[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M796MKBTHY2CS9KDQQFJ2432":3,"skill-same:01M796MKBTHY2CS9KDQQFJ2432":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},"01M796MKBTHY2CS9KDQQFJ2432","serving-cli",null,"Use the Transformers CLI for download, chat, local serving, OpenAI-compatible endpoints, server options, model warmup, continuous batching flags, CORS, reasoning, and safe preflight checks.","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\u002Ftransformers\u002Fsub-skills\u002Fserving-cli\u002FSKILL.md",[],[],"2026-11-15 14:19:17",false,"---\nname: serving-cli\ndescription: \"Use the Transformers CLI for download, chat, local serving,\n  OpenAI-compatible endpoints, server options, model warmup, continuous batching\n  flags, CORS, reasoning, and safe preflight checks.\"\ndisable-model-invocation: true\nmetadata:\n  disco-role: operating\nlicense: Apache 2.0\n---\n\n# Serving CLI\n\nUse this sub-skill when an agent needs `transformers` command-line workflows: `download`, `chat`, `serve`, local OpenAI-compatible HTTP APIs, model warmup, server options, CORS, reasoning flags, or a safe preflight before starting long-lived services.\n\n## Route First\n\n- Use this sub-skill for `transformers --help`, `transformers download`, `transformers chat`, `transformers serve`, `\u002Fv1\u002Fchat\u002Fcompletions`, `\u002Fv1\u002Fcompletions`, `\u002Fv1\u002Fresponses`, `\u002Fv1\u002Faudio\u002Ftranscriptions`, `\u002Fv1\u002Fmodels`, `\u002Fload_model`, and server startup diagnostics.\n- Use sibling [inference-pipelines](..\u002Finference-pipelines\u002FSKILL.md) for in-process `pipeline(...)` task selection, batching, dtype\u002Fdevice choices, and no-server inference.\n- Use sibling [generation](..\u002Fgeneration\u002FSKILL.md) for `GenerationConfig`, `generate()`, streamers, decoding strategy, chat templates, and continuous batching internals outside the CLI server.\n- Use sibling [tokenizers-processors](..\u002Ftokenizers-processors\u002FSKILL.md) for tokenizer\u002Fprocessor loading, multimodal preprocessing, chat templates, and local asset validation.\n- Use sibling [training](..\u002Ftraining\u002FSKILL.md) for `Trainer`, `TrainingArguments`, fine-tuning, distributed training, and Hub pushing from training jobs.\n- Use sibling [quantization-integrations](..\u002Fquantization-integrations\u002FSKILL.md) for bitsandbytes, GPTQ\u002FAWQ\u002FQuanto, Accelerate placement, and backend-specific memory reductions.\n- Use sibling [model-extension](..\u002Fmodel-extension\u002FSKILL.md) for new architectures, custom modeling files, dynamic modules, and `trust_remote_code` review.\n\n## Core CLI Map\n\nThe inspected package exposes the console script:\n\n```bash\ntransformers --help\n```\n\nIt dispatches to the CLI app at `transformers.cli.transformers:main` with commands including:\n\n- `transformers download MODEL_ID [--cache-dir DIR] [--force-download] [--trust-remote-code]`\n- `transformers chat MODEL_ID [BASE_URL] [--system-prompt TEXT] [--save-folder DIR] [generate_flag=value ...]`\n- `transformers serve [MODEL_ID] [server\u002Fmodel\u002Fcontinuous-batching options]`\n- `transformers env` and `transformers version` for environment reporting.\n\nFor option groups, command examples, and expected signals, use [references\u002Fcli-reference.md](references\u002Fcli-reference.md).\n\n## Safe Preflight\n\nBefore asking an agent to start a server, run the bundled dry-run helper:\n\n```bash\npython sub-skills\u002Fserving-cli\u002Fscripts\u002Fcli_preflight.py --check-serving --check-clients\n```\n\nExpected success signal: `OK serving CLI preflight passed` with imported modules and optional package availability. Expected failure signal: non-zero exit with `ERROR` lines naming missing hard requirements or requested extras. The script does not download models, bind ports, or start `uvicorn` unless a future caller edits it; by default it only imports modules, checks optional packages, validates argument combinations, and optionally probes a health URL.\n\nUse `--port 8000` to test whether a local port appears available before starting `transformers serve`. Use `--base-url http:\u002F\u002Flocalhost:8000 --probe-health` only when a server is expected to already be running.\n\n## Starting A Server Safely\n\nDefault `transformers serve` starts a long-lived FastAPI\u002FUvicorn server on `localhost:8000`. Do not launch it by default in automated verification; prefer preflight and command construction unless the user explicitly asks to run a service.\n\nMinimal command:\n\n```bash\ntransformers serve --host localhost --port 8000 --log-level info\n```\n\nCommon production-like local command:\n\n```bash\ntransformers serve Qwen\u002FQwen2.5-0.5B-Instruct \\\n  --host localhost \\\n  --port 8000 \\\n  --dtype auto \\\n  --device auto \\\n  --model-timeout 300\n```\n\nDecision points:\n\n- Add positional `MODEL_ID` to force one preloaded model and prevent idle unload for that model.\n- Omit positional `MODEL_ID` for on-demand model loading driven by request bodies.\n- Keep `--trust-remote-code` off unless custom repository code was reviewed.\n- Use `--enable-cors` only for browser-based clients that need it; it enables permissive CORS.\n- Use `--model-timeout SECONDS` to unload idle on-demand models; it is ignored when a force model is provided.\n- Use `--default-seed N` when reproducibility matters and the backend supports deterministic behavior.\n\nFull serving workflows are in [references\u002Fserving-workflows.md](references\u002Fserving-workflows.md).\n\n## Chat CLI\n\n`transformers chat` is an interactive client for an already running `transformers serve` server. It calls `\u002Fload_model` to show model download\u002Floading progress, then streams chat completions through the server.\n\n```bash\ntransformers chat Qwen\u002FQwen2.5-0.5B-Instruct http:\u002F\u002Flocalhost:8000 max_new_tokens=128 do_sample=False\n```\n\nInside the interactive prompt:\n\n- `!help` prints commands.\n- `!status` prints model and generation settings.\n- `!clear` resets conversation history.\n- `!set temperature=0.7 max_new_tokens=256` updates generation settings.\n- `!save NAME` saves conversation and settings.\n- `!exit` ends the chat client.\n\nIf chat fails immediately, verify that the server is reachable at the positional `BASE_URL`, `\u002Fhealth` returns a response, and `\u002Fload_model` can stream a `ready` event for the chosen model.\n\n## OpenAI-Compatible APIs\n\n`transformers serve` implements local endpoints compatible with common OpenAI client patterns:\n\n- `POST \u002Fv1\u002Fchat\u002Fcompletions` for chat, multimodal messages, streaming, tools where supported, and reasoning fields.\n- `POST \u002Fv1\u002Fcompletions` for legacy prompt completion and `choices[].text` outputs.\n- `POST \u002Fv1\u002Fresponses` for Responses API-style text, image, audio, video, multi-turn, and streaming workflows.\n- `POST \u002Fv1\u002Faudio\u002Ftranscriptions` for multipart audio transcription.\n- `GET \u002Fv1\u002Fmodels` for locally cached generative model discovery.\n- `POST \u002Fload_model` for SSE model warmup and download\u002Floading progress.\n- `GET \u002Fhealth` for server liveness.\n\nUse [references\u002Fopenai-api.md](references\u002Fopenai-api.md) for request bodies, Python client snippets, streaming expectations, unsupported-field behavior, and response checks.\n\n## Continuous Batching And Optimizations\n\nEnable continuous batching only after confirming backend and model compatibility:\n\n```bash\ntransformers serve \\\n  --continuous-batching \\\n  --attn-implementation sdpa \\\n  --dtype bfloat16 \\\n  --cb-block-size 16 \\\n  --cb-max-batch-tokens 4096\n```\n\nImportant constraints:\n\n- `--compile` improves some decode loops but is documented as incompatible with continuous batching.\n- `--quantization bnb-4bit` and `--quantization bnb-8bit` request runtime bitsandbytes quantization; pre-quantized models may not need a server flag.\n- `--attn-implementation flash_attention_2` or `sdpa` must match installed kernels and hardware.\n- `--cb-use-cuda-graph`, `--cb-num-blocks`, `--cb-max-memory-percent`, and related KV cache sizing flags are performance tuning knobs, not correctness fixes.\n\nFor deeper generation and continuous batching concepts, route to [generation](..\u002Fgeneration\u002FSKILL.md). For quantization package choices, route to [quantization-integrations](..\u002Fquantization-integrations\u002FSKILL.md).\n\n## Optional Dependency Boundary\n\nA minimal Transformers import is not enough for all CLI workflows. Known boundaries:\n\n- CLI app import can fail if base CLI dependencies such as `requests` are missing, because `transformers chat` imports HTTP clients.\n- `transformers serve` requires serving extras such as `fastapi`, `uvicorn`, `pydantic`, and `openai`; install with `pip install \"transformers[serving]\"` when available.\n- `transformers chat` uses `requests`, `httpx`, `huggingface_hub`, optional `rich`, and server-side `\u002Fload_model`.\n- Model-backed serving requires backend packages such as `torch`; PyTorch-dependent classes raise optional dependency `ImportError` when torch is absent.\n- Audio and multimodal serving may need additional packages for audio\u002Fvideo\u002Fimage decoding beyond the base serving stack.\n\nTroubleshooting paths and minimal install suggestions are in [references\u002Ftroubleshooting.md](references\u002Ftroubleshooting.md).\n\n## Fast Validation Checklist\n\n1. Run `python sub-skills\u002Fserving-cli\u002Fscripts\u002Fcli_preflight.py --check-serving --check-clients --port 8000`.\n2. Check `transformers --help` only after `requests` and Typer-related CLI dependencies import successfully.\n3. Construct the intended `transformers serve` command without running it.\n4. Confirm port availability, host binding, CORS need, model source, `trust_remote_code`, dtype\u002Fdevice, and timeout policy.\n5. If a server is already running, probe `GET \u002Fhealth` and optionally `GET \u002Fv1\u002Fmodels` before sending generation requests.\n6. Warm up with `POST \u002Fload_model` only when network\u002Fmodel downloads and backend memory are acceptable.\n7. Send a tiny non-streaming request before enabling streaming, continuous batching, quantization, or browser clients.\n\n## Evidence Base\n\nThis sub-skill distills Transformers 5.13.0.dev0 behavior from CLI entrypoints, chat\u002Fdownload\u002Fserve command implementations, serving handlers, serve CLI docs, conversation docs, optimization docs, and CLI tests. Runtime guidance is self-contained and does not require opening source repository files.\n","# Serving CLI\n\nUse this sub-skill when an agent needs `transformers` command-line workflows: `download`, `chat`, `serve`, local OpenAI-compatible HTTP APIs, model warmup, server options, CORS, reasoning flags, or a safe preflight before starting long-lived services.\n\n## Route First\n\n- Use this sub-skill for `transformers --help`, `transformers download`, `transformers chat`, `transformers serve`, `\u002Fv1\u002Fchat\u002Fcompletions`, `\u002Fv1\u002Fcompletions`, `\u002Fv1\u002Fresponses`, `\u002Fv1\u002Faudio\u002Ftranscriptions`, `\u002Fv1\u002Fmodels`, `\u002Fload_model`, and server startup diagnostics.\n- Use sibling [inference-pipelines](..\u002Finference-pipelines\u002FSKILL.md) for in-process `pipeline(...)` task selection, batching, dtype\u002Fdevice choices, and no-server inference.\n- Use sibling [generation](..\u002Fgeneration\u002FSKILL.md) for `GenerationConfig`, `generate()`, streamers, decoding strategy, chat templates, and continuous batching internals outside the CLI server.\n- Use sibling [tokenizers-processors](..\u002Ftokenizers-processors\u002FSKILL.md) for tokenizer\u002Fprocessor loading, multimodal preprocessing, chat templates, and local asset validation.\n- Use sibling [training](..\u002Ftraining\u002FSKILL.md) for `Trainer`, `TrainingArguments`, fine-tuning, distributed training, and Hub pushing from training jobs.\n- Use sibling [quantization-integrations](..\u002Fquantization-integrations\u002FSKILL.md) for bitsandbytes, GPTQ\u002FAWQ\u002FQuanto, Accelerate placement, and backend-specific memory reductions.\n- Use sibling [model-extension](..\u002Fmodel-extension\u002FSKILL.md) for new architectures, custom modeling files, dynamic modules, and `trust_remote_code` review.\n\n## Core CLI Map\n\nThe inspected package exposes the console script:\n\n```bash\ntransformers --help\n```\n\nIt dispatches to the CLI app at `transformers.cli.transformers:main` with commands including:\n\n- `transformers download MODEL_ID [--cache-dir DIR] [--force-download] [--trust-remote-code]`\n- `transformers chat MODEL_ID [BASE_URL] [--system-prompt TEXT] [--save-folder DIR] [generate_flag=value ...]`\n- `transformers serve [MODEL_ID] [server\u002Fmodel\u002Fcontinuous-batching options]`\n- `transformers env` and `transformers version` for environment reporting.\n\nFor option groups, command examples, and expected signals, use [references\u002Fcli-reference.md](references\u002Fcli-reference.md).\n\n## Safe Preflight\n\nBefore asking an agent to start a server, run the bundled dry-run helper:\n\n```bash\npython sub-skills\u002Fserving-cli\u002Fscripts\u002Fcli_preflight.py --check-serving --check-clients\n```\n\nExpected success signal: `OK serving CLI preflight passed` with imported modules and optional package availability. Expected failure signal: non-zero exit with `ERROR` lines naming missing hard requirements or requested extras. The script does not download models, bind ports, or start `uvicorn` unless a future caller edits it; by default it only imports modules, checks optional packages, validates argument combinations, and optionally probes a health URL.\n\nUse `--port 8000` to test whether a local port appears available before starting `transformers serve`. Use `--base-url http:\u002F\u002Flocalhost:8000 --probe-health` only when a server is expected to already be running.\n\n## Starting A Server Safely\n\nDefault `transformers serve` starts a long-lived FastAPI\u002FUvicorn server on `localhost:8000`. Do not launch it by default in automated verification; prefer preflight and command construction unless the user explicitly asks to run a service.\n\nMinimal command:\n\n```bash\ntransformers serve --host localhost --port 8000 --log-level info\n```\n\nCommon production-like local command:\n\n```bash\ntransformers serve Qwen\u002FQwen2.5-0.5B-Instruct \\\n  --host localhost \\\n  --port 8000 \\\n  --dtype auto \\\n  --device auto \\\n  --model-timeout 300\n```\n\nDecision points:\n\n- Add positional `MODEL_ID` to force one preloaded model and prevent idle unload for that model.\n- Omit positional `MODEL_ID` for on-demand model loading driven by request bodies.\n- Keep `--trust-remote-code` off unless custom repository code was reviewed.\n- Use `--enable-cors` only for browser-based clients that need it; it enables permissive CORS.\n- Use `--model-timeout SECONDS` to unload idle on-demand models; it is ignored when a force model is provided.\n- Use `--default-seed N` when reproducibility matters and the backend supports deterministic behavior.\n\nFull serving workflows are in [references\u002Fserving-workflows.md](references\u002Fserving-workflows.md).\n\n## Chat CLI\n\n`transformers chat` is an interactive client for an already running `transformers serve` server. It calls `\u002Fload_model` to show model download\u002Floading progress, then streams chat completions through the server.\n\n```bash\ntransformers chat Qwen\u002FQwen2.5-0.5B-Instruct http:\u002F\u002Flocalhost:8000 max_new_tokens=128 do_sample=False\n```\n\nInside the interactive prompt:\n\n- `!help` prints commands.\n- `!status` prints model and generation settings.\n- `!clear` resets conversation history.\n- `!set temperature=0.7 max_new_tokens=256` updates generation settings.\n- `!save NAME` saves conversation and settings.\n- `!exit` ends the chat client.\n\nIf chat fails immediately, verify that the server is reachable at the positional `BASE_URL`, `\u002Fhealth` returns a response, and `\u002Fload_model` can stream a `ready` event for the chosen model.\n\n## OpenAI-Compatible APIs\n\n`transformers serve` implements local endpoints compatible with common OpenAI client patterns:\n\n- `POST \u002Fv1\u002Fchat\u002Fcompletions` for chat, multimodal messages, streaming, tools where supported, and reasoning fields.\n- `POST \u002Fv1\u002Fcompletions` for legacy prompt completion and `choices[].text` outputs.\n- `POST \u002Fv1\u002Fresponses` for Responses API-style text, image, audio, video, multi-turn, and streaming workflows.\n- `POST \u002Fv1\u002Faudio\u002Ftranscriptions` for multipart audio transcription.\n- `GET \u002Fv1\u002Fmodels` for locally cached generative model discovery.\n- `POST \u002Fload_model` for SSE model warmup and download\u002Floading progress.\n- `GET \u002Fhealth` for server liveness.\n\nUse [references\u002Fopenai-api.md](references\u002Fopenai-api.md) for request bodies, Python client snippets, streaming expectations, unsupported-field behavior, and response checks.\n\n## Continuous Batching And Optimizations\n\nEnable continuous batching only after confirming backend and model compatibility:\n\n```bash\ntransformers serve \\\n  --continuous-batching \\\n  --attn-implementation sdpa \\\n  --dtype bfloat16 \\\n  --cb-block-size 16 \\\n  --cb-max-batch-tokens 4096\n```\n\nImportant constraints:\n\n- `--compile` improves some decode loops but is documented as incompatible with continuous batching.\n- `--quantization bnb-4bit` and `--quantization bnb-8bit` request runtime bitsandbytes quantization; pre-quantized models may not need a server flag.\n- `--attn-implementation flash_attention_2` or `sdpa` must match installed kernels and hardware.\n- `--cb-use-cuda-graph`, `--cb-num-blocks`, `--cb-max-memory-percent`, and related KV cache sizing flags are performance tuning knobs, not correctness fixes.\n\nFor deeper generation and continuous batching concepts, route to [generation](..\u002Fgeneration\u002FSKILL.md). For quantization package choices, route to [quantization-integrations](..\u002Fquantization-integrations\u002FSKILL.md).\n\n## Optional Dependency Boundary\n\nA minimal Transformers import is not enough for all CLI workflows. Known boundaries:\n\n- CLI app import can fail if base CLI dependencies such as `requests` are missing, because `transformers chat` imports HTTP clients.\n- `transformers serve` requires serving extras such as `fastapi`, `uvicorn`, `pydantic`, and `openai`; install with `pip install \"transformers[serving]\"` when available.\n- `transformers chat` uses `requests`, `httpx`, `huggingface_hub`, optional `rich`, and server-side `\u002Fload_model`.\n- Model-backed serving requires backend packages such as `torch`; PyTorch-dependent classes raise optional dependency `ImportError` when torch is absent.\n- Audio and multimodal serving may need additional packages for audio\u002Fvideo\u002Fimage decoding beyond the base serving stack.\n\nTroubleshooting paths and minimal install suggestions are in [references\u002Ftroubleshooting.md](references\u002Ftroubleshooting.md).\n\n## Fast Validation Checklist\n\n1. Run `python sub-skills\u002Fserving-cli\u002Fscripts\u002Fcli_preflight.py --check-serving --check-clients --port 8000`.\n2. Check `transformers --help` only after `requests` and Typer-related CLI dependencies import successfully.\n3. Construct the intended `transformers serve` command without running it.\n4. Confirm port availability, host binding, CORS need, model source, `trust_remote_code`, dtype\u002Fdevice, and timeout policy.\n5. If a server is already running, probe `GET \u002Fhealth` and optionally `GET \u002Fv1\u002Fmodels` before sending generation requests.\n6. Warm up with `POST \u002Fload_model` only when network\u002Fmodel downloads and backend memory are acceptable.\n7. Send a tiny non-streaming request before enabling streaming, continuous batching, quantization, or browser clients.\n\n## Evidence Base\n\nThis sub-skill distills Transformers 5.13.0.dev0 behavior from CLI entrypoints, chat\u002Fdownload\u002Fserve command implementations, serving handlers, serve CLI docs, conversation docs, optimization docs, and CLI tests. Runtime guidance is self-contained and does not require opening source repository files.\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\u002Ftransformers\u002Fsub-skills\u002Fserving-cli",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 serving-cli",{"tool":40,"command":41},"bunx","bunx skills add VectorSpaceLab\u002FAREX-Skill --skill serving-cli",{"tool":43,"command":44},"pnpm","pnpm dlx skills add VectorSpaceLab\u002FAREX-Skill --skill serving-cli","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fskills\u002Frepositories\u002Frepo-skills\u002Ftransformers\u002Fsub-skills\u002Fserving-cli\u002FSKILL.md",[47,53,57,61,65],{"path":48,"resourceType":49,"resourceTypeDesc":50,"sizeBytes":51,"sourceHtmlUrl":52},"references\u002Fcli-reference.md","REFERENCE","参考文档",5089,"https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fskills\u002Frepositories\u002Frepo-skills\u002Ftransformers\u002Fsub-skills\u002Fserving-cli\u002Freferences\u002Fcli-reference.md",{"path":54,"resourceType":49,"resourceTypeDesc":50,"sizeBytes":55,"sourceHtmlUrl":56},"references\u002Fopenai-api.md",5366,"https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fskills\u002Frepositories\u002Frepo-skills\u002Ftransformers\u002Fsub-skills\u002Fserving-cli\u002Freferences\u002Fopenai-api.md",{"path":58,"resourceType":49,"resourceTypeDesc":50,"sizeBytes":59,"sourceHtmlUrl":60},"references\u002Fserving-workflows.md",5153,"https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fskills\u002Frepositories\u002Frepo-skills\u002Ftransformers\u002Fsub-skills\u002Fserving-cli\u002Freferences\u002Fserving-workflows.md",{"path":62,"resourceType":49,"resourceTypeDesc":50,"sizeBytes":63,"sourceHtmlUrl":64},"references\u002Ftroubleshooting.md",5803,"https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fskills\u002Frepositories\u002Frepo-skills\u002Ftransformers\u002Fsub-skills\u002Fserving-cli\u002Freferences\u002Ftroubleshooting.md",{"path":66,"resourceType":67,"resourceTypeDesc":68,"sizeBytes":69,"sourceHtmlUrl":70},"scripts\u002Fcli_preflight.py","SCRIPT","脚本",6769,"https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fskills\u002Frepositories\u002Frepo-skills\u002Ftransformers\u002Fsub-skills\u002Fserving-cli\u002Fscripts\u002Fcli_preflight.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",[],[]]