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
causal_conv1d_varlen_fwd_shape
def causal_conv1d_varlen_fwd_shape(x: T, weight: T, bias: T, conv_states: T, query_start_loc: T, cache_indices: T, has_initial_state: T) -> IndexList[Int(2)]
Computes the output shape for the causal_conv1d_varlen_fwd graph op.
Args:
- x (
T): Input tensor of concatenated sequence elements with shape (dim, total_seqlen). - weight (
T): Convolution weight tensor with shape (dim, width). - bias (
T): Per-channel bias tensor with shape (dim,). - conv_states (
T): Slot-indexed in/out pool of conv states with shape (batch, dim, width - 1). - query_start_loc (
T): Cumulative sequence lengths with shape (batch + 1,). - cache_indices (
T): Indices into the conv_states pool with shape (batch,). - has_initial_state (
T): Whether each sequence has an initial state with shape (batch,).
Returns: