Skip to content

TIL: fix trash output for code llama for long prompt #24

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 7 commits into from
Oct 8, 2023
Merged

TIL: fix trash output for code llama for long prompt #24

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
7 commits
Select commit Hold shift + click to select a range
94a24da
commit
sfc-gh-zhwang Oct 8, 2023
404cc95
commit
sfc-gh-zhwang Oct 8, 2023
63002c8
commit
sfc-gh-zhwang Oct 8, 2023
c3f9b0f
commit
sfc-gh-zhwang Oct 8, 2023
94cd065
commit
sfc-gh-zhwang Oct 8, 2023
783b025
commit
sfc-gh-zhwang Oct 8, 2023
cc7b81c
commit
sfc-gh-zhwang Oct 8, 2023
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
10 changes: 6 additions & 4 deletions .../llama/decoder_masked_groupedquery_attention/decoder_masked_groupedquery_attention_128.cu
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -25,10 +25,13 @@
////////////////////////////////////////////////////////////////////////////////////////////////////

#define MGQA_LAUNCH_KERNEL( \
T, Dh, Dh_MAX, THDS_PER_KEY, THDS_PER_VALUE, THDS_PER_BLOCK, HAS_BEAMS, stream) \
size_t smem_sz = mmha::smem_size_in_bytes<T>(params, THDS_PER_VALUE, THDS_PER_BLOCK); \
T, Dh, Dh_MAX, THDS_PER_KEY, THDS_PER_VALUE, THDS_PER_BLOCK, HAS_BEAMS, stream) \
size_t smem_sz = mmha::smem_size_in_bytes<T>(params, THDS_PER_VALUE, THDS_PER_BLOCK); \
dim3 grid(params.num_heads, params.batch_size); \
mmha::masked_groupedquery_attention_kernel<T, \
cudaFuncSetAttribute(mmha::masked_groupedquery_attention_kernel<T, \
Dh, Dh_MAX, THDS_PER_KEY, THDS_PER_VALUE, THDS_PER_BLOCK, HAS_BEAMS>, \
cudaFuncAttributeMaxDynamicSharedMemorySize, smem_sz); \
mmha::masked_groupedquery_attention_kernel<T, \
Dh, \
Dh_MAX, \
THDS_PER_KEY, \
Expand All @@ -43,7 +46,6 @@ void mgqa_launch_kernel(const KERNEL_PARAMS_TYPE& params, const cudaStream_t& st
{
constexpr int THREADS_PER_VALUE = threads_per_value_t<T, Dh_MAX>::value;
int tlength = params.timestep;
// printf("tlength, CROSS_ATTENTION = %d, %d\n", tlength);
if (params.cache_indir == nullptr) {
if (tlength < 32) {
MGQA_LAUNCH_KERNEL(T, Dh, Dh_MAX, 4, THREADS_PER_VALUE, 64, false, stream);
Expand Down
9 changes: 6 additions & 3 deletions .../llama/decoder_masked_groupedquery_attention/decoder_masked_groupedquery_attention_144.cu
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -25,10 +25,13 @@
////////////////////////////////////////////////////////////////////////////////////////////////////

#define MGQA_LAUNCH_KERNEL( \
T, Dh, Dh_MAX, THDS_PER_KEY, THDS_PER_VALUE, THDS_PER_BLOCK, HAS_BEAMS, stream) \
size_t smem_sz = mmha::smem_size_in_bytes<T>(params, THDS_PER_VALUE, THDS_PER_BLOCK); \
T, Dh, Dh_MAX, THDS_PER_KEY, THDS_PER_VALUE, THDS_PER_BLOCK, HAS_BEAMS, stream) \
size_t smem_sz = mmha::smem_size_in_bytes<T>(params, THDS_PER_VALUE, THDS_PER_BLOCK); \
dim3 grid(params.num_heads, params.batch_size); \
mmha::masked_groupedquery_attention_kernel<T, \
cudaFuncSetAttribute(mmha::masked_groupedquery_attention_kernel<T, \
Dh, Dh_MAX, THDS_PER_KEY, THDS_PER_VALUE, THDS_PER_BLOCK, HAS_BEAMS>, \
cudaFuncAttributeMaxDynamicSharedMemorySize, smem_sz); \
mmha::masked_groupedquery_attention_kernel<T, \
Dh, \
Dh_MAX, \
THDS_PER_KEY, \
Expand Down
9 changes: 6 additions & 3 deletions .../llama/decoder_masked_groupedquery_attention/decoder_masked_groupedquery_attention_160.cu
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -25,10 +25,13 @@
////////////////////////////////////////////////////////////////////////////////////////////////////

#define MGQA_LAUNCH_KERNEL( \
T, Dh, Dh_MAX, THDS_PER_KEY, THDS_PER_VALUE, THDS_PER_BLOCK, HAS_BEAMS, stream) \
size_t smem_sz = mmha::smem_size_in_bytes<T>(params, THDS_PER_VALUE, THDS_PER_BLOCK); \
T, Dh, Dh_MAX, THDS_PER_KEY, THDS_PER_VALUE, THDS_PER_BLOCK, HAS_BEAMS, stream) \
size_t smem_sz = mmha::smem_size_in_bytes<T>(params, THDS_PER_VALUE, THDS_PER_BLOCK); \
dim3 grid(params.num_heads, params.batch_size); \
mmha::masked_groupedquery_attention_kernel<T, \
cudaFuncSetAttribute(mmha::masked_groupedquery_attention_kernel<T, \
Dh, Dh_MAX, THDS_PER_KEY, THDS_PER_VALUE, THDS_PER_BLOCK, HAS_BEAMS>, \
cudaFuncAttributeMaxDynamicSharedMemorySize, smem_sz); \
mmha::masked_groupedquery_attention_kernel<T, \
Dh, \
Dh_MAX, \
THDS_PER_KEY, \
Expand Down
9 changes: 6 additions & 3 deletions .../llama/decoder_masked_groupedquery_attention/decoder_masked_groupedquery_attention_192.cu
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -25,10 +25,13 @@
////////////////////////////////////////////////////////////////////////////////////////////////////

#define MGQA_LAUNCH_KERNEL( \
T, Dh, Dh_MAX, THDS_PER_KEY, THDS_PER_VALUE, THDS_PER_BLOCK, HAS_BEAMS, stream) \
size_t smem_sz = mmha::smem_size_in_bytes<T>(params, THDS_PER_VALUE, THDS_PER_BLOCK); \
T, Dh, Dh_MAX, THDS_PER_KEY, THDS_PER_VALUE, THDS_PER_BLOCK, HAS_BEAMS, stream) \
size_t smem_sz = mmha::smem_size_in_bytes<T>(params, THDS_PER_VALUE, THDS_PER_BLOCK); \
dim3 grid(params.num_heads, params.batch_size); \
mmha::masked_groupedquery_attention_kernel<T, \
cudaFuncSetAttribute(mmha::masked_groupedquery_attention_kernel<T, \
Dh, Dh_MAX, THDS_PER_KEY, THDS_PER_VALUE, THDS_PER_BLOCK, HAS_BEAMS>, \
cudaFuncAttributeMaxDynamicSharedMemorySize, smem_sz); \
mmha::masked_groupedquery_attention_kernel<T, \
Dh, \
Dh_MAX, \
THDS_PER_KEY, \
Expand Down
9 changes: 6 additions & 3 deletions .../llama/decoder_masked_groupedquery_attention/decoder_masked_groupedquery_attention_224.cu
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -25,10 +25,13 @@
////////////////////////////////////////////////////////////////////////////////////////////////////

#define MGQA_LAUNCH_KERNEL( \
T, Dh, Dh_MAX, THDS_PER_KEY, THDS_PER_VALUE, THDS_PER_BLOCK, HAS_BEAMS, stream) \
size_t smem_sz = mmha::smem_size_in_bytes<T>(params, THDS_PER_VALUE, THDS_PER_BLOCK); \
T, Dh, Dh_MAX, THDS_PER_KEY, THDS_PER_VALUE, THDS_PER_BLOCK, HAS_BEAMS, stream) \
size_t smem_sz = mmha::smem_size_in_bytes<T>(params, THDS_PER_VALUE, THDS_PER_BLOCK); \
dim3 grid(params.num_heads, params.batch_size); \
mmha::masked_groupedquery_attention_kernel<T, \
cudaFuncSetAttribute(mmha::masked_groupedquery_attention_kernel<T, \
Dh, Dh_MAX, THDS_PER_KEY, THDS_PER_VALUE, THDS_PER_BLOCK, HAS_BEAMS>, \
cudaFuncAttributeMaxDynamicSharedMemorySize, smem_sz); \
mmha::masked_groupedquery_attention_kernel<T, \
Dh, \
Dh_MAX, \
THDS_PER_KEY, \
Expand Down
9 changes: 6 additions & 3 deletions .../llama/decoder_masked_groupedquery_attention/decoder_masked_groupedquery_attention_256.cu
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -25,10 +25,13 @@
////////////////////////////////////////////////////////////////////////////////////////////////////

#define MGQA_LAUNCH_KERNEL( \
T, Dh, Dh_MAX, THDS_PER_KEY, THDS_PER_VALUE, THDS_PER_BLOCK, HAS_BEAMS, stream) \
size_t smem_sz = mmha::smem_size_in_bytes<T>(params, THDS_PER_VALUE, THDS_PER_BLOCK); \
T, Dh, Dh_MAX, THDS_PER_KEY, THDS_PER_VALUE, THDS_PER_BLOCK, HAS_BEAMS, stream) \
size_t smem_sz = mmha::smem_size_in_bytes<T>(params, THDS_PER_VALUE, THDS_PER_BLOCK); \
dim3 grid(params.num_heads, params.batch_size); \
mmha::masked_groupedquery_attention_kernel<T, \
cudaFuncSetAttribute(mmha::masked_groupedquery_attention_kernel<T, \
Dh, Dh_MAX, THDS_PER_KEY, THDS_PER_VALUE, THDS_PER_BLOCK, HAS_BEAMS>, \
cudaFuncAttributeMaxDynamicSharedMemorySize, smem_sz); \
mmha::masked_groupedquery_attention_kernel<T, \
Dh, \
Dh_MAX, \
THDS_PER_KEY, \
Expand Down
9 changes: 6 additions & 3 deletions ...s/llama/decoder_masked_groupedquery_attention/decoder_masked_groupedquery_attention_32.cu
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -25,10 +25,13 @@
////////////////////////////////////////////////////////////////////////////////////////////////////

#define MGQA_LAUNCH_KERNEL( \
T, Dh, Dh_MAX, THDS_PER_KEY, THDS_PER_VALUE, THDS_PER_BLOCK, HAS_BEAMS, stream) \
size_t smem_sz = mmha::smem_size_in_bytes<T>(params, THDS_PER_VALUE, THDS_PER_BLOCK); \
T, Dh, Dh_MAX, THDS_PER_KEY, THDS_PER_VALUE, THDS_PER_BLOCK, HAS_BEAMS, stream) \
size_t smem_sz = mmha::smem_size_in_bytes<T>(params, THDS_PER_VALUE, THDS_PER_BLOCK); \
dim3 grid(params.num_heads, params.batch_size); \
mmha::masked_groupedquery_attention_kernel<T, \
cudaFuncSetAttribute(mmha::masked_groupedquery_attention_kernel<T, \
Dh, Dh_MAX, THDS_PER_KEY, THDS_PER_VALUE, THDS_PER_BLOCK, HAS_BEAMS>, \
cudaFuncAttributeMaxDynamicSharedMemorySize, smem_sz); \
mmha::masked_groupedquery_attention_kernel<T, \
Dh, \
Dh_MAX, \
THDS_PER_KEY, \
Expand Down
9 changes: 6 additions & 3 deletions ...s/llama/decoder_masked_groupedquery_attention/decoder_masked_groupedquery_attention_48.cu
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -25,10 +25,13 @@
////////////////////////////////////////////////////////////////////////////////////////////////////

#define MGQA_LAUNCH_KERNEL( \
T, Dh, Dh_MAX, THDS_PER_KEY, THDS_PER_VALUE, THDS_PER_BLOCK, HAS_BEAMS, stream) \
size_t smem_sz = mmha::smem_size_in_bytes<T>(params, THDS_PER_VALUE, THDS_PER_BLOCK); \
T, Dh, Dh_MAX, THDS_PER_KEY, THDS_PER_VALUE, THDS_PER_BLOCK, HAS_BEAMS, stream) \
size_t smem_sz = mmha::smem_size_in_bytes<T>(params, THDS_PER_VALUE, THDS_PER_BLOCK); \
dim3 grid(params.num_heads, params.batch_size); \
mmha::masked_groupedquery_attention_kernel<T, \
cudaFuncSetAttribute(mmha::masked_groupedquery_attention_kernel<T, \
Dh, Dh_MAX, THDS_PER_KEY, THDS_PER_VALUE, THDS_PER_BLOCK, HAS_BEAMS>, \
cudaFuncAttributeMaxDynamicSharedMemorySize, smem_sz); \
mmha::masked_groupedquery_attention_kernel<T, \
Dh, \
Dh_MAX, \
THDS_PER_KEY, \
Expand Down
9 changes: 6 additions & 3 deletions ...s/llama/decoder_masked_groupedquery_attention/decoder_masked_groupedquery_attention_64.cu
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -25,10 +25,13 @@
////////////////////////////////////////////////////////////////////////////////////////////////////

#define MGQA_LAUNCH_KERNEL( \
T, Dh, Dh_MAX, THDS_PER_KEY, THDS_PER_VALUE, THDS_PER_BLOCK, HAS_BEAMS, stream) \
size_t smem_sz = mmha::smem_size_in_bytes<T>(params, THDS_PER_VALUE, THDS_PER_BLOCK); \
T, Dh, Dh_MAX, THDS_PER_KEY, THDS_PER_VALUE, THDS_PER_BLOCK, HAS_BEAMS, stream) \
size_t smem_sz = mmha::smem_size_in_bytes<T>(params, THDS_PER_VALUE, THDS_PER_BLOCK); \
dim3 grid(params.num_heads, params.batch_size); \
mmha::masked_groupedquery_attention_kernel<T, \
cudaFuncSetAttribute(mmha::masked_groupedquery_attention_kernel<T, \
Dh, Dh_MAX, THDS_PER_KEY, THDS_PER_VALUE, THDS_PER_BLOCK, HAS_BEAMS>, \
cudaFuncAttributeMaxDynamicSharedMemorySize, smem_sz); \
mmha::masked_groupedquery_attention_kernel<T, \
Dh, \
Dh_MAX, \
THDS_PER_KEY, \
Expand Down
9 changes: 6 additions & 3 deletions ...s/llama/decoder_masked_groupedquery_attention/decoder_masked_groupedquery_attention_80.cu
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -25,10 +25,13 @@
////////////////////////////////////////////////////////////////////////////////////////////////////

#define MGQA_LAUNCH_KERNEL( \
T, Dh, Dh_MAX, THDS_PER_KEY, THDS_PER_VALUE, THDS_PER_BLOCK, HAS_BEAMS, stream) \
size_t smem_sz = mmha::smem_size_in_bytes<T>(params, THDS_PER_VALUE, THDS_PER_BLOCK); \
T, Dh, Dh_MAX, THDS_PER_KEY, THDS_PER_VALUE, THDS_PER_BLOCK, HAS_BEAMS, stream) \
size_t smem_sz = mmha::smem_size_in_bytes<T>(params, THDS_PER_VALUE, THDS_PER_BLOCK); \
dim3 grid(params.num_heads, params.batch_size); \
mmha::masked_groupedquery_attention_kernel<T, \
cudaFuncSetAttribute(mmha::masked_groupedquery_attention_kernel<T, \
Dh, Dh_MAX, THDS_PER_KEY, THDS_PER_VALUE, THDS_PER_BLOCK, HAS_BEAMS>, \
cudaFuncAttributeMaxDynamicSharedMemorySize, smem_sz); \
mmha::masked_groupedquery_attention_kernel<T, \
Dh, \
Dh_MAX, \
THDS_PER_KEY, \
Expand Down
9 changes: 6 additions & 3 deletions ...s/llama/decoder_masked_groupedquery_attention/decoder_masked_groupedquery_attention_96.cu
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -25,10 +25,13 @@
////////////////////////////////////////////////////////////////////////////////////////////////////

#define MGQA_LAUNCH_KERNEL( \
T, Dh, Dh_MAX, THDS_PER_KEY, THDS_PER_VALUE, THDS_PER_BLOCK, HAS_BEAMS, stream) \
size_t smem_sz = mmha::smem_size_in_bytes<T>(params, THDS_PER_VALUE, THDS_PER_BLOCK); \
T, Dh, Dh_MAX, THDS_PER_KEY, THDS_PER_VALUE, THDS_PER_BLOCK, HAS_BEAMS, stream) \
size_t smem_sz = mmha::smem_size_in_bytes<T>(params, THDS_PER_VALUE, THDS_PER_BLOCK); \
dim3 grid(params.num_heads, params.batch_size); \
mmha::masked_groupedquery_attention_kernel<T, \
cudaFuncSetAttribute(mmha::masked_groupedquery_attention_kernel<T, \
Dh, Dh_MAX, THDS_PER_KEY, THDS_PER_VALUE, THDS_PER_BLOCK, HAS_BEAMS>, \
cudaFuncAttributeMaxDynamicSharedMemorySize, smem_sz); \
mmha::masked_groupedquery_attention_kernel<T, \
Dh, \
Dh_MAX, \
THDS_PER_KEY, \
Expand Down