SparseWake/sparsewake
SparseWake SparseWake is a synthetic benchmark for sparse temporal hydrodynamic sensing. ICLR 2027 release The expanded release adds controlled multi-source mixtures and common-prior nearest-source tasks, with complete core data banks, reference checkpoints, a small review supplement, and reproduction code with a frozen wake-library input. Download release iclr2027-v1.0rc2 The version page lists the three archives, exact sizes, checksums, extraction instructions… See the full description on the dataset page: https://huggingface.co/datasets/SparseWake/sparsewake.
0279
1from __future__ import annotations2 3import torch4from torch import nn5 6 7class TemporalMLP(nn.Module):8 def __init__(self, input_dim: int, output_dim: int = 2, hidden: tuple[int, ...] = (256, 256, 128), dropout: float = 0.05):9 super().__init__()10 layers: list[nn.Module] = []11 last = input_dim12 for width in hidden:13 layers.extend([nn.Linear(last, width), nn.LayerNorm(width), nn.GELU(), nn.Dropout(dropout)])14 last = width15 layers.append(nn.Linear(last, output_dim))16 self.net = nn.Sequential(*layers)17 18 def forward(self, x: torch.Tensor) -> torch.Tensor:19 return self.net(x)20 21