VoltBackboneV2

class VoltBackboneV2(in_channels: int = 4, stem_channels: int = 32, embed_dim: int = 384, depth: int = 12, num_heads: int = 6, mlp_ratio: float = 4.0, init_values: float | None = None, qk_norm: bool = True, drop_path: float = 0.3, increase_drop_path: bool = True, stride: int = 5, kernel_size: int = 5, mask_token: bool = True, decoder_dim: int = 256, rope_max_grid_size: tuple = (1024, 1024, 1024), rope_freq_split: tuple = (11, 11, 10))

Bases: PointModel

Flat Volt encoder + transposed-conv decoder, returning per-input-point features.

VoltBackboneV2.forward(point)