flashinfer.cake_causal_conv1d_det_train.cake_causal_conv1d_det_train_backward_workspace

flashinfer.cake_causal_conv1d_det_train.cake_causal_conv1d_det_train_backward_workspace(S: int, C: int, device: device, *, bwd_grid_mult: int = 1) → Tensor

Allocate a backward workspace (see ..._workspace_bytes; no zeroing is needed).