Skip to content

block_multihead_attention support V100 GQA #68104

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 9 commits into from
Sep 11, 2024
Merged

block_multihead_attention support V100 GQA #68104

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
9 commits
Select commit Hold shift + click to select a range
1782350
test v100 no flash
zhink Aug 16, 2024
8123aaf
test v100 no flash 2
zhink Aug 16, 2024
b0a8bc3
test v100 no flash 3
zhink Aug 16, 2024
352b7a9
Merge commit 'aea5c200a4f665fc8a462f096a6532aa88757e58' into v100
zhink Aug 30, 2024
4206ceb
block attention will by variable_length_memory_efficient_attention wh…
zhink Sep 1, 2024
a43e114
block attention support encoder stage with GQA when sm < 80
zhink Sep 9, 2024
78d03dd
Merge branch 'develop' into v100
zhink Sep 9, 2024
9e05345
处理冲突
zhink Sep 9, 2024
1d00758
codestyle
zhink Sep 9, 2024
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
50 changes: 32 additions & 18 deletions paddle/phi/kernels/fusion/gpu/block_attn.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -4038,10 +4038,10 @@ __global__ void fusedQKV_transpose_split_kernel(T *q_buf,
const int seq_len,
const int pre_cache_length,
const int token_num,
const int head_num,
const int q_head_num,
const int kv_head_num,
const int size_per_head) {
const int32_t hidden_size = head_num * size_per_head;
const int32_t fused_hidden_size = 3 * hidden_size;
const int fused_hidden_size = (q_head_num + 2 * kv_head_num) * size_per_head;
int64_t global_thread_idx = blockDim.x * blockIdx.x + threadIdx.x;
using LoadT = phi::AlignedVector<T, VecSize>;
LoadT src_vec;
Expand All @@ -4059,22 +4059,33 @@ __global__ void fusedQKV_transpose_split_kernel(T *q_buf,
if (seq_lens[target_batch_id] == 0) continue;
const int32_t seq_id = ori_token_idx % seq_len;

const int32_t qkv_id = bias_idx / hidden_size;
const int32_t head_id = (linear_index % hidden_size) / size_per_head;
const int32_t head_id = bias_idx / size_per_head;
const int32_t size_id = linear_index % size_per_head;

const int tmp_max_len_this_time =
max_len_this_time + (qkv_id == 0 ? 0 : pre_cache_length);
const int tmp_seq_id = qkv_id == 0 ? seq_id : seq_id + pre_cache_length;
const int write_idx =
target_batch_id * head_num * tmp_max_len_this_time * size_per_head +
head_id * tmp_max_len_this_time * size_per_head +
tmp_seq_id * size_per_head + size_id;
if (qkv_id == 0) {
max_len_this_time + (head_id < q_head_num ? 0 : pre_cache_length);
const int tmp_seq_id =
head_id < q_head_num ? seq_id : seq_id + pre_cache_length;

if (head_id < q_head_num) {
const int write_idx =
target_batch_id * q_head_num * tmp_max_len_this_time * size_per_head +
head_id * tmp_max_len_this_time * size_per_head +
tmp_seq_id * size_per_head + size_id;
phi::Store<T, VecSize>(src_vec, &q_buf[write_idx]);
} else if (qkv_id == 1) {
} else if (head_id < q_head_num + kv_head_num) {
const int write_idx =
target_batch_id * kv_head_num * tmp_max_len_this_time *
size_per_head +
(head_id - q_head_num) * tmp_max_len_this_time * size_per_head +
tmp_seq_id * size_per_head + size_id;
phi::Store<T, VecSize>(src_vec, &k_buf[write_idx]);
} else {
const int write_idx = target_batch_id * kv_head_num *
tmp_max_len_this_time * size_per_head +
(head_id - q_head_num - kv_head_num) *
tmp_max_len_this_time * size_per_head +
tmp_seq_id * size_per_head + size_id;
phi::Store<T, VecSize>(src_vec, &v_buf[write_idx]);
}
}
Expand All @@ -4093,12 +4104,14 @@ void qkv_transpose_split(
const int *seq_lens,
const int token_num,
const int batch_size,
const int head_num,
const int q_head_num,
const int kv_head_num,
const int max_len_this_time,
const int seq_len,
const int pre_cache_length,
const int size_per_head) {
int32_t elem_cnt = token_num * head_num * size_per_head * 3;
int32_t elem_cnt = token_num * (q_head_num + kv_head_num * 2) * size_per_head;

constexpr int PackSize = VEC_16B / sizeof(T);
PADDLE_ENFORCE_EQ(size_per_head % PackSize,
0,
Expand All @@ -4123,11 +4136,12 @@ void qkv_transpose_split(
seq_len,
pre_cache_length,
token_num,
head_num,
q_head_num,
kv_head_num,
size_per_head);
if (pre_key_cache) {
// stage 2: write pre_cache to kv_buf
elem_cnt = batch_size * head_num * pre_cache_length * size_per_head * 2;
elem_cnt = batch_size * q_head_num * pre_cache_length * size_per_head * 2;
pack_num = elem_cnt / PackSize;
GetNumBlocks(pack_num, &grid_size);
write_pre_cahe_to_kv_buffer<T, PackSize>
Expand All @@ -4138,7 +4152,7 @@ void qkv_transpose_split(
seq_lens,
batch_size,
pre_cache_length,
head_num,
kv_head_num,
size_per_head,
max_len_this_time,
elem_cnt);
Expand Down
7 changes: 2 additions & 5 deletions paddle/phi/kernels/fusion/gpu/block_multi_head_attention_kernel.cu
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -570,12 +570,7 @@ void DispatchWithDtype(
// Reshape fmha_buf back (to 2-D), to not affect following codes.
fmha_buf.Resize(fmha_shape);
} else {
// NOTE: not support gqa
if (sm < 80 && !use_pre_cache) {
if (q_num_head != kv_num_head) {
PADDLE_THROW(common::errors::Unimplemented(
"Only supported MHA on Volta/Turing(sm < 80) now."));
}
qkv_transpose_split<T>(
dev_ctx,
q_trans.data<T>(),
Expand All @@ -589,6 +584,7 @@ void DispatchWithDtype(
token_num,
bsz,
q_num_head,
kv_num_head,
max_enc_len_this_time_data,
max_seq_len,
pre_cache_length,
Expand All @@ -607,6 +603,7 @@ void DispatchWithDtype(
token_num,
bsz,
q_num_head,
kv_num_head,
max_enc_len_this_time_data,
max_seq_len,
pre_cache_length,
Expand Down
16 changes: 1 addition & 15 deletions test/legacy_test/test_block_multihead_attention_gqa.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -21,6 +21,7 @@
create_attn_mask,
get_cuda_version,
get_padding_offset,
is_sm_supported,
remove_padding,
)

Expand All @@ -34,21 +35,6 @@
np.random.seed(2024)


is_sm8x = (
core.is_compiled_with_cuda()
and paddle.device.cuda.get_device_capability()[0] == 8
and paddle.device.cuda.get_device_capability()[1] >= 0
)

is_sm9x = (
core.is_compiled_with_cuda()
and paddle.device.cuda.get_device_capability()[0] == 9
and paddle.device.cuda.get_device_capability()[1] >= 0
)

is_sm_supported = is_sm9x or is_sm8x


def naive_attention_impl(
query,
key,
Expand Down