evaluation-conversion-deployment
Plan Sana metrics, checkpoint conversion/export, Hugging Face upload/download utilities, `sana-run` SLURM launches, and deployment routes without running heavy jobs.
- License 许可
- Apache-2.0
- 最近更新
- 2026-10-08
- Stars
- ★ 11
- Forks
- 3
01 前言 Front Matter
- name
- evaluation-conversion-deployment
- license
- Apache 2.0
- metadata
- { "disco-role": "operating" }
- description
- Plan Sana metrics, checkpoint conversion/export, Hugging Face upload/download utilities, `sana-run` SLURM launches, and deployment routes without running heavy jobs.
- disable-model-invocation
- true
02 资源文件 4
- 参考文档references/cli-and-deployment.md5.1 KB
- 参考文档references/conversion-and-export.md4.6 KB
- 参考文档references/metrics-and-benchmarks.md5.6 KB
- 参考文档references/troubleshooting.md4.5 KB
03 SKILL.md 正文
Evaluation, Conversion, and Deployment
Use this sub-skill when you need to plan a Sana evaluation, export, upload, or deployment command without executing the benchmark or conversion job.
Use for
- Metric planning for FID/CLIP, GenEval, DPG-Bench, and ImageReward.
- Checkpoint export planning for image, video, and SVDQuant/Nunchaku pipelines.
- Safe inspection of
sana-runandsana-upload. - Deployment route selection for SGLang, ComfyUI, and Gradio demos.
Do not use for
- Actual image/video generation; route to the image or video inference sub-skills.
- Training or dataset schema work; route to
training-data-configs.
Start here
references/metrics-and-benchmarks.mdreferences/conversion-and-export.mdreferences/cli-and-deployment.mdreferences/troubleshooting.md
Safe planners
scripts/plan_metrics_command.pyscripts/plan_conversion_command.pyscripts/inspect_sana_cli.py
Operating notes
- Prefer a dry, command-planning pass before any benchmark, upload, or conversion.
- Require separate metric environments and datasets when a benchmark needs them.
- Treat HF tokens, SLURM credentials, and upload destinations as secrets.
- If a requested path is missing or a model family/precision combo is inconsistent, stop at planning and explain the gap rather than guessing.
Provenance labels used while distilling
docs/metrics_toolkit.mddocs/sglang.mddocs/ComfyUI/comfyui.mddocs/model_zoo.mddocs/4bit_sana.mdscripts/bash_run_inference_metric*.shscripts/inference_*metric*.pytools/metrics/*tools/convert_scripts/*.pysana/cli/run.pysana/cli/upload2hf.py.github/workflows/ci.yaml
04 同仓库更多 Skills
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.
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.
python-api
Programmatic Object-Detection-Metrics API guidance for BoundingBox, BoundingBoxes, Evaluator, enums, coordinate conversion, AP/mAP semantics, and source-style import diagnostics.
file-evaluation
Folder-based PASCAL VOC AP/mAP evaluation from ground-truth and detection text files.
object-detection-metrics
Guides legacy Object-Detection-Metrics PASCAL VOC AP/mAP evaluation from detection text folders and direct Python API classes.
nx-workspace-patterns
Configure and optimize Nx monorepo workspaces. Use when setting up Nx, configuring project boundaries, optimizing build caching, or implementing affected commands.