","tasks","api","pinia","work","options","ssr","typescript","as","vite","load","standard"]