Configuration

torchgeo-bench is driven by Hydra. The primary config lives at src/torchgeo_bench/conf/config.yaml and is composed with a model preset selected from src/torchgeo_bench/conf/model/.

Every value in the config can be overridden on the command line using Hydra’s dotted-path syntax:

$ torchgeo-bench run \
    model=timm/resnet50 \
    dataset.names=[m-eurosat] \
    eval.bootstrap=100 \
    eval.skip_linear=true \
    device=cuda:1

Config tree

The packaged config tree is shipped inside the wheel:

src/torchgeo_bench/conf/
├── config.yaml          # primary config (defaults below)
└── model/
    ├── rcf.yaml
    ├── imagestats.yaml
    ├── sam3_encoder.yaml
    ├── olmoearth_{base,large}.yaml
    ├── timm/
    │   └── ...          # ResNet, ConvNeXt, EfficientNet, ViT/Swin/DeiT, ...
    └── torchgeo/
        └── ...          # SSL backbones, ScaleMAE, DOFA, Satlas, EarthLoc, ...

See Models for an operator-facing tour of the available presets.

Top-level options

Key	Meaning
seed	Global RNG seed (numpy + torch).
device	PyTorch device string (e.g. cuda:0, cpu).
output	Path to the appended results CSV.
verbose	Toggle progress logging.
resume	Skip already-computed (dataset, method, model, config) combos.

dataset block

dataset:
  names: all                    # or a list, e.g. [m-eurosat, m-pv4ger]
  partition: default            # alternative GeoBench V1 partitions when supported
  batch_size: 64
  bands: rgb                    # rgb | all | [red, green, blue, nir]
  image_size: 224               # null disables resizing
  interpolation: bilinear       # bilinear | bicubic | nearest

See Datasets for the full list of available dataset names and band selection semantics.

eval block

eval:
  bootstrap: 200                # bootstrap resamples for KNN/linear CIs
  c_range: [-6, 4, 40]          # log10 sweep start, stop, num samples for linear probe
  merge_val: true               # merge train+val before training the final logistic head
  skip_linear: false            # skip the (slower) linear probe entirely

  intrinsic_dim:                # optional ID metrics on extracted embeddings
    enabled: false
    estimators: [TwoNN, MLE, lPCA]
    splits: [train]
    max_samples: 10000
    device: null                # null = auto (cuda if available)

  segmentation:                 # used only for segmentation datasets
    head_type: fpn              # linear | fpn | dpt | conv-block
    layers: []                  # backbone layers to extract; [] = use defaults
    lr: 1e-3
    epochs: 10
    criterion:
      _target_: torch.nn.CrossEntropyLoss
      ignore_index: 255
    lr_scheduler: cosine
    cache_features: true
    cache_dtype: float16
    save_viz: false
    viz_dir: viz
    n_viz_samples: 8

See also

Segmentation backbone layer reference lists the verified layers values for every supported timm backbone family, with spatial sizes and notes on stages that share resolution (common in EfficientNet / MobileNet).

Refer to Evaluation for the runtime functions that consume each sub-block.

model block

The default model: rcf selects Random Convolutional Features. Browse src/torchgeo_bench/conf/model/ for the full list of presets. Each preset can ship its own eval overrides (for example, segmentation backbones often pin head_type: fpn and bump the learning rate); these are merged on top of the global eval block per dataset.