Remove 200us slow concat kernel (part 1: kernel) (#7145)
This commit is contained in:
@@ -105,7 +105,8 @@ void merge_state_v2(
|
||||
at::Tensor v_a, at::Tensor s_a, at::Tensor v_b, at::Tensor s_b, at::Tensor v_merged, at::Tensor s_merged);
|
||||
void cutlass_mla_decode(
|
||||
torch::Tensor const& out,
|
||||
torch::Tensor const& q_nope_and_q_pe,
|
||||
torch::Tensor const& q_nope,
|
||||
torch::Tensor const& q_pe,
|
||||
torch::Tensor const& kv_c_and_k_pe_cache,
|
||||
torch::Tensor const& seq_lens,
|
||||
torch::Tensor const& page_table,
|
||||
|
||||
Reference in New Issue
Block a user