204d17086a
feat(graph-mixer): implement L0 sparsity with Hard-Concrete gate for channel selection
2025-09-11 16:50:58 +08:00
5fc0da4239
refactor(mamba): add residual connection to Mamba2Encoder layers
2025-09-10 21:23:17 +08:00
598fdaadbc
feat(mamba): extract last time step from Mamba2Encoder output
2025-09-10 21:03:35 +08:00
b139f711bc
refactor(mamba): adjust Mamba2Encoder layer configuration
2025-09-10 21:00:31 +08:00
9787badd25
feat(mambaseries): allow stacking multiple Mamba2 layers
2025-09-10 20:56:24 +08:00
ff987da4c6
refactor(xPatch): remove redundant encoder argument from season_net call
2025-09-10 16:03:43 +08:00
1044c60fe7
refactor(SeasonPatch): unify encoder and head initialization
2025-09-10 16:00:52 +08:00
908d3a7080
feat(mamba): add Mamba2 encoder option to SeasonPatch
2025-09-10 15:51:28 +08:00
96c40c6ab6
feat: add xpatch_sparsechannel test script
2025-09-10 10:54:50 +08:00
9f7fb24beb
refactor(graphmixer): enhance channel graph attention with ST-Gumbel
2025-09-06 00:06:26 +08:00
ef307a57e9
feat: add mamba and dynamic chunking related code and test code
2025-09-04 01:32:13 +00:00
12cb7652cf
feat: add sigmoid learning rate scheduler
...
Add new 'sigmoid' learning rate adjustment method to utils/tools.py
with logistic growth rate and warm-up smoothing for better training
convergence.
2025-09-03 07:13:25 +00:00
a069c9a874
feat: add PEMS and Solar dataset support
...
- Add Dataset_PEMS and Dataset_Solar classes for PEMS and Solar datasets
- Update data_factory.py to include new dataset mappings
- Fix M4 dataset handling with proper numpy array dtype
- Add PEMS-specific loss function (L1Loss) and inverse transform support
- Update validation logic for PEMS dataset with inverse scaling
- Fix M4 data loader insample mask calculation bug
Changes support new traffic and solar energy datasets while maintaining
backward compatibility with existing datasets.
2025-09-03 07:01:32 +00:00
d6dd462886
first commit
2025-08-28 10:17:59 +00:00
