For the complete documentation index, see llms.txt. Markdown versions of all pages are available by appending .md to any URL (e.g. /max/get-started.md).

Mojo function

causal_conv1d_varlen_update_shape

def causal_conv1d_varlen_update_shape[dtype: DType](x: ManagedTensorSlice[IOSpec[_, _].Input, static_spec=x.static_spec], weight: ManagedTensorSlice[IOSpec[_, _].Input, static_spec=weight.static_spec], bias: ManagedTensorSlice[IOSpec[_, _].Input, static_spec=bias.static_spec], cache_seqlens: ManagedTensorSlice[IOSpec[_, _].Input, static_spec=cache_seqlens.static_spec], conv_state_indices: ManagedTensorSlice[IOSpec[_, _].Input, static_spec=conv_state_indices.static_spec]) -> IndexList[Int(3)]

Returns the output shape for the causal_conv1d_varlen_update op.

The output shape equals the input shape (batch, dim, seqlen).

Parameters:

dtype (DType): Element type of the x, weight, and bias input tensors.

Args:

x (ManagedTensorSlice[IOSpec[_, _].Input, static_spec=x.static_spec]): Input tensor with shape (batch, dim, seqlen).
weight (ManagedTensorSlice[IOSpec[_, _].Input, static_spec=weight.static_spec]): Convolution weights with shape (dim, width).
bias (ManagedTensorSlice[IOSpec[_, _].Input, static_spec=bias.static_spec]): Per-channel bias with shape (dim,).
cache_seqlens (ManagedTensorSlice[IOSpec[_, _].Input, static_spec=cache_seqlens.static_spec]): Current sequence lengths per batch entry with shape (batch,).
conv_state_indices (ManagedTensorSlice[IOSpec[_, _].Input, static_spec=conv_state_indices.static_spec]): Indices into the conv state buffer with shape (batch,).

Returns:

IndexList[Int(3)]: The output tensor shape, equal to x.shape().