For the complete documentation index, see llms.txt. Markdown versions of all pages are available by appending .md to any URL (e.g. /get-started.md).
Mojo function
gated_delta_conv1d_fwd_shape
def gated_delta_conv1d_fwd_shape(qkv_input_ragged: T, conv_weight: T, conv_state: T, slot_idx: T, input_row_offsets: T) -> IndexList[Int(2)]
Computes the output shape for the gated_delta_conv1d_fwd graph op.
Args:
- qkv_input_ragged (
T): Ragged QKV input tensor of shape[total_seq_len, conv_dim]. - conv_weight (
T): Convolution filter of shape[conv_dim, kernel_size]. - conv_state (
T): Mutable conv-state pool of shape[max_slots, conv_dim, kernel_size-1]. - slot_idx (
T): Per-batch slot indices into the conv-state pool, shape[batch_size]. - input_row_offsets (
T): Cumulative row offsets per batch, shape[batch_size + 1].
Returns: