[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M7FDGVCWX1RGSYKVDK2CDHVC":3,"skill-same:01M7FDGVCWX1RGSYKVDK2CDHVC":63},{"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":28,"commitSha":29,"skillDirectory":30,"forksCount":31,"repositoryHtmlUrl":32,"repositoryHomepageUrl":6,"ownerHtmlUrl":33,"ownerBlogUrl":6,"ownerDescription":34,"updatedAt":35,"install":36,"resources":48},"01M7FDGVCWX1RGSYKVDK2CDHVC","training-and-model",null,"Train, fine-tune, and understand zi2zi's TensorFlow 1.x conditional adversarial U-Net, experiment layout, losses, label shuffling, and checkpoint monitoring.","lza6\u002FArmor-piercing-system-prompts","lza6","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F128880206?v=4","01M7EB3K7WPX5T28EJ2NVHE6EG","Chinese-tingfeng","01M7EB3K97E1KQ16QWKJMFNGX8",11,"Apache-2.0","https:\u002F\u002Fgithub.com\u002Flza6\u002FArmor-piercing-system-prompts\u002Fblob\u002Fmain\u002F06-前端与设计\u002Fzi2zi\u002Fsub-skills\u002Ftraining-and-model\u002FSKILL.md",[],[],"2026-11-16 10:05:34",false,"---\nname: training-and-model\ndescription: \"Train, fine-tune, and understand zi2zi's TensorFlow 1.x\n  conditional adversarial U-Net, experiment layout, losses, label shuffling, and\n  checkpoint monitoring.\"\nmetadata:\n  disco-role: operating\ndisable-model-invocation: true\nlicense: Apache 2.0\n---\n\n# zi2zi training and model\n\nUse this sub-skill when a task is about zi2zi training, fine-tuning, model\nstructure, loss terms, checkpoint management, or experiment diagnostics. It does\nnot run long training automatically; it explains the legacy workflow and gives\nsafe command planning and preflight checks.\n\n## Use this sub-skill when\n\n- The user wants to train zi2zi from `train.obj` and `val.obj`.\n- The user wants to fine-tune only a subset of labels or freeze the encoder.\n- The task mentions `train.py`, `experiment_dir`, `checkpoint`, `sample`,\n  `logs`, `flip_labels`, `fine_tune`, `inst_norm`, `freeze_encoder`, or loss\n  names such as L1, constant, category, or total variation.\n- The user needs to understand why `d_loss` saturates, why label shuffling was\n  added, or how checkpoints and TensorBoard outputs are organized.\n\n## Route elsewhere\n\n- Data rendering and `.obj` packaging go to\n  [data-preparation](..\u002Fdata-preparation\u002FSKILL.md).\n- Inference, interpolation, and export go to\n  [inference-and-export](..\u002Finference-and-export\u002FSKILL.md).\n- Cross-cutting compatibility issues go to the root\n  [compatibility](..\u002F..\u002Freferences\u002Fcompatibility.md) and\n  [troubleshooting](..\u002F..\u002Freferences\u002Ftroubleshooting.md) references.\n\n## Core training model facts\n\n- `UNet` is a pix2pix-like encoder\u002Fdecoder generator with skip connections and\n  a discriminator that predicts real\u002Ffake and style category.\n- The model combines several losses: adversarial cheat loss, L1 loss, category\n  loss, constant\u002Fencoding loss, and optional total variation loss.\n- `flip_labels=1` enables the no-target-source branch used for label shuffling\n  and saturation recovery.\n- `freeze_encoder=1` excludes encoder variables from generator updates.\n- `fine_tune` accepts a comma-separated list of integer labels, which the data\n  provider uses to filter training and validation records.\n- `inst_norm=1` switches decoder normalization to conditional instance\n  normalization instead of batch normalization.\n\n## Standard experiment layout\n\nThe README and source expect an experiment tree like this:\n\n```text\nexperiment\u002F\n  data\u002F\n    train.obj\n    val.obj\n  checkpoint\u002F\n  logs\u002F\n  sample\u002F\n```\n\n`train.py` creates the checkpoint, log, and sample directories if they do not\nexist. Checkpoints are saved under a batch-size-specific model directory named\n`experiment_\u003Cid>_batch_\u003Cbatch>`.\n\n## Recommended workflow\n\n1. Confirm `train.obj` and `val.obj` are present and non-empty.\n2. Choose `--embedding_num` large enough for the highest label in the data.\n3. Decide whether you want batch norm or conditional instance norm.\n4. For a normal full run, set batch size, learning rate, epoch count, and sample\n  \u002Fcheckpoint cadence deliberately rather than accepting defaults blindly.\n5. Use `--resume=1` only when you want to continue an existing experiment.\n6. Use `--fine_tune` for label-subset adaptation and `--flip_labels=1` when the\n   discriminator saturates during later fine-tuning.\n7. Watch TensorBoard logs and the `sample\u002F` images for collapse or label issues.\n\nRead [training-workflow.md](references\u002Ftraining-workflow.md) for command shapes,\npreflight checks, and safe run-planning. Read [model-architecture.md](references\u002Fmodel-architecture.md)\nfor architecture and loss details. Read [troubleshooting.md](references\u002Ftroubleshooting.md)\nfor TensorFlow, checkpoint, OOM, and data-layout failures.\n\n## Bundled helper\n\n- [scripts\u002Fplan_zi2zi_training.py](scripts\u002Fplan_zi2zi_training.py) validates a\n  training command and prints the intended legacy zi2zi invocation without\n  launching the TensorFlow loop.\n\n## What not to promise automatically\n\n- Do not promise a successful long training run without a compatible legacy\n  TensorFlow environment, valid `.obj` files, and explicit runtime budget.\n- Do not claim GPU verification from a CPU-only graph build.\n- Do not infer checkpoint compatibility across different batch sizes,\n  embedding sizes, or normalization settings without checking the variable\n  names and shapes.\n","# zi2zi training and model\n\nUse this sub-skill when a task is about zi2zi training, fine-tuning, model\nstructure, loss terms, checkpoint management, or experiment diagnostics. It does\nnot run long training automatically; it explains the legacy workflow and gives\nsafe command planning and preflight checks.\n\n## Use this sub-skill when\n\n- The user wants to train zi2zi from `train.obj` and `val.obj`.\n- The user wants to fine-tune only a subset of labels or freeze the encoder.\n- The task mentions `train.py`, `experiment_dir`, `checkpoint`, `sample`,\n  `logs`, `flip_labels`, `fine_tune`, `inst_norm`, `freeze_encoder`, or loss\n  names such as L1, constant, category, or total variation.\n- The user needs to understand why `d_loss` saturates, why label shuffling was\n  added, or how checkpoints and TensorBoard outputs are organized.\n\n## Route elsewhere\n\n- Data rendering and `.obj` packaging go to\n  [data-preparation](..\u002Fdata-preparation\u002FSKILL.md).\n- Inference, interpolation, and export go to\n  [inference-and-export](..\u002Finference-and-export\u002FSKILL.md).\n- Cross-cutting compatibility issues go to the root\n  [compatibility](..\u002F..\u002Freferences\u002Fcompatibility.md) and\n  [troubleshooting](..\u002F..\u002Freferences\u002Ftroubleshooting.md) references.\n\n## Core training model facts\n\n- `UNet` is a pix2pix-like encoder\u002Fdecoder generator with skip connections and\n  a discriminator that predicts real\u002Ffake and style category.\n- The model combines several losses: adversarial cheat loss, L1 loss, category\n  loss, constant\u002Fencoding loss, and optional total variation loss.\n- `flip_labels=1` enables the no-target-source branch used for label shuffling\n  and saturation recovery.\n- `freeze_encoder=1` excludes encoder variables from generator updates.\n- `fine_tune` accepts a comma-separated list of integer labels, which the data\n  provider uses to filter training and validation records.\n- `inst_norm=1` switches decoder normalization to conditional instance\n  normalization instead of batch normalization.\n\n## Standard experiment layout\n\nThe README and source expect an experiment tree like this:\n\n```text\nexperiment\u002F\n  data\u002F\n    train.obj\n    val.obj\n  checkpoint\u002F\n  logs\u002F\n  sample\u002F\n```\n\n`train.py` creates the checkpoint, log, and sample directories if they do not\nexist. Checkpoints are saved under a batch-size-specific model directory named\n`experiment_\u003Cid>_batch_\u003Cbatch>`.\n\n## Recommended workflow\n\n1. Confirm `train.obj` and `val.obj` are present and non-empty.\n2. Choose `--embedding_num` large enough for the highest label in the data.\n3. Decide whether you want batch norm or conditional instance norm.\n4. For a normal full run, set batch size, learning rate, epoch count, and sample\n  \u002Fcheckpoint cadence deliberately rather than accepting defaults blindly.\n5. Use `--resume=1` only when you want to continue an existing experiment.\n6. Use `--fine_tune` for label-subset adaptation and `--flip_labels=1` when the\n   discriminator saturates during later fine-tuning.\n7. Watch TensorBoard logs and the `sample\u002F` images for collapse or label issues.\n\nRead [training-workflow.md](references\u002Ftraining-workflow.md) for command shapes,\npreflight checks, and safe run-planning. Read [model-architecture.md](references\u002Fmodel-architecture.md)\nfor architecture and loss details. Read [troubleshooting.md](references\u002Ftroubleshooting.md)\nfor TensorFlow, checkpoint, OOM, and data-layout failures.\n\n## Bundled helper\n\n- [scripts\u002Fplan_zi2zi_training.py](scripts\u002Fplan_zi2zi_training.py) validates a\n  training command and prints the intended legacy zi2zi invocation without\n  launching the TensorFlow loop.\n\n## What not to promise automatically\n\n- Do not promise a successful long training run without a compatible legacy\n  TensorFlow environment, valid `.obj` files, and explicit runtime budget.\n- Do not claim GPU verification from a CPU-only graph build.\n- Do not infer checkpoint compatibility across different batch sizes,\n  embedding sizes, or normalization settings without checking the variable\n  names and shapes.\n",{"name":5,"license":24,"metadata":25,"description":7,"disable-model-invocation":27},"Apache 2.0",{"disco-role":26},"operating",true,"main","d69096e239f2a89aabc2cd078c47664a9d3b1e11","06-前端与设计\u002Fzi2zi\u002Fsub-skills\u002Ftraining-and-model",3,"https:\u002F\u002Fgithub.com\u002Flza6\u002FArmor-piercing-system-prompts","https:\u002F\u002Fgithub.com\u002Flza6","个人vx：Tf00798\r\n个人✈：@TFGY999","2026-10-08 14:29:22",{"source":8,"skillName":5,"commands":37,"prompt":47,"zipAvailable":27},[38,41,44],{"tool":39,"command":40},"npx","npx skills add lza6\u002FArmor-piercing-system-prompts --skill training-and-model",{"tool":42,"command":43},"bunx","bunx skills add lza6\u002FArmor-piercing-system-prompts --skill training-and-model",{"tool":45,"command":46},"pnpm","pnpm dlx skills add lza6\u002FArmor-piercing-system-prompts --skill training-and-model","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002Flza6\u002FArmor-piercing-system-prompts\u002Fblob\u002Fmain\u002F06-前端与设计\u002Fzi2zi\u002Fsub-skills\u002Ftraining-and-model\u002FSKILL.md",[49,55,59],{"path":50,"resourceType":51,"resourceTypeDesc":52,"sizeBytes":53,"sourceHtmlUrl":54},"references\u002Fmodel-architecture.md","REFERENCE","参考文档",2648,"https:\u002F\u002Fgithub.com\u002Flza6\u002FArmor-piercing-system-prompts\u002Fblob\u002Fmain\u002F06-前端与设计\u002Fzi2zi\u002Fsub-skills\u002Ftraining-and-model\u002Freferences\u002Fmodel-architecture.md",{"path":56,"resourceType":51,"resourceTypeDesc":52,"sizeBytes":57,"sourceHtmlUrl":58},"references\u002Ftraining-workflow.md",2869,"https:\u002F\u002Fgithub.com\u002Flza6\u002FArmor-piercing-system-prompts\u002Fblob\u002Fmain\u002F06-前端与设计\u002Fzi2zi\u002Fsub-skills\u002Ftraining-and-model\u002Freferences\u002Ftraining-workflow.md",{"path":60,"resourceType":51,"resourceTypeDesc":52,"sizeBytes":61,"sourceHtmlUrl":62},"references\u002Ftroubleshooting.md",2308,"https:\u002F\u002Fgithub.com\u002Flza6\u002FArmor-piercing-system-prompts\u002Fblob\u002Fmain\u002F06-前端与设计\u002Fzi2zi\u002Fsub-skills\u002Ftraining-and-model\u002Freferences\u002Ftroubleshooting.md",[64,72,77,84,91,98],{"skillNo":65,"name":66,"nameZh":6,"description":67,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":68,"licenseFileUrl":6,"sourceHtmlUrl":69,"categories":70,"keywords":71,"publishedAt":19},"01M7FFG485KY83TD6HSC4YPCMC","obsidian-clipper-template-creator","Guide for creating templates for the Obsidian Web Clipper. Use when you want to create a new clipping template, understand available variables, or format clipped content.","MIT","https:\u002F\u002Fgithub.com\u002Flza6\u002FArmor-piercing-system-prompts\u002Fblob\u002Fmain\u002F07-后端与架构\u002Fobsidian-clipper-template-creator\u002FSKILL.md",[],[],{"skillNo":73,"name":66,"nameZh":6,"description":67,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":68,"licenseFileUrl":6,"sourceHtmlUrl":74,"categories":75,"keywords":76,"publishedAt":19},"01M7FFG0T6AZ0STEJPSB3CX84K","https:\u002F\u002Fgithub.com\u002Flza6\u002FArmor-piercing-system-prompts\u002Fblob\u002Fmain\u002F07-后端与架构\u002Fobsidian-clipper-template-creator-antigravity-awesome-skills-main\u002FSKILL.md",[],[],{"skillNo":78,"name":79,"nameZh":6,"description":80,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":68,"licenseFileUrl":6,"sourceHtmlUrl":81,"categories":82,"keywords":83,"publishedAt":19},"01M7FFFY0QEM14SKWPDYGMXM2P","python-api","Programmatic Object-Detection-Metrics API guidance for BoundingBox, BoundingBoxes, Evaluator, enums, coordinate conversion, AP\u002FmAP semantics, and source-style import diagnostics.","https:\u002F\u002Fgithub.com\u002Flza6\u002FArmor-piercing-system-prompts\u002Fblob\u002Fmain\u002F07-后端与架构\u002Fobject-detection-metrics\u002Fsub-skills\u002Fpython-api\u002FSKILL.md",[],[],{"skillNo":85,"name":86,"nameZh":6,"description":87,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":68,"licenseFileUrl":6,"sourceHtmlUrl":88,"categories":89,"keywords":90,"publishedAt":19},"01M7FFFVZQZ2GJ0C0TGP7JHD20","file-evaluation","Folder-based PASCAL VOC AP\u002FmAP evaluation from ground-truth and detection text files.","https:\u002F\u002Fgithub.com\u002Flza6\u002FArmor-piercing-system-prompts\u002Fblob\u002Fmain\u002F07-后端与架构\u002Fobject-detection-metrics\u002Fsub-skills\u002Ffile-evaluation\u002FSKILL.md",[],[],{"skillNo":92,"name":93,"nameZh":6,"description":94,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":68,"licenseFileUrl":6,"sourceHtmlUrl":95,"categories":96,"keywords":97,"publishedAt":19},"01M7FFFSFNYSE77SJ6PE1CK2DA","object-detection-metrics","Guides legacy Object-Detection-Metrics PASCAL VOC AP\u002FmAP evaluation from detection text folders and direct Python API classes.","https:\u002F\u002Fgithub.com\u002Flza6\u002FArmor-piercing-system-prompts\u002Fblob\u002Fmain\u002F07-后端与架构\u002Fobject-detection-metrics\u002FSKILL.md",[],[],{"skillNo":99,"name":100,"nameZh":6,"description":101,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":68,"licenseFileUrl":6,"sourceHtmlUrl":102,"categories":103,"keywords":104,"publishedAt":19},"01M7FFFP0GCK1RVXBCPQ8APBMT","nx-workspace-patterns","Configure and optimize Nx monorepo workspaces. Use when setting up Nx, configuring project boundaries, optimizing build caching, or implementing affected commands.","https:\u002F\u002Fgithub.com\u002Flza6\u002FArmor-piercing-system-prompts\u002Fblob\u002Fmain\u002F07-后端与架构\u002Fnx-workspace-patterns\u002FSKILL.md",[],[]]