fix typos

Author: zigzagcai

csrc/selective_scan/selective_scan_bwd_kernel.cuh

@@ -136,7 +136,7 @@ void selective_scan_bwd_kernel(SSMParamsBwd params) {
         : reinterpret_cast<scan_t *>(params.x_ptr) + (batch_id * params.dim + dim_id) * (params.n_chunks) * params.dstate;
     float dD_val = 0;
     float ddelta_bias_val = 0;
-    long *cu_seqlens = reinterpret_cast<long *>(params.cu_seqlens_ptr) + batch_id * params.u_batch_stride
+    long *cu_seqlens = reinterpret_cast<long *>(params.cu_seqlens_ptr) + batch_id * params.u_batch_stride;
 
     constexpr int kChunkSize = kNThreads * kNItems;
     u += (params.n_chunks - 1) * kChunkSize;

csrc/selective_scan/selective_scan_fwd_kernel.cuh

@@ -107,7 +107,7 @@ void selective_scan_fwd_kernel(SSMParamsBase params) {
     weight_t *C = reinterpret_cast<weight_t *>(params.C_ptr) + dim_id * kNRows * params.C_d_stride;
     input_t *Cvar = reinterpret_cast<input_t *>(params.C_ptr) + batch_id * params.C_batch_stride + group_id * params.C_group_stride;
     scan_t *x = reinterpret_cast<scan_t *>(params.x_ptr) + (batch_id * params.dim + dim_id * kNRows) * params.n_chunks * params.dstate;
-    long *cu_seqlens = reinterpret_cast<long *>(params.cu_seqlens_ptr) + batch_id * params.u_batch_stride
+    long *cu_seqlens = reinterpret_cast<long *>(params.cu_seqlens_ptr) + batch_id * params.u_batch_stride;
 
     float D_val[kNRows] = {0};
     if (params.D_ptr != nullptr) {