Model loading guide¶
This page documents how to load each model exported by rshf.
Discover repository IDs¶
When a model uses from_pretrained, pass a Hugging Face repository ID.
You can find matching IDs in the curated collection with:
from rshf import list_models
list_models("satmae")
The underlying collection is:
MVRL/remote-sensing-foundation-models-664e8fcd67d8ca8c03f42d00.
Per-model loading reference¶
Model |
Import |
Load |
|---|---|---|
BioCLIP |
|
|
Climplicit |
|
|
CLIP |
|
|
CROMA |
|
|
DinoV3 Sat |
|
|
GeoCLAP |
|
|
GeoCLIP |
|
|
Presto |
|
|
Prithvi |
|
|
ProM3E |
|
|
RCME |
|
|
RemoteCLIP |
|
|
RVSA |
|
|
Sat2Cap |
|
|
SatClip |
|
|
SatMAE |
|
|
SatMAE multi-spectral pretrain |
|
|
SatMAE multi-spectral finetune |
|
|
SatMAE++ |
|
|
ScaleMAE |
|
|
SenCLIP |
|
|
SINR |
|
|
StreetCLIP |
|
|
TaxaBind |
|
|
Additional CLIP exports¶
The rshf.clip module also exports CLIPProcessor, CLIPTextModel,
and CLIPVisionModel:
from rshf.clip import CLIPProcessor, CLIPTextModel, CLIPVisionModel
processor = CLIPProcessor.from_pretrained("<hf-repo-id>")
text_encoder = CLIPTextModel.from_pretrained("<hf-repo-id>")
vision_encoder = CLIPVisionModel.from_pretrained("<hf-repo-id>")