feat: add OlmoEarth v1/v1.1 embedder (S2 L2A 12-band, FlexiViT)#85
Open
amrit110 wants to merge 4 commits into
Open
feat: add OlmoEarth v1/v1.1 embedder (S2 L2A 12-band, FlexiViT)#85amrit110 wants to merge 4 commits into
amrit110 wants to merge 4 commits into
Conversation
Adds a new on-the-fly embedder for the OlmoEarth foundation model family (Allen AI) trained on the Major TOM dataset. Supports all 7 released variants: nano/tiny/base/large (v1) and nano/tiny/base (v1.1). Key implementation details: - Fetches 12 S2 L2A bands from GEE in OlmoEarth band-set order (10 m: B2/B3/B4/B8, 20 m: B5/B6/B7/B8A/B11/B12, 60 m: B1/B9) - Per-band normalization via OlmoEarth COMPUTED strategy (mean±2σ) - FlexiViT encoder accepts configurable patch_size (1-8, default 4) and image_size (default 256, the training tile size) - Timestamps derived automatically from the temporal midpoint date - Both pooled (global mean/max over spatial+temporal+band-set dims) and grid (spatial token map D×H'×W') output modes supported - True batch inference with shape-grouped forward passes - olmoearth-pretrain-minimal added as an optional extra [olmoearth] Also fixes two pre-existing broken anchor links in prithvi.md and satmae.md.
Contributor
Author
|
@Dinghye, please review! |
Collaborator
|
@amrit110 crazy! I will check this today. Thank you soooo much! |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
olmoearthas a new on-the-fly embedder for the OlmoEarth foundation model family (Allen AI), trained on the Major TOM datasetnano,tiny,base,large(v1) andnano_v1_1,tiny_v1_1,base_v1_1(v1.1) — embedding dims 128/192/768/1024pooled(global pool over spatial+temporal+band-set token dims) andgrid(spatial token map D×H'×W') output modes supportedpatch_size(1–8, default 4) andimage_size(default 256, matching training tile size)olmoearth-pretrain-minimaladded as an optional extra (pip install rs-embed[olmoearth])docs/models/olmoearth.md; catalog/reference tables updatedprithvi.mdandsatmae.mdTest plan
list_models(),describe_model(),Modelclassmkdocs build --strict)