Skip to content

issues Search Results · language:Edge language:Python linked:pr language:JavaScript language:TypeScript language:Python

Filter by

7.9M results  (353 ms)

7.9M results

Issue 030: Document strict MAE model and pretraining workflow Goal Add concise documentation for the strict 3D attribute-set MAE model and Stage 1 pretraining workflow. Context After Issues 020-029, ...

Issue 028: Implement MAE pretraining engine with checkpointing Goal Implement the Stage 1 strict MAE pretraining loop. Context The repository currently has dry-run entrypoints. After this issue, proc/train_mae.py ...

Issue 029: Integrate real MAE training into proc/train_mae.py Goal Replace the current dry-run-only proc/train_mae.py behavior with a thin entrypoint that can either validate config or launch real MAE ...

Issue 027: Implement torch collate and DataLoader helpers for MAE pretraining Goal Bridge the NumPy dataset outputs to PyTorch training batches, including variable input attribute counts. Context NopimsAttributePretrainDataset ...

Issue 026: Implement MAE reconstruction and gradient losses Goal Implement loss functions needed for Stage 1 strict MAE pretraining. Context The model predicts patch-level reconstructions: pred_patches: ...

Issue 025: Implement StrictAttributeSetMAE3D model Goal Implement the core strict 3D attribute-set MAE model for MVP pretraining. Context The model should use: - variable input attribute subset ...

Issue 024: Implement transformer blocks and 3D positional encodings Goal Add reusable transformer blocks and 3D positional encoding utilities for the strict MAE encoder and decoder. Context The MVP ...

Issue 023: Implement context token pooling for local+context MAE Goal Implement a lightweight context token pooler that turns low-resolution context crop tokens into a small set of global context tokens. ...

Issue 021: Implement 3D patchify / unpatchify utilities Goal Implement reusable 3D patch utilities for strict MAE targets and decoder outputs. The utilities must use grid order [x, y, z] consistently. ...

Issue 022: Implement AttributePatchTokenizer3D Goal Implement the MVP attribute-set tokenizer that converts a variable number of input attribute volumes into fixed spatial tokens for strict 3D MAE. ...
Issue origami icon

Learn how you can use GitHub Issues to plan and track your work.

Save views for sprints, backlogs, teams, or releases. Rank, sort, and filter issues to suit the occasion. The possibilities are endless.Learn more about GitHub Issues
ProTip! Restrict your search to the title by using the in:title qualifier.
Issue origami icon

Learn how you can use GitHub Issues to plan and track your work.

Save views for sprints, backlogs, teams, or releases. Rank, sort, and filter issues to suit the occasion. The possibilities are endless.Learn more about GitHub Issues
ProTip! Restrict your search to the title by using the in:title qualifier.