[Kernel] Change interface to Mamba causal_conv1d_update for continuous batching (#8012)
This commit is contained in:
committed by
GitHub
parent
09deb4721f
commit
8110e44529
@@ -36,6 +36,10 @@ struct ConvParamsBase {
|
||||
|
||||
void *__restrict__ conv_state_ptr;
|
||||
|
||||
// For the continuous batching case. Makes it so that the mamba state for
|
||||
// the current batch doesn't need to be a contiguous tensor.
|
||||
int32_t *__restrict__ conv_state_indices_ptr;
|
||||
|
||||
void *__restrict__ seq_idx_ptr;
|
||||
|
||||
// No __restrict__ since initial_states could be the same as final_states.
|
||||
|
||||
Reference in New Issue
Block a user