inference-api
Use the installed sgm.inference.api Python API for SDXL text-to-image, image-to-image, refiner handoff, sampler choices, model specs, and safe API inspection.
- License 许可
- MIT
- 最近更新
- 2026-09-03
- Stars
- ★ 351
- Forks
- 27
01 前言 Front Matter
- name
- inference-api
- license
- MIT
- metadata
- { "disco-role": "operating" }
- description
- Use the installed sgm.inference.api Python API for SDXL text-to-image, image-to-image, refiner handoff, sampler choices, model specs, and safe API inspection.
- disable-model-invocation
- true
02 资源文件 4
- 参考文档references/api-reference.md6.7 KB
- 参考文档references/troubleshooting.md5.8 KB
- 参考文档references/workflows.md6.4 KB
- 脚本scripts/inspect_inference_api.py7.7 KB
03 SKILL.md 正文
Inference API
Use this sub-skill when a task asks for the installed sgm.inference.api Python API: constructing SamplingPipeline, choosing SDXL base/refiner architectures, selecting samplers/discretizations/guiders, using SamplingParams, returning latents, running image_to_image with strength, or inspecting the API safely without loading checkpoints.
Route Here For
SamplingPipelineconstruction and the defaultmodel_path="checkpoints",config_path="configs/inference",device="cuda",use_fp16=Truebehavior.- SDXL text-to-image, image-to-image, and base-to-refiner plans that use
SamplingParamsandreturn_latents=True. - API-level sampler selection among
EulerEDMSampler,HeunEDMSampler,EulerAncestralSampler,DPMPP2SAncestralSampler,DPMPP2MSampler, andLinearMultistepSampler. - Diagnosing checkpoint/config mismatches, CPU-only environments, invalid enum strings,
img2img_strength, image tensor shape, and missing optional dependencies.
Start With
- Read references/api-reference.md for exact signatures, enum values,
SamplingParamsdefaults, model specs, and helper behavior. - Read references/workflows.md for SDXL txt2img, img2img, base+refiner, dry-run inspection, and validation patterns.
- Read references/troubleshooting.md when a pipeline fails to initialize, sample, load a checkpoint, or run in CPU-only conditions.
- Run
python scripts/inspect_inference_api.py --json --assert-expectedfrom this sub-skill directory, or call the script by path from any cwd, to inspect importable API facts without loading configs or checkpoints.
Boundaries And Cross-Links
- Use the root generative-models skill for repository-level routing and package context.
- Route standalone SVD, SV3D, SV4D, and video sampling scripts to video-sampling.
- Route training configuration authoring and training YAML design to training-and-configs.
- Route UI demos, Streamlit surfaces, and watermark detection workflows to demos-and-watermarking.
- Keep this sub-skill focused on the installed SDXL inference API and its helpers; do not require future agents to reopen source repository files.
04 同仓库更多 Skills
passnet-triton-opt
Make a single PassNet Triton kernel fast AND numerically faithful: performance model (when a replacement can win at all), block/grid/warp 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.
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.
passnet-skill
PassNet GPU kernel optimization via compiler passes. Design and implement Triton-based optimization passes, create pass files under ./pass_dir/, self-evaluate with pass_evaluator, and iterate to maximize GPU speedup.
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()/replacement_args()/replacement_func() files, when a pass "failed to match", or when deciding how to fuse multiple kernels/ops.
passnet-orchestrate
Round/batch planning and decision support for PassNet graph optimization. Use for multi-sample triage, worker allocation, eval-budget policy, and post-eval keep/revert 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.
passnet-feedback
Fast, deterministic feedback for PassNet work: pre-flight pattern/match 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).