[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M7A77S2HEEKWSC7D7K281GJW":3,"skill-same:01M7A77S2HEEKWSC7D7K281GJW":125},{"skillNo":4,"name":5,"nameZh":6,"description":7,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":16,"categories":17,"keywords":18,"publishedAt":19,"linkOnly":20,"markdownContent":21,"markdownBody":22,"frontMatter":23,"refName":32,"commitSha":33,"skillDirectory":34,"forksCount":35,"repositoryHtmlUrl":36,"repositoryHomepageUrl":37,"ownerHtmlUrl":38,"ownerBlogUrl":39,"ownerDescription":40,"updatedAt":41,"install":42,"resources":55},"01M7A77S2HEEKWSC7D7K281GJW","tao-train-image-classification",null,"PyTorch-based TAO image classification. Supports a wide range of backbones (FAN, EfficientNet, ResNet, etc.) with distillation and quantization for deployment. Use when training, evaluating, distilling, quantizing, exporting, or running inference for a TAO image-classification (PyT) model. Trigger phrases include \"train image classifier\", \"TAO classification\", \"ResNet\u002FEfficientNet\u002FFAN backbone classifier\", \"classification-pyt\".","themusebook\u002Fskills","themusebook","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F349927247?v=4","01M7A5C1RV5S3V5FF6ZACVDNRN","musebook","01M7A5C1SFC3KRZAXNN60XJ7KM",0,"Apache-2.0","https:\u002F\u002Fgithub.com\u002Fthemusebook\u002Fskills\u002Fblob\u002Fmain\u002Fpublic\u002Fapi\u002Fskills\u002Fnvidia\u002Ftao-train-image-classification\u002FSKILL.md",[],[],"2026-11-15 14:19:17",false,"---\nname: tao-train-image-classification\ndescription: PyTorch-based TAO image classification. Supports a wide range of backbones (FAN, EfficientNet, ResNet, etc.)\n  with distillation and quantization for deployment. Use when training, evaluating, distilling, quantizing, exporting, or\n  running inference for a TAO image-classification (PyT) model. Trigger phrases include \"train image classifier\",\n  \"TAO classification\", \"ResNet\u002FEfficientNet\u002FFAN backbone classifier\", \"classification-pyt\".\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- image\n- classification\n---\n\n# Classification PyT\n\nPyTorch image classification. Supports a wide range of backbones (FAN, EfficientNet, ResNet, etc.) with distillation and quantization for deployment.\n\nSet model.backbone.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-image-classification.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## Training Requirements\n\n- **Dataset type:** image_classification\n- **Formats:** classification_pyt\n- **Monitoring metric:** val_acc_1\n\n### Per-Action Dataset Requirements\n\n| Action | Spec Key | Source | Files | List? |\n|---|---|---|---|---|\n| distill | dataset.train_dataset.images_dir | train_datasets | images_train.tar.gz | No |\n| distill | dataset.classes_file | train_datasets | classes.txt | No |\n| distill | dataset.val_dataset.images_dir | eval_dataset | images_val.tar.gz | No |\n| evaluate | dataset.val_dataset.images_dir | eval_dataset | images_val.tar.gz | No |\n| evaluate | dataset.classes_file | eval_dataset | classes.txt | No |\n| evaluate | dataset.test_dataset.images_dir | inference_dataset | images_test.tar.gz | No |\n| export | dataset.root_dir | train_datasets |  | No |\n| inference | dataset.val_dataset.images_dir | eval_dataset | images_val.tar.gz | No |\n| inference | dataset.classes_file | eval_dataset | classes.txt | No |\n| inference | dataset.test_dataset.images_dir | inference_dataset | images_test.tar.gz | No |\n| quantize | dataset.train_dataset.images_dir | train_datasets | images_train.tar.gz | No |\n| quantize | dataset.classes_file | train_datasets | classes.txt | No |\n| quantize | dataset.val_dataset.images_dir | eval_dataset | images_val.tar.gz | No |\n| quantize | dataset.quant_calibration_dataset.images_dir | calibration_dataset | images_train.tar.gz | No |\n| train | dataset.train_dataset.images_dir | train_datasets | images_train.tar.gz | No |\n| train | dataset.classes_file | train_datasets | classes.txt | No |\n| train | dataset.val_dataset.images_dir | eval_dataset | images_val.tar.gz | 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\nTRAIN_IMAGES_DIR = \"\u002Fworkspace\u002Fdata\u002Fextracted\u002Ftrain\u002Fimages_train\"\nVAL_IMAGES_DIR = \"\u002Fworkspace\u002Fdata\u002Fextracted\u002Fval\u002Fimages_val\"\nTEST_IMAGES_DIR = \"\u002Fworkspace\u002Fdata\u002Fextracted\u002Ftest\u002Fimages_test\"\nCLASSES_FILE = \"\u002Fworkspace\u002Fdata\u002Fs3\u002Fclasses.txt\"\n```\n\nFor local Docker, download the S3 archives, extract them first, and point\n`dataset.*.images_dir` at the extracted class-root folder. Do not pass\n`images_train.tar.gz`, `images_val.tar.gz`, or `images_test.tar.gz` directly to\nlocal Docker specs; the skill metadata declares these inputs as folders.\n\n**train (mandatory data sources):**\n```python\n{\n    \"train.num_epochs\": 2,\n    \"train.validation_interval\": 2,\n    \"train.checkpoint_interval\": 2,\n    \"train.num_gpus\": 1,\n    \"dataset.train_dataset.images_dir\": TRAIN_IMAGES_DIR,\n    \"dataset.classes_file\": CLASSES_FILE,\n    \"dataset.val_dataset.images_dir\": VAL_IMAGES_DIR,\n}\n```\n\n**export (mandatory data sources):**\n```python\n{\n    \"export.input_height\": 224,\n    \"export.input_width\": 224,\n    \"dataset.root_dir\": \"\u002Fworkspace\u002Fdata\u002Fextracted\",\n}\n```\n\n**gen_trt_engine:**\n```python\n{\n    \"gen_trt_engine.tensorrt.data_type\": \"fp16\",\n}\n```\n\n**inference (mandatory data sources):**\n```python\n{\n    \"dataset.batch_size\": 1,\n    \"dataset.val_dataset.images_dir\": VAL_IMAGES_DIR,\n    \"dataset.classes_file\": CLASSES_FILE,\n    \"dataset.test_dataset.images_dir\": TEST_IMAGES_DIR,\n}\n```\n\n**distill (mandatory data sources):**\n```python\n{\n    \"dataset.train_dataset.images_dir\": TRAIN_IMAGES_DIR,\n    \"dataset.classes_file\": CLASSES_FILE,\n    \"dataset.val_dataset.images_dir\": VAL_IMAGES_DIR,\n    \"train.optim.policy\": \"step\",\n}\n```\n\n**evaluate (mandatory data sources):**\n```python\n{\n    \"dataset.val_dataset.images_dir\": VAL_IMAGES_DIR,\n    \"dataset.classes_file\": CLASSES_FILE,\n    \"dataset.test_dataset.images_dir\": TEST_IMAGES_DIR,\n}\n```\n\n**quantize (mandatory data sources):**\n```python\n{\n    \"dataset.train_dataset.images_dir\": TRAIN_IMAGES_DIR,\n    \"dataset.classes_file\": CLASSES_FILE,\n    \"dataset.val_dataset.images_dir\": VAL_IMAGES_DIR,\n    \"dataset.quant_calibration_dataset.images_dir\": TRAIN_IMAGES_DIR,\n}\n```\n## Eval Dataset\n\nOptional. Validation images are provided as a separate tar alongside training images.\nFor small smoke datasets that do not provide a separate `images_test.tar.gz`,\nset `dataset.test_dataset.images_dir` to the validation archive so evaluate and\ninference still exercise the checkpoint handoff.\n\n## Important Parameters\n\n- **dataset.num_classes**: Number of classes. Default 20. Must match the number of subdirectories in your image tarballs.\n- **model.backbone.type**: Default fan_small_12_p4_hybrid. Supported backbones and their head in_channels (from model_params_mapping.py): FAN: fan_tiny, fan_small_12_p4_hybrid, fan_base_16_p4_hybrid, fan_large_16_p4_hybrid. GCViT: gcvit_tiny through gcvit_large. FasterViT: fastervit_0 through fastervit_6. ViT\u002FEVA\u002FDINO: vit_large_patch14_dinov2, eva02_large_patch14, etc. SigLIP-CLIPA: ViT-H-14-SigLIP-CLIPA-224, etc. Some backbones require non-default input resolution (384, 512, 768).\n- **dataset.classes_file**: Path to classes.txt listing class names.\n- **train.optim.lr**: Learning rate. Default 6e-5.\n- **dataset.img_size**: Input image size. Default 224.\n- **dataset.batch_size**: Per-GPU batch size. Default 8.\n\n## Multi-GPU \u002F Multi-Node\n\n**Launch method:** Lightning-managed (single `python` process, Lightning spawns workers).\n\n| Spec Key | Description | Default |\n|----------|-------------|---------|\n| `train.num_gpus` | Number of GPUs | 1 |\n| `train.gpu_ids` | GPU device indices | [0] |\n| `train.num_nodes` | Number of nodes | 1 |\n\n- Multi-GPU strategy: `ddp_find_unused_parameters_true`\n- No fsdp support\n\n**Multi-node env vars** (set by orchestrator): `WORLD_SIZE`, `NODE_RANK`, `MASTER_ADDR`, `MASTER_PORT`, `NUM_GPU_PER_NODE`.\n\n## Hardware\n\nMinimum 1 GPU(s), recommended 2 GPU(s). 16GB+ (V100 or A100) VRAM per GPU. Classification is generally lightweight. Most backbones at 224x224 fit well on 16GB GPUs with batch_size=8.\n\n## Error Patterns\n\n**CUDA out of memory**: Reduce batch_size or use a smaller backbone.\n\n**num_classes mismatch**: Ensure dataset.num_classes matches the actual class directories in your image tarballs and classes.txt.\n\n**Empty class directory**: Every class in classes.txt must have at least one image in the corresponding subdirectory.\n\n**Distill scheduler default**: The bundled distill template and schema use\n`train.optim.policy: step`. Keep that setting for distill specs unless the\ncontainer implementation is updated; the 7.0 PyT distiller does not assign a\nscheduler interval for `train.optim.policy: linear`.\n\n**Checkpoint handoff**: Training produces `model_epoch_*.pth` checkpoints and a\n`classifier_model_latest.pth` symlink. For evaluate, inference, export, quantize,\ndistill, and resume, select the exact intended epoch checkpoint through the SDK\nresolver; use the latest symlink only when the user explicitly requests latest.\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 `classification_pyt.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 | `results_dir` | `output_dir` | current job results directory |\n| evaluate | `evaluate.checkpoint` | `parent_model` | model file inferred from the parent job results folder |\n| evaluate | `results_dir` | `output_dir` | current job results directory |\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| gen_trt_engine | `gen_trt_engine.onnx_file` | `parent_model` | model file inferred from the parent job results folder |\n| gen_trt_engine | `gen_trt_engine.trt_engine` | `create_engine_file` | output TensorRT engine path |\n| gen_trt_engine | `results_dir` | `output_dir` | current job results directory |\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 | `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 | `model.backbone.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-image-classification](references\u002Ftao-deploy-image-classification.md)\n","# Classification PyT\n\nPyTorch image classification. Supports a wide range of backbones (FAN, EfficientNet, ResNet, etc.) with distillation and quantization for deployment.\n\nSet model.backbone.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-image-classification.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## Training Requirements\n\n- **Dataset type:** image_classification\n- **Formats:** classification_pyt\n- **Monitoring metric:** val_acc_1\n\n### Per-Action Dataset Requirements\n\n| Action | Spec Key | Source | Files | List? |\n|---|---|---|---|---|\n| distill | dataset.train_dataset.images_dir | train_datasets | images_train.tar.gz | No |\n| distill | dataset.classes_file | train_datasets | classes.txt | No |\n| distill | dataset.val_dataset.images_dir | eval_dataset | images_val.tar.gz | No |\n| evaluate | dataset.val_dataset.images_dir | eval_dataset | images_val.tar.gz | No |\n| evaluate | dataset.classes_file | eval_dataset | classes.txt | No |\n| evaluate | dataset.test_dataset.images_dir | inference_dataset | images_test.tar.gz | No |\n| export | dataset.root_dir | train_datasets |  | No |\n| inference | dataset.val_dataset.images_dir | eval_dataset | images_val.tar.gz | No |\n| inference | dataset.classes_file | eval_dataset | classes.txt | No |\n| inference | dataset.test_dataset.images_dir | inference_dataset | images_test.tar.gz | No |\n| quantize | dataset.train_dataset.images_dir | train_datasets | images_train.tar.gz | No |\n| quantize | dataset.classes_file | train_datasets | classes.txt | No |\n| quantize | dataset.val_dataset.images_dir | eval_dataset | images_val.tar.gz | No |\n| quantize | dataset.quant_calibration_dataset.images_dir | calibration_dataset | images_train.tar.gz | No |\n| train | dataset.train_dataset.images_dir | train_datasets | images_train.tar.gz | No |\n| train | dataset.classes_file | train_datasets | classes.txt | No |\n| train | dataset.val_dataset.images_dir | eval_dataset | images_val.tar.gz | 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\nTRAIN_IMAGES_DIR = \"\u002Fworkspace\u002Fdata\u002Fextracted\u002Ftrain\u002Fimages_train\"\nVAL_IMAGES_DIR = \"\u002Fworkspace\u002Fdata\u002Fextracted\u002Fval\u002Fimages_val\"\nTEST_IMAGES_DIR = \"\u002Fworkspace\u002Fdata\u002Fextracted\u002Ftest\u002Fimages_test\"\nCLASSES_FILE = \"\u002Fworkspace\u002Fdata\u002Fs3\u002Fclasses.txt\"\n```\n\nFor local Docker, download the S3 archives, extract them first, and point\n`dataset.*.images_dir` at the extracted class-root folder. Do not pass\n`images_train.tar.gz`, `images_val.tar.gz`, or `images_test.tar.gz` directly to\nlocal Docker specs; the skill metadata declares these inputs as folders.\n\n**train (mandatory data sources):**\n```python\n{\n    \"train.num_epochs\": 2,\n    \"train.validation_interval\": 2,\n    \"train.checkpoint_interval\": 2,\n    \"train.num_gpus\": 1,\n    \"dataset.train_dataset.images_dir\": TRAIN_IMAGES_DIR,\n    \"dataset.classes_file\": CLASSES_FILE,\n    \"dataset.val_dataset.images_dir\": VAL_IMAGES_DIR,\n}\n```\n\n**export (mandatory data sources):**\n```python\n{\n    \"export.input_height\": 224,\n    \"export.input_width\": 224,\n    \"dataset.root_dir\": \"\u002Fworkspace\u002Fdata\u002Fextracted\",\n}\n```\n\n**gen_trt_engine:**\n```python\n{\n    \"gen_trt_engine.tensorrt.data_type\": \"fp16\",\n}\n```\n\n**inference (mandatory data sources):**\n```python\n{\n    \"dataset.batch_size\": 1,\n    \"dataset.val_dataset.images_dir\": VAL_IMAGES_DIR,\n    \"dataset.classes_file\": CLASSES_FILE,\n    \"dataset.test_dataset.images_dir\": TEST_IMAGES_DIR,\n}\n```\n\n**distill (mandatory data sources):**\n```python\n{\n    \"dataset.train_dataset.images_dir\": TRAIN_IMAGES_DIR,\n    \"dataset.classes_file\": CLASSES_FILE,\n    \"dataset.val_dataset.images_dir\": VAL_IMAGES_DIR,\n    \"train.optim.policy\": \"step\",\n}\n```\n\n**evaluate (mandatory data sources):**\n```python\n{\n    \"dataset.val_dataset.images_dir\": VAL_IMAGES_DIR,\n    \"dataset.classes_file\": CLASSES_FILE,\n    \"dataset.test_dataset.images_dir\": TEST_IMAGES_DIR,\n}\n```\n\n**quantize (mandatory data sources):**\n```python\n{\n    \"dataset.train_dataset.images_dir\": TRAIN_IMAGES_DIR,\n    \"dataset.classes_file\": CLASSES_FILE,\n    \"dataset.val_dataset.images_dir\": VAL_IMAGES_DIR,\n    \"dataset.quant_calibration_dataset.images_dir\": TRAIN_IMAGES_DIR,\n}\n```\n## Eval Dataset\n\nOptional. Validation images are provided as a separate tar alongside training images.\nFor small smoke datasets that do not provide a separate `images_test.tar.gz`,\nset `dataset.test_dataset.images_dir` to the validation archive so evaluate and\ninference still exercise the checkpoint handoff.\n\n## Important Parameters\n\n- **dataset.num_classes**: Number of classes. Default 20. Must match the number of subdirectories in your image tarballs.\n- **model.backbone.type**: Default fan_small_12_p4_hybrid. Supported backbones and their head in_channels (from model_params_mapping.py): FAN: fan_tiny, fan_small_12_p4_hybrid, fan_base_16_p4_hybrid, fan_large_16_p4_hybrid. GCViT: gcvit_tiny through gcvit_large. FasterViT: fastervit_0 through fastervit_6. ViT\u002FEVA\u002FDINO: vit_large_patch14_dinov2, eva02_large_patch14, etc. SigLIP-CLIPA: ViT-H-14-SigLIP-CLIPA-224, etc. Some backbones require non-default input resolution (384, 512, 768).\n- **dataset.classes_file**: Path to classes.txt listing class names.\n- **train.optim.lr**: Learning rate. Default 6e-5.\n- **dataset.img_size**: Input image size. Default 224.\n- **dataset.batch_size**: Per-GPU batch size. Default 8.\n\n## Multi-GPU \u002F Multi-Node\n\n**Launch method:** Lightning-managed (single `python` process, Lightning spawns workers).\n\n| Spec Key | Description | Default |\n|----------|-------------|---------|\n| `train.num_gpus` | Number of GPUs | 1 |\n| `train.gpu_ids` | GPU device indices | [0] |\n| `train.num_nodes` | Number of nodes | 1 |\n\n- Multi-GPU strategy: `ddp_find_unused_parameters_true`\n- No fsdp support\n\n**Multi-node env vars** (set by orchestrator): `WORLD_SIZE`, `NODE_RANK`, `MASTER_ADDR`, `MASTER_PORT`, `NUM_GPU_PER_NODE`.\n\n## Hardware\n\nMinimum 1 GPU(s), recommended 2 GPU(s). 16GB+ (V100 or A100) VRAM per GPU. Classification is generally lightweight. Most backbones at 224x224 fit well on 16GB GPUs with batch_size=8.\n\n## Error Patterns\n\n**CUDA out of memory**: Reduce batch_size or use a smaller backbone.\n\n**num_classes mismatch**: Ensure dataset.num_classes matches the actual class directories in your image tarballs and classes.txt.\n\n**Empty class directory**: Every class in classes.txt must have at least one image in the corresponding subdirectory.\n\n**Distill scheduler default**: The bundled distill template and schema use\n`train.optim.policy: step`. Keep that setting for distill specs unless the\ncontainer implementation is updated; the 7.0 PyT distiller does not assign a\nscheduler interval for `train.optim.policy: linear`.\n\n**Checkpoint handoff**: Training produces `model_epoch_*.pth` checkpoints and a\n`classifier_model_latest.pth` symlink. For evaluate, inference, export, quantize,\ndistill, and resume, select the exact intended epoch checkpoint through the SDK\nresolver; use the latest symlink only when the user explicitly requests latest.\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 `classification_pyt.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 | `results_dir` | `output_dir` | current job results directory |\n| evaluate | `evaluate.checkpoint` | `parent_model` | model file inferred from the parent job results folder |\n| evaluate | `results_dir` | `output_dir` | current job results directory |\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| gen_trt_engine | `gen_trt_engine.onnx_file` | `parent_model` | model file inferred from the parent job results folder |\n| gen_trt_engine | `gen_trt_engine.trt_engine` | `create_engine_file` | output TensorRT engine path |\n| gen_trt_engine | `results_dir` | `output_dir` | current job results directory |\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 | `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 | `model.backbone.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-image-classification](references\u002Ftao-deploy-image-classification.md)\n",{"name":5,"tags":24,"license":15,"metadata":27,"description":7,"allowed-tools":30,"compatibility":31},[25,26],"image","classification",{"author":28,"version":29},"NVIDIA Corporation","0.1.0","Read Bash","Requires docker + nvidia-container-toolkit.","main","1c523aba5cdf867a79f078847075492d7778092a","public\u002Fapi\u002Fskills\u002Fnvidia\u002Ftao-train-image-classification",1,"https:\u002F\u002Fgithub.com\u002Fthemusebook\u002Fskills","https:\u002F\u002Fskills.x402.wtf","https:\u002F\u002Fgithub.com\u002Fthemusebook","https:\u002F\u002Fmusebook.trade","The official Github of musebook","2026-11-02 23:42:26",{"source":8,"skillName":5,"commands":43,"prompt":53,"zipAvailable":54},[44,47,50],{"tool":45,"command":46},"npx","npx skills add themusebook\u002Fskills --skill tao-train-image-classification",{"tool":48,"command":49},"bunx","bunx skills add themusebook\u002Fskills --skill tao-train-image-classification",{"tool":51,"command":52},"pnpm","pnpm dlx skills add themusebook\u002Fskills --skill tao-train-image-classification","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002Fthemusebook\u002Fskills\u002Fblob\u002Fmain\u002Fpublic\u002Fapi\u002Fskills\u002Fnvidia\u002Ftao-train-image-classification\u002FSKILL.md",true,[56,62,66,70,74,78,82,86,90,94,98,102,106,109,113,117,121],{"path":57,"resourceType":58,"resourceTypeDesc":59,"sizeBytes":60,"sourceHtmlUrl":61},"BENCHMARK.md","REFERENCE","参考文档",4149,"https:\u002F\u002Fgithub.com\u002Fthemusebook\u002Fskills\u002Fblob\u002Fmain\u002Fpublic\u002Fapi\u002Fskills\u002Fnvidia\u002Ftao-train-image-classification\u002FBENCHMARK.md",{"path":63,"resourceType":58,"resourceTypeDesc":59,"sizeBytes":64,"sourceHtmlUrl":65},"metadata.json",679,"https:\u002F\u002Fgithub.com\u002Fthemusebook\u002Fskills\u002Fblob\u002Fmain\u002Fpublic\u002Fapi\u002Fskills\u002Fnvidia\u002Ftao-train-image-classification\u002Fmetadata.json",{"path":67,"resourceType":58,"resourceTypeDesc":59,"sizeBytes":68,"sourceHtmlUrl":69},"references\u002Fskill_info.yaml",7049,"https:\u002F\u002Fgithub.com\u002Fthemusebook\u002Fskills\u002Fblob\u002Fmain\u002Fpublic\u002Fapi\u002Fskills\u002Fnvidia\u002Ftao-train-image-classification\u002Freferences\u002Fskill_info.yaml",{"path":71,"resourceType":58,"resourceTypeDesc":59,"sizeBytes":72,"sourceHtmlUrl":73},"references\u002Fspec_template_deploy_evaluate.yaml",200,"https:\u002F\u002Fgithub.com\u002Fthemusebook\u002Fskills\u002Fblob\u002Fmain\u002Fpublic\u002Fapi\u002Fskills\u002Fnvidia\u002Ftao-train-image-classification\u002Freferences\u002Fspec_template_deploy_evaluate.yaml",{"path":75,"resourceType":58,"resourceTypeDesc":59,"sizeBytes":76,"sourceHtmlUrl":77},"references\u002Fspec_template_deploy_gen_trt_engine.yaml",217,"https:\u002F\u002Fgithub.com\u002Fthemusebook\u002Fskills\u002Fblob\u002Fmain\u002Fpublic\u002Fapi\u002Fskills\u002Fnvidia\u002Ftao-train-image-classification\u002Freferences\u002Fspec_template_deploy_gen_trt_engine.yaml",{"path":79,"resourceType":58,"resourceTypeDesc":59,"sizeBytes":80,"sourceHtmlUrl":81},"references\u002Fspec_template_deploy_inference.yaml",168,"https:\u002F\u002Fgithub.com\u002Fthemusebook\u002Fskills\u002Fblob\u002Fmain\u002Fpublic\u002Fapi\u002Fskills\u002Fnvidia\u002Ftao-train-image-classification\u002Freferences\u002Fspec_template_deploy_inference.yaml",{"path":83,"resourceType":58,"resourceTypeDesc":59,"sizeBytes":84,"sourceHtmlUrl":85},"references\u002Fspec_template_distill.yaml",3010,"https:\u002F\u002Fgithub.com\u002Fthemusebook\u002Fskills\u002Fblob\u002Fmain\u002Fpublic\u002Fapi\u002Fskills\u002Fnvidia\u002Ftao-train-image-classification\u002Freferences\u002Fspec_template_distill.yaml",{"path":87,"resourceType":58,"resourceTypeDesc":59,"sizeBytes":88,"sourceHtmlUrl":89},"references\u002Fspec_template_evaluate.yaml",3185,"https:\u002F\u002Fgithub.com\u002Fthemusebook\u002Fskills\u002Fblob\u002Fmain\u002Fpublic\u002Fapi\u002Fskills\u002Fnvidia\u002Ftao-train-image-classification\u002Freferences\u002Fspec_template_evaluate.yaml",{"path":91,"resourceType":58,"resourceTypeDesc":59,"sizeBytes":92,"sourceHtmlUrl":93},"references\u002Fspec_template_export.yaml",3279,"https:\u002F\u002Fgithub.com\u002Fthemusebook\u002Fskills\u002Fblob\u002Fmain\u002Fpublic\u002Fapi\u002Fskills\u002Fnvidia\u002Ftao-train-image-classification\u002Freferences\u002Fspec_template_export.yaml",{"path":95,"resourceType":58,"resourceTypeDesc":59,"sizeBytes":96,"sourceHtmlUrl":97},"references\u002Fspec_template_gen_trt_engine.yaml",3405,"https:\u002F\u002Fgithub.com\u002Fthemusebook\u002Fskills\u002Fblob\u002Fmain\u002Fpublic\u002Fapi\u002Fskills\u002Fnvidia\u002Ftao-train-image-classification\u002Freferences\u002Fspec_template_gen_trt_engine.yaml",{"path":99,"resourceType":58,"resourceTypeDesc":59,"sizeBytes":100,"sourceHtmlUrl":101},"references\u002Fspec_template_inference.yaml",3186,"https:\u002F\u002Fgithub.com\u002Fthemusebook\u002Fskills\u002Fblob\u002Fmain\u002Fpublic\u002Fapi\u002Fskills\u002Fnvidia\u002Ftao-train-image-classification\u002Freferences\u002Fspec_template_inference.yaml",{"path":103,"resourceType":58,"resourceTypeDesc":59,"sizeBytes":104,"sourceHtmlUrl":105},"references\u002Fspec_template_quantize.yaml",3012,"https:\u002F\u002Fgithub.com\u002Fthemusebook\u002Fskills\u002Fblob\u002Fmain\u002Fpublic\u002Fapi\u002Fskills\u002Fnvidia\u002Ftao-train-image-classification\u002Freferences\u002Fspec_template_quantize.yaml",{"path":107,"resourceType":58,"resourceTypeDesc":59,"sizeBytes":104,"sourceHtmlUrl":108},"references\u002Fspec_template_train.yaml","https:\u002F\u002Fgithub.com\u002Fthemusebook\u002Fskills\u002Fblob\u002Fmain\u002Fpublic\u002Fapi\u002Fskills\u002Fnvidia\u002Ftao-train-image-classification\u002Freferences\u002Fspec_template_train.yaml",{"path":110,"resourceType":58,"resourceTypeDesc":59,"sizeBytes":111,"sourceHtmlUrl":112},"references\u002Ftao-deploy-image-classification.md",5785,"https:\u002F\u002Fgithub.com\u002Fthemusebook\u002Fskills\u002Fblob\u002Fmain\u002Fpublic\u002Fapi\u002Fskills\u002Fnvidia\u002Ftao-train-image-classification\u002Freferences\u002Ftao-deploy-image-classification.md",{"path":114,"resourceType":58,"resourceTypeDesc":59,"sizeBytes":115,"sourceHtmlUrl":116},"references\u002Ftao-deploy-image-classification.skill_info.yaml",2423,"https:\u002F\u002Fgithub.com\u002Fthemusebook\u002Fskills\u002Fblob\u002Fmain\u002Fpublic\u002Fapi\u002Fskills\u002Fnvidia\u002Ftao-train-image-classification\u002Freferences\u002Ftao-deploy-image-classification.skill_info.yaml",{"path":118,"resourceType":58,"resourceTypeDesc":59,"sizeBytes":119,"sourceHtmlUrl":120},"skill-card.md",3736,"https:\u002F\u002Fgithub.com\u002Fthemusebook\u002Fskills\u002Fblob\u002Fmain\u002Fpublic\u002Fapi\u002Fskills\u002Fnvidia\u002Ftao-train-image-classification\u002Fskill-card.md",{"path":122,"resourceType":58,"resourceTypeDesc":59,"sizeBytes":123,"sourceHtmlUrl":124},"verification.json",4071,"https:\u002F\u002Fgithub.com\u002Fthemusebook\u002Fskills\u002Fblob\u002Fmain\u002Fpublic\u002Fapi\u002Fskills\u002Fnvidia\u002Ftao-train-image-classification\u002Fverification.json",[126,135,142,149,156,163],{"skillNo":127,"name":128,"nameZh":6,"description":129,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":130,"licenseFileUrl":131,"sourceHtmlUrl":132,"categories":133,"keywords":134,"publishedAt":19},"01M7AB34SPKVSDRRG009PM1CZH","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","MIT","https:\u002F\u002Fgithub.com\u002Fthemusebook\u002Fskills\u002Fblob\u002Fmain\u002Fskills\u002Fyoutube-clipper\u002FLICENSE","https:\u002F\u002Fgithub.com\u002Fthemusebook\u002Fskills\u002Fblob\u002Fmain\u002Fskills\u002Fyoutube-clipper\u002FSKILL.md",[],[],{"skillNo":136,"name":137,"nameZh":6,"description":138,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":130,"licenseFileUrl":6,"sourceHtmlUrl":139,"categories":140,"keywords":141,"publishedAt":19},"01M7AB18BH6N7KR7939HRNWVMQ","viem-integration","Integrate EVM blockchains using viem. Use when user says \"read blockchain data\", \"send transaction\", \"interact with smart contract\", \"connect to Ethereum\", \"use viem\", \"use wagmi\", \"wallet integration\", \"viem setup\", or mentions blockchain\u002FEVM development with TypeScript.","https:\u002F\u002Fgithub.com\u002Fthemusebook\u002Fskills\u002Fblob\u002Fmain\u002Fskills\u002Fviem-integration\u002FSKILL.md",[],[],{"skillNo":143,"name":144,"nameZh":6,"description":145,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":130,"licenseFileUrl":6,"sourceHtmlUrl":146,"categories":147,"keywords":148,"publishedAt":19},"01M7AB12GHGP48WCNXDQQ9222H","v4-security-foundations","Security-first Uniswap v4 hook development. Use when user mentions \"v4 hooks\", \"hook security\", \"PoolManager\", \"beforeSwap\", \"afterSwap\", or asks about V4 hook best practices, vulnerabilities, or audit requirements.","https:\u002F\u002Fgithub.com\u002Fthemusebook\u002Fskills\u002Fblob\u002Fmain\u002Fskills\u002Fv4-security-foundations\u002FSKILL.md",[],[],{"skillNo":150,"name":151,"nameZh":6,"description":152,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":130,"licenseFileUrl":6,"sourceHtmlUrl":153,"categories":154,"keywords":155,"publishedAt":19},"01M7AB0XM3PSR7V3Q2ZRR0Q0FB","v4-sdk-integration","App-layer SDK guide for building swap and liquidity experiences directly with the Uniswap v4 SDK. Use when user asks about \"v4 sdk\", \"uniswap v4\", \"v4 swap\", \"v4 liquidity\", \"PoolManager\", \"V4Planner\", \"StateView\", \"PositionManager\", \"pool state\", \"v4 position\", \"uniswap sdk\", or when building swap\u002Fliquidity UX directly with SDKs rather than via the Trading API.","https:\u002F\u002Fgithub.com\u002Fthemusebook\u002Fskills\u002Fblob\u002Fmain\u002Fskills\u002Fv4-sdk-integration\u002FSKILL.md",[],[],{"skillNo":157,"name":158,"nameZh":6,"description":159,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":130,"licenseFileUrl":6,"sourceHtmlUrl":160,"categories":161,"keywords":162,"publishedAt":19},"01M7AB0SGA93ASA588DJW4P9WT","v4-hook-generator","Generate Uniswap v4 hook contracts via OpenZeppelin MCP. Use when building custom swap logic, async swaps, hook-owned liquidity, custom curves, dynamic fees, MEV protection, limit orders, or oracle hooks.","https:\u002F\u002Fgithub.com\u002Fthemusebook\u002Fskills\u002Fblob\u002Fmain\u002Fskills\u002Fv4-hook-generator\u002FSKILL.md",[],[],{"skillNo":164,"name":165,"nameZh":6,"description":166,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":130,"licenseFileUrl":6,"sourceHtmlUrl":167,"categories":168,"keywords":169,"publishedAt":19},"01M7AAZYH75H4CS03WSW2EZ8DM","testing","For testing with Light Protocol programs and clients on localnet, devnet, and mainnet validation.","https:\u002F\u002Fgithub.com\u002Fthemusebook\u002Fskills\u002Fblob\u002Fmain\u002Fskills\u002Ftesting\u002FSKILL.md",[],[]]