Skip to content

Commit

Permalink
fix lint
Browse files Browse the repository at this point in the history
  • Loading branch information
grimoire committed Sep 18, 2023
1 parent c1ec6e3 commit a2d96b0
Show file tree
Hide file tree
Showing 2 changed files with 8 additions and 8 deletions.
Original file line number Diff line number Diff line change
Expand Up @@ -1422,8 +1422,8 @@ __global__ void masked_multihead_attention_kernel(Multihead_attention_params<T>
// Trigger the stores to global memory.
if (Dh == Dh_MAX || co < Dh / QK_ELTS_IN_16B) {

size_t offset = params.kv_cache_per_sample_offset + kvhi * params.memory_max_len * Dh + tlength_circ * Dh
+ co * QK_ELTS_IN_16B + ci;
size_t offset = params.kv_cache_per_sample_offset + kvhi * params.memory_max_len * Dh
+ tlength_circ * Dh + co * QK_ELTS_IN_16B + ci;

if (!QUANT_POLICY) {
*reinterpret_cast<Qk_vec_m*>(&params.k_cache_per_sample[bi][offset]) =
Expand Down
12 changes: 6 additions & 6 deletions src/turbomind/models/llama/llama_kernels.h
Original file line number Diff line number Diff line change
Expand Up @@ -80,12 +80,12 @@ void invokeMyCopyInt(int* dst, const int* src, size_t count, cudaStream_t st);

template<typename T>
struct BaseAttentionLayout {
int stride_batch;
int stride_seq;
int stride_head;
bool use_seqlens = false;
size_t batch_seqs_offset = 0;
T** batch_seqs = nullptr;
int stride_batch;
int stride_seq;
int stride_head;
bool use_seqlens = false;
size_t batch_seqs_offset = 0;
T** batch_seqs = nullptr;
};

template<typename T>
Expand Down

0 comments on commit a2d96b0

Please sign in to comment.