alma is an open-weight model from Geospatial Data Science Lab, released under MIT License. Its published files total 1.4 GB.
ALMA: Airborne LiDAR Masked Autoencoding Framework for Self Supervised Representation Learning on 3DEP Point Cloud Arnav Goel and Jinha Jung reconstruction models, and complete final segmentation models.
Model Card
By Geospatial Data Science Lab, published under mit, revision 48ee80ffb7af.
ALMA: Airborne LiDAR Masked Autoencoding Framework for Self Supervised Representation Learning on 3DEP Point Cloud Arnav Goel and Jinha Jung reconstruction models, and complete final segmentation models. Download and verify using the code repository's manifest/downloader. SHA256SUMS lists the exact exported weight hashes. Optimizer states are omitted. Inputs are 20,480 XYZ points. Coordinates must be in feet. Pretraining uses bounding-box-centered XYZ divided by [50,50,25] (QL1) / [100,100,75] (QL2). Downstream uses mean-centered XY divided by half crop width and minimum-shifted Z in feet. Configurations are not interchangeable. The segmentation decoder predicts LAS 2–6; other training…
Read Geospatial Data Science Lab's full model card
ALMA pretrained models
ALMA: Airborne LiDAR Masked Autoencoding Framework for Self Supervised Representation Learning on 3DEP Point Cloud
Arnav Goel and Jinha Jung
Code and usage: https://github.com/gdslab/alma
Model weights: https://huggingface.co/gdslab/alma
Six artifacts are supplied: QL1/QL2 hierarchical encoders, full MAE reconstruction models, and complete final segmentation models. Download and verify using the code repository's manifest/downloader. SHA256SUMS lists the exact exported weight hashes. Optimizer states are omitted.
| Regime | Crop | Pretraining best epoch | Segmentation best epoch |
|---|---|---|---|
| Hamilton QL1 | 100 ft | 235 | 93 |
| Indiana 3DEP QL2 | 200 ft | 80 (stage 2) | 65 |
Inputs are 20,480 XYZ points. Coordinates must be in feet. Pretraining uses bounding-box-centered XYZ divided by [50,50,25] (QL1) / [100,100,75] (QL2). Downstream uses mean-centered XY divided by half crop width and minimum-shifted Z in feet. Configurations are not interchangeable.
The segmentation decoder predicts LAS 2–6; other training labels are ignored, with geometry retained. Confidence-thresholded unclassified is post-processing. QL2 supervised evaluation uses a geometry-aware thinned QL1 proxy; native QL2 accuracy is not established. Models are regional and require validation on new data.
See the included model cards for provenance and limitations. The source checkpoints and every exported tensor were verified locally. The user reported all 12 release tests passing on 2026-09-30, including QL1/QL2 encoder and segmentation decoder GPU forwards. Fresh installation, full training, MAE CUDA loss/backward and real-tile execution have not been validated by this release audit.
Model weights are released under the MIT license; see WEIGHTS_LICENSE. Code retains its upstream MIT license and attribution; see LICENSE and NOTICE.
Identity and Version
- Repository
- gdslab/alma
- Publisher
- Geospatial Data Science Lab
- Task
- Not stated by the source
- Modality
- Other
- Library
- Not stated by the source
- Parameters
- Not stated by the source
- Languages
- en
- Revision
- 48ee80ffb7aff1582fbbc679be920f84b00a2394
- First published
- 2026-09-30
- Last updated
- 2026-10-01
Files and Weights
36 files, 1.4 GB in total. The weights are 6 files totalling 1.4 GB in pth.
Every file
| File | Type | Size | SHA-256 |
|---|---|---|---|
| ALMA-QL1-Indiana-encoder.pth | Weights | 137.6 MB | 8af44fb35a40 |
| ALMA-QL1-Indiana-mae.pth | Weights | 178.0 MB | bdd5b35de0a7 |
| ALMA-QL1-Indiana-seg.pth | Weights | 179.9 MB | 1e8c6bc90926 |
| ALMA-QL2-Indiana-encoder.pth | Weights | 265.6 MB | c6c686043200 |
| ALMA-QL2-Indiana-mae.pth | Weights | 337.1 MB | 1e4a7e367cac |
| ALMA-QL2-Indiana-seg.pth | Weights | 339.2 MB | 16c31924136c |
| configs/ql1_encoder_exp6.yaml | Configuration | 525 B | — |
| configs/ql1_pretraining.yaml | Configuration | 2.1 KB | — |
| configs/ql1_segmentation.yaml | Configuration | 1.9 KB | — |
| configs/ql1_segmentation_eval.yaml | Configuration | 2.0 KB | — |
| configs/ql2_encoder_exp10_stage2.yaml | Configuration | 522 B | — |
| configs/ql2_pretraining_stage1.yaml | Configuration | 2.2 KB | — |
| configs/ql2_pretraining_stage2.yaml | Configuration | 2.3 KB | — |
| configs/ql2_segmentation.yaml | Configuration | 2.2 KB | — |
| configs/ql2_segmentation_eval.yaml | Configuration | 2.2 KB | — |
| manifest.yaml | Configuration | 3.9 KB | — |
| LICENSE | Documentation | 1.1 KB | — |
| NOTICE | Documentation | 541 B | — |
| README.md | Documentation | 2.1 KB | — |
| WEIGHTS_LICENSE | Documentation | 1.1 KB | — |
| docs/data_preparation.md | Documentation | 679 B | — |
| docs/finetuning.md | Documentation | 1.4 KB | — |
| docs/inference.md | Documentation | 1.5 KB | — |
| docs/pretraining.md | Documentation | 1.9 KB | — |
| docs/publishing.md | Documentation | 6.3 KB | — |
| docs/release_audit.md | Documentation | 4.1 KB | — |
| docs/reproducibility.md | Documentation | 552 B | — |
| model_cards/ALMA-QL1-Indiana-encoder.md | Documentation | 990 B | — |
| model_cards/ALMA-QL1-Indiana-seg.md | Documentation | 966 B | — |
| model_cards/ALMA-QL2-Indiana-encoder.md | Documentation | 892 B | — |
| model_cards/ALMA-QL2-Indiana-seg.md | Documentation | 955 B | — |
| model_cards/huggingface.md | Documentation | 2.1 KB | — |
| CITATION.cff | Other | 452 B | — |
| CODE_CHANGE_LOG.txt | Other | 10.1 KB | — |
| SHA256SUMS | Other | 554 B | — |
| .gitattributes | Repository | 1.5 KB | — |
License and Download
- License
- mit
- Access
- Open weights, no gate
- Download size
- 1.4 GB
Released by Geospatial Data Science Lab through its official repository on Hugging Face. Read the license.
Memory Requirements
| Precision | Weights in memory |
|---|---|
| As published | 1.4 GB |
Weights only, from the published parameter count; the key-value cache and runtime add to this.
Questions About alma
Can I use alma commercially?
Yes. alma is released under MIT License. The MIT License is a short permissive license. It permits commercial use, modification and redistribution, provided the copyright notice and permission notice are included.