[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M7964ME3K2T2RNEW6MKP76G1":3,"skill-same:01M7964ME3K2T2RNEW6MKP76G1":71},{"skillNo":4,"name":5,"nameZh":6,"description":7,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":15,"categories":16,"keywords":17,"publishedAt":18,"linkOnly":19,"markdownContent":20,"markdownBody":21,"frontMatter":22,"refName":27,"commitSha":28,"skillDirectory":29,"forksCount":30,"repositoryHtmlUrl":31,"repositoryHomepageUrl":6,"ownerHtmlUrl":32,"ownerBlogUrl":6,"ownerDescription":6,"updatedAt":33,"install":34,"resources":46},"01M7964ME3K2T2RNEW6MKP76G1","training",null,"Use Tensorpack ModelDesc, TrainConfig, trainers, callbacks, summaries, layers, and example training recipes.","VectorSpaceLab\u002FAREX-Skill","VectorSpaceLab","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F180637102?v=4","01M7538C5WPSJGWF7KHR4EFZ9F","01M7538C6R2GFP53NPXTCZNBRT",351,"Apache-2.0","https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fskills\u002Frepositories\u002Frepo-skills\u002Ftensorpack\u002Fsub-skills\u002Ftraining\u002FSKILL.md",[],[],"2026-11-15 14:19:17",false,"---\nname: training\ndescription: \"Use Tensorpack ModelDesc, TrainConfig, trainers, callbacks,\n  summaries, layers, and example training recipes.\"\ndisable-model-invocation: true\nmetadata:\n  disco-role: operating\nlicense: Apache 2.0\n---\n\n# Tensorpack training sub-skill\n\nUse this sub-skill when a task needs Tensorpack graph training: defining a\n`ModelDesc`, choosing a trainer, configuring `TrainConfig`, composing callbacks\nand monitors, using Tensorpack symbolic layers, adapting known training recipes,\nor debugging TF1-compatible tower training.\n\n## Start here\n\n1. If the user needs a training script skeleton, read\n   [references\u002Fworkflows.md](references\u002Fworkflows.md#minimal-modeldesc--trainconfig-quickstart)\n   and consider the bundled smoke helper\n   [scripts\u002Fminimal_training_smoke.py](scripts\u002Fminimal_training_smoke.py).\n2. If the user asks for an API name, signature, or role, read\n   [references\u002Fapi-reference.md](references\u002Fapi-reference.md).\n3. If the user asks how to adapt an example family such as MNIST, ResNet,\n   Faster\u002FMask R-CNN, GAN, Atari RL, TIMIT, Char-RNN, PTB, or Keras, read\n   [references\u002Fexample-recipes.md](references\u002Fexample-recipes.md).\n4. If the user reports an error, wrong metric, checkpoint\u002Fresume issue, GPU\n   scaling surprise, Keras incompatibility, or TF2\u002Feager-mode failure, read\n   [references\u002Ftroubleshooting.md](references\u002Ftroubleshooting.md).\n5. For deeper patterns, use\n   [references\u002Fworkflows.md](references\u002Fworkflows.md) before proposing code.\n\n## Route to this sub-skill when\n\n- The task mentions `ModelDesc`, `ModelDescBase`, `inputs()`, `build_graph()`,\n  `optimizer()`, `TrainConfig`, `AutoResumeTrainConfig`, or\n  `launch_train_with_config`.\n- The task asks how Tensorpack follows the define-then-run paradigm, how an\n  epoch\u002Fstep loop works, or how callbacks fit around training iterations.\n- The task asks which trainer to use: `SimpleTrainer`,\n  `SyncMultiGPUTrainer`, `SyncMultiGPUTrainerReplicated`,\n  `SyncMultiGPUTrainerParameterServer`, `AsyncMultiGPUTrainer`, distributed\n  parameter-server\u002Freplicated trainers, `HorovodTrainer`, or BytePS-style\n  distributed trainers.\n- The task involves tower-function rules: variable scopes, variable reuse,\n  collections, train\u002Finference towers, `TowerContext`, or multiple calls to\n  the same tower function.\n- The task involves callbacks\u002Fmonitors such as `ModelSaver`, `InferenceRunner`,\n  `ScalarStats`, `ClassificationError`, `ScheduledHyperParamSetter`,\n  `MaxSaver`, `MinSaver`, `MergeAllSummaries`, TensorBoard event files,\n  JSON histories, scalar printing, progress bars, or update ops.\n- The task involves training summaries: `tf.summary.*`,\n  `summary.add_moving_summary`, `summary.add_param_summary`, moving averages,\n  default monitors, or logging into TensorBoard\u002FJSON\u002Fstdout.\n- The task involves Tensorpack symbolic layers such as `Conv2D`,\n  `FullyConnected`, `BatchNorm`, `argscope`, `LinearWrap`, `regularize_cost`,\n  `BNReLU`, pooling layers, dropout, or caveats around `tf.layers`\u002FKeras.\n- The task asks how to adapt a Tensorpack example family for training, while\n  preserving data, dependency, and backend caveats.\n\n## Route elsewhere\n\n- For DataFlow construction, dataset layout, serializers, augmentors, input\n  performance, queue stalls, or `MultiProcessRunnerZMQ`, route to\n  [..\u002Fdataflow\u002FSKILL.md](..\u002Fdataflow\u002FSKILL.md).\n- For offline prediction, `PredictConfig`, `OfflinePredictor`, checkpoint\n  variable inspection, `.npz` model-zoo files, Serving export, compact frozen\n  graphs, or Caffe conversion, route to\n  [..\u002Finference-export\u002FSKILL.md](..\u002Finference-export\u002FSKILL.md).\n- For full benchmark\u002Fperformance reproduction, ask for the actual hardware,\n  exact dataset, cached weights, and dependency versions. Without those, only\n  provide implementation guidance and verification limits.\n\n## Operating checklist\n\n1. **Normalize the objective.** Identify whether the user wants a new training\n   script, a trainer choice, a callback\u002Fmonitor setup, an example adaptation, or\n   a debugging plan.\n2. **Select the model interface.** Prefer `ModelDesc` + `TrainConfig` for\n   single-cost gradient training. Use the raw trainer interface when callbacks\n   must be constructed after graph setup or the iteration is not single-cost.\n3. **Define inputs explicitly.** `inputs()` should return a list of\n   `tf.TensorSpec` or placeholders created inside `inputs()`. The order and\n   names must match the datapoints yielded by the input source.\n4. **Build a tower-safe graph.** `build_graph()` is a tower function: it may be\n   called once per GPU and again for inference. Use `tf.get_variable`, respect\n   the enclosing variable scope, avoid global-state mutation, and return the\n   final cost for `ModelDesc` training.\n5. **Create the optimizer once.** `optimizer()` must return a\n   `tf.train.Optimizer`. If using scheduled learning rates, create a named,\n   non-trainable variable such as `learning_rate` and target it from callbacks.\n6. **Choose the trainer by topology.** Start with `SimpleTrainer` for CPU or one\n   local GPU. Use synchronized multi-GPU trainers for data-parallel SGD, async\n   trainers for intentionally asynchronous updates, and Horovod\u002FBytePS only when\n   the launcher, dependency, and log-directory behavior are controlled.\n7. **Set `steps_per_epoch` deliberately.** Tensorpack epochs mainly schedule\n   callbacks. If input size is not known, set `TrainConfig(steps_per_epoch=...)`.\n   For multi-GPU, remember each GPU consumes one input batch, so total batch size\n   is per-tower batch times the number of towers.\n8. **Add callbacks and monitors.** Defaults already include moving summaries,\n   progress, merged summaries, update ops, TensorBoard event writing, JSON\n   history, and scalar printing. Add model saving, validation inference,\n   hyperparameter schedules, best-checkpoint savers, or custom callbacks as\n   needed.\n9. **Respect initialization\u002Fresume ownership.** Put checkpoint or `.npz` loading\n   in `TrainConfig(session_init=...)` or `AutoResumeTrainConfig`; route variable\n   inspection\u002Fexport details to inference-export.\n10. **Bound verification.** For a quick Tensorpack sanity check, run the bundled\n    helper with fake data and a temporary log directory. Do not claim benchmark\n    reproduction from the smoke helper.\n\n## Bundled smoke helper\n\nThe helper [scripts\u002Fminimal_training_smoke.py](scripts\u002Fminimal_training_smoke.py)\nis a safe, fake-data training script adapted for the generated skill tree. It\nhas no network or dataset dependency and is intended for parser\u002Fhelp checks or a\nvery short one-epoch run:\n\n```bash\npython scripts\u002Fminimal_training_smoke.py --workdir \u002Ftmp\u002Ftensorpack-smoke --steps-per-epoch 2 --max-epoch 1\n```\n\nUse it to demonstrate the shape of a Tensorpack `ModelDesc`, `TrainConfig`,\ncallbacks, summaries, and trainer launch. Do not use it as evidence that a\nreal dataset, GPU backend, Horovod, or paper benchmark is reproducible.\n\n## Quick answer patterns\n\n- **\"How do I start a model?\"** Point to the ModelDesc quickstart in\n  [references\u002Fworkflows.md](references\u002Fworkflows.md#minimal-modeldesc--trainconfig-quickstart)\n  and explain `inputs()`, `build_graph()`, and `optimizer()`.\n- **\"Which trainer for N GPUs?\"** Use the trainer matrix in\n  [references\u002Fworkflows.md](references\u002Fworkflows.md#trainer-selection) and the\n  API signatures in [references\u002Fapi-reference.md](references\u002Fapi-reference.md#trainers-and-launch).\n- **\"Why is validation\u002Fcheckpoint missing?\"** Check callback ordering and\n  monitor names in [references\u002Fworkflows.md](references\u002Fworkflows.md#callbacks-monitors-and-summaries),\n  then use [references\u002Ftroubleshooting.md](references\u002Ftroubleshooting.md#callbacks-monitors-and-validation).\n- **\"Why did multi-GPU change my epoch semantics?\"** Explain per-tower batches\n  and `steps_per_epoch`; use\n  [references\u002Ftroubleshooting.md](references\u002Ftroubleshooting.md#gpu-and-multi-gpu-semantics).\n- **\"Can I run the ResNet\u002FFasterRCNN\u002FGAN\u002FRL example?\"** Read\n  [references\u002Fexample-recipes.md](references\u002Fexample-recipes.md) and surface\n  required data, optional dependencies, backend needs, and verification status.\n\n## Known limits\n\n- This sub-skill documents GPU and distributed trainers from API\u002Fsource evidence\n  but the generated Tensorpack skill pass used CPU verification as the required\n  backend. Treat GPU, Horovod, BytePS, and large-dataset recipes as documented\n  but not smoke-verified unless the user supplies matching hardware and data.\n- Tensorpack is TF1-graph-oriented even when installed with a TF2 package. Use\n  `tensorpack.tfv1` \u002F `tf.compat.v1` style graph code and disable eager execution\n  before graph construction when needed.\n- Example-family commands are catalogs for adaptation. They are not bundled as\n  runnable dependencies and they require the stated datasets\u002Fbackends.\n","# Tensorpack training sub-skill\n\nUse this sub-skill when a task needs Tensorpack graph training: defining a\n`ModelDesc`, choosing a trainer, configuring `TrainConfig`, composing callbacks\nand monitors, using Tensorpack symbolic layers, adapting known training recipes,\nor debugging TF1-compatible tower training.\n\n## Start here\n\n1. If the user needs a training script skeleton, read\n   [references\u002Fworkflows.md](references\u002Fworkflows.md#minimal-modeldesc--trainconfig-quickstart)\n   and consider the bundled smoke helper\n   [scripts\u002Fminimal_training_smoke.py](scripts\u002Fminimal_training_smoke.py).\n2. If the user asks for an API name, signature, or role, read\n   [references\u002Fapi-reference.md](references\u002Fapi-reference.md).\n3. If the user asks how to adapt an example family such as MNIST, ResNet,\n   Faster\u002FMask R-CNN, GAN, Atari RL, TIMIT, Char-RNN, PTB, or Keras, read\n   [references\u002Fexample-recipes.md](references\u002Fexample-recipes.md).\n4. If the user reports an error, wrong metric, checkpoint\u002Fresume issue, GPU\n   scaling surprise, Keras incompatibility, or TF2\u002Feager-mode failure, read\n   [references\u002Ftroubleshooting.md](references\u002Ftroubleshooting.md).\n5. For deeper patterns, use\n   [references\u002Fworkflows.md](references\u002Fworkflows.md) before proposing code.\n\n## Route to this sub-skill when\n\n- The task mentions `ModelDesc`, `ModelDescBase`, `inputs()`, `build_graph()`,\n  `optimizer()`, `TrainConfig`, `AutoResumeTrainConfig`, or\n  `launch_train_with_config`.\n- The task asks how Tensorpack follows the define-then-run paradigm, how an\n  epoch\u002Fstep loop works, or how callbacks fit around training iterations.\n- The task asks which trainer to use: `SimpleTrainer`,\n  `SyncMultiGPUTrainer`, `SyncMultiGPUTrainerReplicated`,\n  `SyncMultiGPUTrainerParameterServer`, `AsyncMultiGPUTrainer`, distributed\n  parameter-server\u002Freplicated trainers, `HorovodTrainer`, or BytePS-style\n  distributed trainers.\n- The task involves tower-function rules: variable scopes, variable reuse,\n  collections, train\u002Finference towers, `TowerContext`, or multiple calls to\n  the same tower function.\n- The task involves callbacks\u002Fmonitors such as `ModelSaver`, `InferenceRunner`,\n  `ScalarStats`, `ClassificationError`, `ScheduledHyperParamSetter`,\n  `MaxSaver`, `MinSaver`, `MergeAllSummaries`, TensorBoard event files,\n  JSON histories, scalar printing, progress bars, or update ops.\n- The task involves training summaries: `tf.summary.*`,\n  `summary.add_moving_summary`, `summary.add_param_summary`, moving averages,\n  default monitors, or logging into TensorBoard\u002FJSON\u002Fstdout.\n- The task involves Tensorpack symbolic layers such as `Conv2D`,\n  `FullyConnected`, `BatchNorm`, `argscope`, `LinearWrap`, `regularize_cost`,\n  `BNReLU`, pooling layers, dropout, or caveats around `tf.layers`\u002FKeras.\n- The task asks how to adapt a Tensorpack example family for training, while\n  preserving data, dependency, and backend caveats.\n\n## Route elsewhere\n\n- For DataFlow construction, dataset layout, serializers, augmentors, input\n  performance, queue stalls, or `MultiProcessRunnerZMQ`, route to\n  [..\u002Fdataflow\u002FSKILL.md](..\u002Fdataflow\u002FSKILL.md).\n- For offline prediction, `PredictConfig`, `OfflinePredictor`, checkpoint\n  variable inspection, `.npz` model-zoo files, Serving export, compact frozen\n  graphs, or Caffe conversion, route to\n  [..\u002Finference-export\u002FSKILL.md](..\u002Finference-export\u002FSKILL.md).\n- For full benchmark\u002Fperformance reproduction, ask for the actual hardware,\n  exact dataset, cached weights, and dependency versions. Without those, only\n  provide implementation guidance and verification limits.\n\n## Operating checklist\n\n1. **Normalize the objective.** Identify whether the user wants a new training\n   script, a trainer choice, a callback\u002Fmonitor setup, an example adaptation, or\n   a debugging plan.\n2. **Select the model interface.** Prefer `ModelDesc` + `TrainConfig` for\n   single-cost gradient training. Use the raw trainer interface when callbacks\n   must be constructed after graph setup or the iteration is not single-cost.\n3. **Define inputs explicitly.** `inputs()` should return a list of\n   `tf.TensorSpec` or placeholders created inside `inputs()`. The order and\n   names must match the datapoints yielded by the input source.\n4. **Build a tower-safe graph.** `build_graph()` is a tower function: it may be\n   called once per GPU and again for inference. Use `tf.get_variable`, respect\n   the enclosing variable scope, avoid global-state mutation, and return the\n   final cost for `ModelDesc` training.\n5. **Create the optimizer once.** `optimizer()` must return a\n   `tf.train.Optimizer`. If using scheduled learning rates, create a named,\n   non-trainable variable such as `learning_rate` and target it from callbacks.\n6. **Choose the trainer by topology.** Start with `SimpleTrainer` for CPU or one\n   local GPU. Use synchronized multi-GPU trainers for data-parallel SGD, async\n   trainers for intentionally asynchronous updates, and Horovod\u002FBytePS only when\n   the launcher, dependency, and log-directory behavior are controlled.\n7. **Set `steps_per_epoch` deliberately.** Tensorpack epochs mainly schedule\n   callbacks. If input size is not known, set `TrainConfig(steps_per_epoch=...)`.\n   For multi-GPU, remember each GPU consumes one input batch, so total batch size\n   is per-tower batch times the number of towers.\n8. **Add callbacks and monitors.** Defaults already include moving summaries,\n   progress, merged summaries, update ops, TensorBoard event writing, JSON\n   history, and scalar printing. Add model saving, validation inference,\n   hyperparameter schedules, best-checkpoint savers, or custom callbacks as\n   needed.\n9. **Respect initialization\u002Fresume ownership.** Put checkpoint or `.npz` loading\n   in `TrainConfig(session_init=...)` or `AutoResumeTrainConfig`; route variable\n   inspection\u002Fexport details to inference-export.\n10. **Bound verification.** For a quick Tensorpack sanity check, run the bundled\n    helper with fake data and a temporary log directory. Do not claim benchmark\n    reproduction from the smoke helper.\n\n## Bundled smoke helper\n\nThe helper [scripts\u002Fminimal_training_smoke.py](scripts\u002Fminimal_training_smoke.py)\nis a safe, fake-data training script adapted for the generated skill tree. It\nhas no network or dataset dependency and is intended for parser\u002Fhelp checks or a\nvery short one-epoch run:\n\n```bash\npython scripts\u002Fminimal_training_smoke.py --workdir \u002Ftmp\u002Ftensorpack-smoke --steps-per-epoch 2 --max-epoch 1\n```\n\nUse it to demonstrate the shape of a Tensorpack `ModelDesc`, `TrainConfig`,\ncallbacks, summaries, and trainer launch. Do not use it as evidence that a\nreal dataset, GPU backend, Horovod, or paper benchmark is reproducible.\n\n## Quick answer patterns\n\n- **\"How do I start a model?\"** Point to the ModelDesc quickstart in\n  [references\u002Fworkflows.md](references\u002Fworkflows.md#minimal-modeldesc--trainconfig-quickstart)\n  and explain `inputs()`, `build_graph()`, and `optimizer()`.\n- **\"Which trainer for N GPUs?\"** Use the trainer matrix in\n  [references\u002Fworkflows.md](references\u002Fworkflows.md#trainer-selection) and the\n  API signatures in [references\u002Fapi-reference.md](references\u002Fapi-reference.md#trainers-and-launch).\n- **\"Why is validation\u002Fcheckpoint missing?\"** Check callback ordering and\n  monitor names in [references\u002Fworkflows.md](references\u002Fworkflows.md#callbacks-monitors-and-summaries),\n  then use [references\u002Ftroubleshooting.md](references\u002Ftroubleshooting.md#callbacks-monitors-and-validation).\n- **\"Why did multi-GPU change my epoch semantics?\"** Explain per-tower batches\n  and `steps_per_epoch`; use\n  [references\u002Ftroubleshooting.md](references\u002Ftroubleshooting.md#gpu-and-multi-gpu-semantics).\n- **\"Can I run the ResNet\u002FFasterRCNN\u002FGAN\u002FRL example?\"** Read\n  [references\u002Fexample-recipes.md](references\u002Fexample-recipes.md) and surface\n  required data, optional dependencies, backend needs, and verification status.\n\n## Known limits\n\n- This sub-skill documents GPU and distributed trainers from API\u002Fsource evidence\n  but the generated Tensorpack skill pass used CPU verification as the required\n  backend. Treat GPU, Horovod, BytePS, and large-dataset recipes as documented\n  but not smoke-verified unless the user supplies matching hardware and data.\n- Tensorpack is TF1-graph-oriented even when installed with a TF2 package. Use\n  `tensorpack.tfv1` \u002F `tf.compat.v1` style graph code and disable eager execution\n  before graph construction when needed.\n- Example-family commands are catalogs for adaptation. They are not bundled as\n  runnable dependencies and they require the stated datasets\u002Fbackends.\n",{"name":5,"license":23,"metadata":24,"description":7,"disable-model-invocation":26},"Apache 2.0",{"disco-role":25},"operating",true,"main","ac3fe1afa80fb9a09775ecfb2b6cc3ba850a2db6","skills\u002Frepositories\u002Frepo-skills\u002Ftensorpack\u002Fsub-skills\u002Ftraining",27,"https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill","https:\u002F\u002Fgithub.com\u002FVectorSpaceLab","2026-09-03 10:25:21",{"source":8,"skillName":5,"commands":35,"prompt":45,"zipAvailable":26},[36,39,42],{"tool":37,"command":38},"npx","npx skills add VectorSpaceLab\u002FAREX-Skill --skill training",{"tool":40,"command":41},"bunx","bunx skills add VectorSpaceLab\u002FAREX-Skill --skill training",{"tool":43,"command":44},"pnpm","pnpm dlx skills add VectorSpaceLab\u002FAREX-Skill --skill training","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fskills\u002Frepositories\u002Frepo-skills\u002Ftensorpack\u002Fsub-skills\u002Ftraining\u002FSKILL.md",[47,53,57,61,65],{"path":48,"resourceType":49,"resourceTypeDesc":50,"sizeBytes":51,"sourceHtmlUrl":52},"references\u002Fapi-reference.md","REFERENCE","参考文档",15488,"https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fskills\u002Frepositories\u002Frepo-skills\u002Ftensorpack\u002Fsub-skills\u002Ftraining\u002Freferences\u002Fapi-reference.md",{"path":54,"resourceType":49,"resourceTypeDesc":50,"sizeBytes":55,"sourceHtmlUrl":56},"references\u002Fexample-recipes.md",17122,"https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fskills\u002Frepositories\u002Frepo-skills\u002Ftensorpack\u002Fsub-skills\u002Ftraining\u002Freferences\u002Fexample-recipes.md",{"path":58,"resourceType":49,"resourceTypeDesc":50,"sizeBytes":59,"sourceHtmlUrl":60},"references\u002Ftroubleshooting.md",14038,"https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fskills\u002Frepositories\u002Frepo-skills\u002Ftensorpack\u002Fsub-skills\u002Ftraining\u002Freferences\u002Ftroubleshooting.md",{"path":62,"resourceType":49,"resourceTypeDesc":50,"sizeBytes":63,"sourceHtmlUrl":64},"references\u002Fworkflows.md",15710,"https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fskills\u002Frepositories\u002Frepo-skills\u002Ftensorpack\u002Fsub-skills\u002Ftraining\u002Freferences\u002Fworkflows.md",{"path":66,"resourceType":67,"resourceTypeDesc":68,"sizeBytes":69,"sourceHtmlUrl":70},"scripts\u002Fminimal_training_smoke.py","SCRIPT","脚本",5430,"https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fskills\u002Frepositories\u002Frepo-skills\u002Ftensorpack\u002Fsub-skills\u002Ftraining\u002Fscripts\u002Fminimal_training_smoke.py",[72,79,86,93,100,107],{"skillNo":73,"name":74,"nameZh":6,"description":75,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":76,"categories":77,"keywords":78,"publishedAt":18},"01M799NYSH1R3RVAHSM2D88Y84","passnet-triton-opt","Make a single PassNet Triton kernel fast AND numerically faithful: performance model (when a replacement can win at all), block\u002Fgrid\u002Fwarp tuning, autotune policy, launch overhead, and per-op numeric recipes to pass the dtype baseline tolerances. Use when a pass matches and is correct but speedup below expected, or when correctness fails by small numeric margins.","https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fskills\u002Ftask-oriented\u002FPassNet\u002Fskills\u002Fpassnet-triton-opt\u002FSKILL.md",[],[],{"skillNo":80,"name":81,"nameZh":6,"description":82,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":83,"categories":84,"keywords":85,"publishedAt":18},"01M799NS5APEC0N1FZ7XZVW6BN","passnet-solve","END-TO-END playbook for solving one PassNet sample: analyze the computation graphs, decide the optimization strategy (what to fuse, what to replace, what to leave alone), drive the iteration loop, and maximize the sample score. This is the ENTRY skill — invoke it first for any PassNet optimization task; it tells you when to use passnet-pattern-fusion, passnet-triton-opt and passnet-feedback.","https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fskills\u002Ftask-oriented\u002FPassNet\u002Fskills\u002Fpassnet-solve\u002FSKILL.md",[],[],{"skillNo":87,"name":88,"nameZh":6,"description":89,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":90,"categories":91,"keywords":92,"publishedAt":18},"01M799NNY9BKSM21V5R7QVTBJX","passnet-skill","PassNet GPU kernel optimization via compiler passes. Design and implement Triton-based optimization passes, create pass files under .\u002Fpass_dir\u002F, self-evaluate with pass_evaluator, and iterate to maximize GPU speedup.","https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fskills\u002Ftask-oriented\u002FPassNet\u002Fskills\u002Fpassnet-skill\u002FSKILL.md",[],[],{"skillNo":94,"name":95,"nameZh":6,"description":96,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":97,"categories":98,"keywords":99,"publishedAt":18},"01M799NHH6NP7Z1RK77KGXAN6S","passnet-pattern-fusion","Author PassNet pass files: write patterns that actually MATCH the FX graph, pick fusion regions (which ops to absorb into one Triton kernel), and structure multi-pass submissions with the mandatory shared-dispatch architecture. Use when creating or fixing pattern()\u002Freplacement_args()\u002Freplacement_func() files, when a pass \"failed to match\", or when deciding how to fuse multiple kernels\u002Fops.","https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fskills\u002Ftask-oriented\u002FPassNet\u002Fskills\u002Fpassnet-pattern-fusion\u002FSKILL.md",[],[],{"skillNo":101,"name":102,"nameZh":6,"description":103,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":104,"categories":105,"keywords":106,"publishedAt":18},"01M799ND2J4WNC2PGAZ0A06NDF","passnet-orchestrate","Round\u002Fbatch planning and decision support for PassNet graph optimization. Use for multi-sample triage, worker allocation, eval-budget policy, and post-eval keep\u002Frevert decisions. For solving one concrete sample end to end, use passnet-solve as the entry skill; consult this skill only when you need broader planning or a specific decision gate.","https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fskills\u002Ftask-oriented\u002FPassNet\u002Fskills\u002Fpassnet-orchestrate\u002FSKILL.md",[],[],{"skillNo":108,"name":109,"nameZh":6,"description":110,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":111,"categories":112,"keywords":113,"publishedAt":18},"01M799N8SBTMY02ZDG2AZF5ZWE","passnet-feedback","Fast, deterministic feedback for PassNet work: pre-flight pattern\u002Fmatch verification WITHOUT burning a GPU evaluation, per-node bottleneck analysis of a sample's graphs, and eval-log parsing into per-variant status + estimated score + failure classification. Use BEFORE every GPU evaluation (check_pattern), at the START of a sample (analyze_graph), and AFTER every evaluation (parse_eval_log).","https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fskills\u002Ftask-oriented\u002FPassNet\u002Fskills\u002Fpassnet-feedback\u002FSKILL.md",[],[]]