models-and-transforms
Use for TorchGeo model registry, pretrained weights, model builders, timm/SMP integration, and spectral/SAR/color/spatial/temporal transforms.
- License 许可
- MIT
- 最近更新
- 2026-09-03
- Stars
- ★ 351
- Forks
- 27
01 前言 Front Matter
- name
- models-and-transforms
- license
- MIT
- metadata
- { "disco-role": "operating" }
- description
- Use for TorchGeo model registry, pretrained weights, model builders, timm/SMP integration, and spectral/SAR/color/spatial/temporal transforms.
- disable-model-invocation
- true
02 资源文件 1
- 参考文档references/models-and-transforms.md2.2 KB
03 SKILL.md 正文
TorchGeo models and transforms
Use this sub-skill for pre-trained geospatial models, weight enums, model registry queries, transform composition, spectral indices, and optional model dependency troubleshooting.
Model registry
torchgeo.models.api registers named model builders and weights. Core helpers:
list_models()returns registered model names.get_model(name, *args, **kwargs)instantiates a registered model builder.get_model_weights(name_or_builder)returns theWeightsEnumclass associated with a model.get_weight(full_enum_string)returns one concrete weight enum value by its string representation.
Registered families at the distilled commit include Aurora, CopernicusFM, CROMA, DEO, DOFA, EarthLoc, OlmoEarth, Panopticon, Presto, ResNet, SatCLIP, ScaleMAE, Swin, Tessera, TileNet, UNet, and ViT variants.
Weight and dependency rules
- Use
weights=Nonefor construction-only smoke tests. - Passing a TorchGeo weight enum or enum string may download checkpoints through
get_state_dict; check network/cache policy first. - Some model families require optional extras (
microsoft-aurora,olmoearth-pretrain-minimal, or other upstream packages). Guard optional imports with clear user guidance. - When changing input channel counts, update both model construction parameters and any weight-loading adaptation logic.
Transform families
torchgeo.transforms.indicesappends spectral indices such as NDVI, NBR, NDBI, NDSI, NDWI, MNDWI, and related normalized-difference features.- SAR, color, spatial, and temporal transforms are in neighboring modules and are built around tensor/Kornia-style batch operations.
- Verify channel ordering before applying a spectral index. For example, NDVI needs NIR and red channel indexes in the image tensor.
- In Lightning workflows, put GPU-compatible Kornia augmentations in
on_after_batch_transferthrough datamodule augmentation fields.
Read next
- reference for examples and verification candidates.
- Root backend plan before running model/task smoke checks.
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).