[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M78SWDNT83N2MYWB8NH2D1FT":3,"skill-same:01M78SWDNT83N2MYWB8NH2D1FT":70},{"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":26,"commitSha":27,"skillDirectory":28,"forksCount":29,"repositoryHtmlUrl":30,"repositoryHomepageUrl":6,"ownerHtmlUrl":31,"ownerBlogUrl":6,"ownerDescription":6,"updatedAt":32,"install":33,"resources":45},"01M78SWDNT83N2MYWB8NH2D1FT","training",null,"Train Transformer translation models with train.py, safe command construction, data-mode choices, hyperparameters, scheduler behavior, logs, checkpoints, CPU\u002FGPU flags, TensorBoard, and troubleshooting.","VectorSpaceLab\u002FAREX-Skill","VectorSpaceLab","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F180637102?v=4","01M7538C5WPSJGWF7KHR4EFZ9F","01M7538C6R2GFP53NPXTCZNBRT",351,"MIT","https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fskills\u002Frepositories\u002Frepo-skills\u002Fattention-is-all-you-need-pytorch\u002Fsub-skills\u002Ftraining\u002FSKILL.md",[],[],"2026-11-15 14:19:17",false,"---\nname: training\ndescription: \"Train Transformer translation models with train.py, safe command\n  construction, data-mode choices, hyperparameters, scheduler behavior, logs,\n  checkpoints, CPU\u002FGPU flags, TensorBoard, and troubleshooting.\"\ndisable-model-invocation: true\nmetadata:\n  disco-role: operating\nlicense: MIT\n---\n\n# Training\n\nUse this sub-skill when the task is to plan or debug training with `train.py` without starting an expensive run. It covers the training CLI, data input modes, safe command construction, loss\u002Fscheduler smoke checks, logs, checkpoints, device flags, TensorBoard, and recoverable training setup failures.\n\nDo not use this sub-skill to create training data files, translate from a checkpoint, or explain Transformer internals beyond training-facing constructor flags. Route those tasks to the data-preparation, translation, or model-architecture sub-skills.\n\n## Fast routes\n\n- Build a command without running training: use [scripts\u002Fbuild_training_command.py](scripts\u002Fbuild_training_command.py), then confirm the details in [references\u002Fcli-reference.md](references\u002Fcli-reference.md).\n- Validate training helpers and a tiny model\u002Fscheduler path before a long run: use [scripts\u002Ftraining_smoke_check.py](scripts\u002Ftraining_smoke_check.py).\n- Choose between all-in-one pickle training and BPE-prefix training: read [references\u002Fworkflows.md](references\u002Fworkflows.md#data-mode-workflows).\n- Recover from missing `-output_dir`, stale README flags, CUDA failures, weight-sharing assertions, or TensorBoard import problems: read [references\u002Ftroubleshooting.md](references\u002Ftroubleshooting.md).\n\n## Safe operating pattern\n\n1. Decide the data mode:\n   - all-in-one pickle: `-data_pkl DATA.pkl` only;\n   - BPE files: `-data_pkl BPE_VOCAB.pkl -train_path TRAIN_PREFIX -val_path VAL_PREFIX` where `.src` and `.trg` files exist for both prefixes.\n2. Always include `-output_dir`; `train.py` raises if it is omitted.\n3. Prefer `-no_cuda` for CPU-only planning and smoke checks. Omit it only when CUDA is available and desired.\n4. Use `scripts\u002Fbuild_training_command.py` to print the exact command. It is safe by default and does not execute training.\n5. Run `scripts\u002Ftraining_smoke_check.py --repo-root \u003Ccheckout>` in an environment with the legacy training dependencies when you need to verify import, loss, model forward, and `ScheduledOptim` behavior.\n6. Treat full training as a long-running user action requiring prepared data, a writable output directory, and explicit approval.\n\n## Key runtime facts\n\n- `train.py` uses `torchtext.data.Field`, `Dataset`, `BucketIterator`, and `TranslationDataset`, so it expects the legacy torchtext API rather than modern torchtext datapipes.\n- The training loop writes `train.log` and `valid.log` inside `-output_dir`, overwriting existing files with the same names at run start.\n- With default `-save_mode best`, the checkpoint is `OUTPUT_DIR\u002Fmodel.chkpt`. With `-save_mode all`, checkpoint files are written in the process working directory, not in `-output_dir`.\n- `ScheduledOptim` wraps Adam with the Transformer warmup schedule; it updates the learning rate on every `step_and_update_lr()` call.\n- The README examples include a historical `-log` flag, but the current `train.py` parser does not accept it. Omit `-log` from generated commands.\n","# Training\n\nUse this sub-skill when the task is to plan or debug training with `train.py` without starting an expensive run. It covers the training CLI, data input modes, safe command construction, loss\u002Fscheduler smoke checks, logs, checkpoints, device flags, TensorBoard, and recoverable training setup failures.\n\nDo not use this sub-skill to create training data files, translate from a checkpoint, or explain Transformer internals beyond training-facing constructor flags. Route those tasks to the data-preparation, translation, or model-architecture sub-skills.\n\n## Fast routes\n\n- Build a command without running training: use [scripts\u002Fbuild_training_command.py](scripts\u002Fbuild_training_command.py), then confirm the details in [references\u002Fcli-reference.md](references\u002Fcli-reference.md).\n- Validate training helpers and a tiny model\u002Fscheduler path before a long run: use [scripts\u002Ftraining_smoke_check.py](scripts\u002Ftraining_smoke_check.py).\n- Choose between all-in-one pickle training and BPE-prefix training: read [references\u002Fworkflows.md](references\u002Fworkflows.md#data-mode-workflows).\n- Recover from missing `-output_dir`, stale README flags, CUDA failures, weight-sharing assertions, or TensorBoard import problems: read [references\u002Ftroubleshooting.md](references\u002Ftroubleshooting.md).\n\n## Safe operating pattern\n\n1. Decide the data mode:\n   - all-in-one pickle: `-data_pkl DATA.pkl` only;\n   - BPE files: `-data_pkl BPE_VOCAB.pkl -train_path TRAIN_PREFIX -val_path VAL_PREFIX` where `.src` and `.trg` files exist for both prefixes.\n2. Always include `-output_dir`; `train.py` raises if it is omitted.\n3. Prefer `-no_cuda` for CPU-only planning and smoke checks. Omit it only when CUDA is available and desired.\n4. Use `scripts\u002Fbuild_training_command.py` to print the exact command. It is safe by default and does not execute training.\n5. Run `scripts\u002Ftraining_smoke_check.py --repo-root \u003Ccheckout>` in an environment with the legacy training dependencies when you need to verify import, loss, model forward, and `ScheduledOptim` behavior.\n6. Treat full training as a long-running user action requiring prepared data, a writable output directory, and explicit approval.\n\n## Key runtime facts\n\n- `train.py` uses `torchtext.data.Field`, `Dataset`, `BucketIterator`, and `TranslationDataset`, so it expects the legacy torchtext API rather than modern torchtext datapipes.\n- The training loop writes `train.log` and `valid.log` inside `-output_dir`, overwriting existing files with the same names at run start.\n- With default `-save_mode best`, the checkpoint is `OUTPUT_DIR\u002Fmodel.chkpt`. With `-save_mode all`, checkpoint files are written in the process working directory, not in `-output_dir`.\n- `ScheduledOptim` wraps Adam with the Transformer warmup schedule; it updates the learning rate on every `step_and_update_lr()` call.\n- The README examples include a historical `-log` flag, but the current `train.py` parser does not accept it. Omit `-log` from generated commands.\n",{"name":5,"license":14,"metadata":23,"description":7,"disable-model-invocation":25},{"disco-role":24},"operating",true,"main","ac3fe1afa80fb9a09775ecfb2b6cc3ba850a2db6","skills\u002Frepositories\u002Frepo-skills\u002Fattention-is-all-you-need-pytorch\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":34,"prompt":44,"zipAvailable":25},[35,38,41],{"tool":36,"command":37},"npx","npx skills add VectorSpaceLab\u002FAREX-Skill --skill training",{"tool":39,"command":40},"bunx","bunx skills add VectorSpaceLab\u002FAREX-Skill --skill training",{"tool":42,"command":43},"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\u002Fattention-is-all-you-need-pytorch\u002Fsub-skills\u002Ftraining\u002FSKILL.md",[46,52,56,60,66],{"path":47,"resourceType":48,"resourceTypeDesc":49,"sizeBytes":50,"sourceHtmlUrl":51},"references\u002Fcli-reference.md","REFERENCE","参考文档",5822,"https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fskills\u002Frepositories\u002Frepo-skills\u002Fattention-is-all-you-need-pytorch\u002Fsub-skills\u002Ftraining\u002Freferences\u002Fcli-reference.md",{"path":53,"resourceType":48,"resourceTypeDesc":49,"sizeBytes":54,"sourceHtmlUrl":55},"references\u002Ftroubleshooting.md",6143,"https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fskills\u002Frepositories\u002Frepo-skills\u002Fattention-is-all-you-need-pytorch\u002Fsub-skills\u002Ftraining\u002Freferences\u002Ftroubleshooting.md",{"path":57,"resourceType":48,"resourceTypeDesc":49,"sizeBytes":58,"sourceHtmlUrl":59},"references\u002Fworkflows.md",7735,"https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fskills\u002Frepositories\u002Frepo-skills\u002Fattention-is-all-you-need-pytorch\u002Fsub-skills\u002Ftraining\u002Freferences\u002Fworkflows.md",{"path":61,"resourceType":62,"resourceTypeDesc":63,"sizeBytes":64,"sourceHtmlUrl":65},"scripts\u002Fbuild_training_command.py","SCRIPT","脚本",12302,"https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fskills\u002Frepositories\u002Frepo-skills\u002Fattention-is-all-you-need-pytorch\u002Fsub-skills\u002Ftraining\u002Fscripts\u002Fbuild_training_command.py",{"path":67,"resourceType":62,"resourceTypeDesc":63,"sizeBytes":68,"sourceHtmlUrl":69},"scripts\u002Ftraining_smoke_check.py",7970,"https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fskills\u002Frepositories\u002Frepo-skills\u002Fattention-is-all-you-need-pytorch\u002Fsub-skills\u002Ftraining\u002Fscripts\u002Ftraining_smoke_check.py",[71,79,86,93,100,107],{"skillNo":72,"name":73,"nameZh":6,"description":74,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":75,"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.","Apache-2.0","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":75,"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":75,"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":75,"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":75,"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":75,"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",[],[]]