[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M795FJADCJ1H53B2GY747VTC":3,"skill-same:01M795FJADCJ1H53B2GY747VTC":59},{"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},"01M795FJADCJ1H53B2GY747VTC","training",null,"Route for SSD Keras model building, compilation, training, and fine-tuning.","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\u002Fssd-keras\u002Fsub-skills\u002Ftraining\u002FSKILL.md",[],[],"2026-11-15 14:19:17",false,"---\nname: training\ndescription: \"Route for SSD Keras model building, compilation, training, and fine-tuning.\"\nmetadata:\n  disco-role: operating\ndisable-model-invocation: true\nlicense: Apache 2.0\n---\n\n# Training\n\nUse this route when the task is about building a trainable SSD model, loading pretrained weights, compiling the loss, wiring the encoder, or running a fitting loop.\n\n## What this route covers\n\n- building SSD300, SSD512, or SSD7-style trainable models\n- loading pretrained VGG or SSD weights by name\n- compiling with `SSDLoss`\n- creating `SSDInputEncoder` instances that match the model anchors\n- fitting on VOC, CSV, or synthetic data batches\n- adapting classifier weights when the class count changes\n- using the SSD7 template as a starting point for alternate backbones\n\n## What this route excludes\n\n- dataset parsing details and augmentation internals\n- decoding predictions\n- VOC \u002F COCO evaluation\n\nSend those tasks to `data-preparation` or `inference-evaluation`.\n\n## First things to check\n\n1. Open `references\u002Fmodel-architecture.md` to confirm the model \u002F encoder \u002F loss contract.\n2. Open `references\u002Fworkflows.md` for the notebook-derived training sequence.\n3. Open `references\u002Fcompatibility.md` if the old TensorFlow \u002F Keras stack is not importing.\n4. Run `scripts\u002Fsmoke.py` for a tiny synthetic model \u002F encoder \u002F train-step check.\n\n## Typical training workflow\n\n### 1. Pick the model family\n\n- `ssd_300` is the canonical path and matches the main Pascal VOC \u002F COCO notebook recipes.\n- `ssd_512` uses the same design at a larger input size.\n- `build_model` is the smaller SSD7 template and is the best smoke path or custom-backbone starting point.\n\n### 2. Match the anchor settings\n\nThe model builder and the encoder must agree on:\n\n- scales\n- aspect ratios\n- steps\n- offsets\n- variances\n- coordinate convention\n- coordinate normalization\n\nIf these drift apart, the model will build but the targets or decoded boxes will be wrong.\n\n### 3. Load pretrained weights carefully\n\n- Use `load_weights(..., by_name=True)` when the source weights were saved with a different head shape.\n- If the source and target class counts differ, adapt the classifier weights with `sample_tensors` before loading them.\n- The weight-sampling tutorial is the right reference when you are reusing a pretrained detector with a new class list.\n\n### 4. Compile and fit\n\n- Use `SSDLoss.compute_loss` as the model loss.\n- Keep the optimizer settings consistent with the notebook recipe when you need to reproduce the historical results.\n- Feed the model from `DataGenerator` through the data-preparation route.\n- Keep an eye on early training instability: `OOM` and `NaN` are the two common failure modes.\n\n### 5. Save and resume\n\n- Save the whole model when you need to reload custom layers later.\n- If you resume training, keep the loss, encoder, and model config exactly aligned with the original run.\n\n## Fine-tuning \u002F transfer learning\n\nWhen the class set changes, do not edit the pretrained weights by hand.\n\nInstead:\n\n1. Inspect the source classifier kernel and bias shapes.\n2. Select the classes to keep.\n3. Use `sample_tensors` to sub-sample or up-sample the classifier tensors consistently.\n4. Load the adapted weights into the new model head.\n\n## Useful source objects\n\n- `ssd_300`\n- `ssd_512`\n- `build_model`\n- `SSDLoss`\n- `SSDInputEncoder`\n- `sample_tensors`\n- `AnchorBoxes`\n- `L2Normalization`\n- `DecodeDetections`\n- `DecodeDetectionsFast`\n\n## Script path\n\n- `scripts\u002Fsmoke.py` — builds a small model, encodes a synthetic batch, runs one training step, and exercises the weight-sampling utility.\n\n## Quick decision guide\n\n- Need the canonical VOC \u002F COCO recipe? Start with `ssd_300`.\n- Need a faster smoke path or a custom backbone template? Start with `build_model`.\n- Need to change the class count on pretrained weights? Stay in this route and use `sample_tensors` before loading the new head.\n","# Training\n\nUse this route when the task is about building a trainable SSD model, loading pretrained weights, compiling the loss, wiring the encoder, or running a fitting loop.\n\n## What this route covers\n\n- building SSD300, SSD512, or SSD7-style trainable models\n- loading pretrained VGG or SSD weights by name\n- compiling with `SSDLoss`\n- creating `SSDInputEncoder` instances that match the model anchors\n- fitting on VOC, CSV, or synthetic data batches\n- adapting classifier weights when the class count changes\n- using the SSD7 template as a starting point for alternate backbones\n\n## What this route excludes\n\n- dataset parsing details and augmentation internals\n- decoding predictions\n- VOC \u002F COCO evaluation\n\nSend those tasks to `data-preparation` or `inference-evaluation`.\n\n## First things to check\n\n1. Open `references\u002Fmodel-architecture.md` to confirm the model \u002F encoder \u002F loss contract.\n2. Open `references\u002Fworkflows.md` for the notebook-derived training sequence.\n3. Open `references\u002Fcompatibility.md` if the old TensorFlow \u002F Keras stack is not importing.\n4. Run `scripts\u002Fsmoke.py` for a tiny synthetic model \u002F encoder \u002F train-step check.\n\n## Typical training workflow\n\n### 1. Pick the model family\n\n- `ssd_300` is the canonical path and matches the main Pascal VOC \u002F COCO notebook recipes.\n- `ssd_512` uses the same design at a larger input size.\n- `build_model` is the smaller SSD7 template and is the best smoke path or custom-backbone starting point.\n\n### 2. Match the anchor settings\n\nThe model builder and the encoder must agree on:\n\n- scales\n- aspect ratios\n- steps\n- offsets\n- variances\n- coordinate convention\n- coordinate normalization\n\nIf these drift apart, the model will build but the targets or decoded boxes will be wrong.\n\n### 3. Load pretrained weights carefully\n\n- Use `load_weights(..., by_name=True)` when the source weights were saved with a different head shape.\n- If the source and target class counts differ, adapt the classifier weights with `sample_tensors` before loading them.\n- The weight-sampling tutorial is the right reference when you are reusing a pretrained detector with a new class list.\n\n### 4. Compile and fit\n\n- Use `SSDLoss.compute_loss` as the model loss.\n- Keep the optimizer settings consistent with the notebook recipe when you need to reproduce the historical results.\n- Feed the model from `DataGenerator` through the data-preparation route.\n- Keep an eye on early training instability: `OOM` and `NaN` are the two common failure modes.\n\n### 5. Save and resume\n\n- Save the whole model when you need to reload custom layers later.\n- If you resume training, keep the loss, encoder, and model config exactly aligned with the original run.\n\n## Fine-tuning \u002F transfer learning\n\nWhen the class set changes, do not edit the pretrained weights by hand.\n\nInstead:\n\n1. Inspect the source classifier kernel and bias shapes.\n2. Select the classes to keep.\n3. Use `sample_tensors` to sub-sample or up-sample the classifier tensors consistently.\n4. Load the adapted weights into the new model head.\n\n## Useful source objects\n\n- `ssd_300`\n- `ssd_512`\n- `build_model`\n- `SSDLoss`\n- `SSDInputEncoder`\n- `sample_tensors`\n- `AnchorBoxes`\n- `L2Normalization`\n- `DecodeDetections`\n- `DecodeDetectionsFast`\n\n## Script path\n\n- `scripts\u002Fsmoke.py` — builds a small model, encodes a synthetic batch, runs one training step, and exercises the weight-sampling utility.\n\n## Quick decision guide\n\n- Need the canonical VOC \u002F COCO recipe? Start with `ssd_300`.\n- Need a faster smoke path or a custom backbone template? Start with `build_model`.\n- Need to change the class count on pretrained weights? Stay in this route and use `sample_tensors` before loading the new head.\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\u002Fssd-keras\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\u002Fssd-keras\u002Fsub-skills\u002Ftraining\u002FSKILL.md",[47,53],{"path":48,"resourceType":49,"resourceTypeDesc":50,"sizeBytes":51,"sourceHtmlUrl":52},"references\u002Ftroubleshooting.md","REFERENCE","参考文档",2025,"https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fskills\u002Frepositories\u002Frepo-skills\u002Fssd-keras\u002Fsub-skills\u002Ftraining\u002Freferences\u002Ftroubleshooting.md",{"path":54,"resourceType":55,"resourceTypeDesc":56,"sizeBytes":57,"sourceHtmlUrl":58},"scripts\u002Fsmoke.py","SCRIPT","脚本",5060,"https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fskills\u002Frepositories\u002Frepo-skills\u002Fssd-keras\u002Fsub-skills\u002Ftraining\u002Fscripts\u002Fsmoke.py",[60,67,74,81,88,95],{"skillNo":61,"name":62,"nameZh":6,"description":63,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":64,"categories":65,"keywords":66,"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":68,"name":69,"nameZh":6,"description":70,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":71,"categories":72,"keywords":73,"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":75,"name":76,"nameZh":6,"description":77,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":78,"categories":79,"keywords":80,"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":82,"name":83,"nameZh":6,"description":84,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":85,"categories":86,"keywords":87,"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":89,"name":90,"nameZh":6,"description":91,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":92,"categories":93,"keywords":94,"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":96,"name":97,"nameZh":6,"description":98,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":99,"categories":100,"keywords":101,"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",[],[]]