Skip to content

[KVCache] Attention func accepting over-padded qkv and output NDArray #17401

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
tqchen merged 1 commit into from
Sep 22, 2024
Merged

[KVCache] Attention func accepting over-padded qkv and output NDArray #17401

Changes from all commits
Commits
Show all changes
1 commit
Select commit
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
17 changes: 13 additions & 4 deletions src/runtime/relax_vm/paged_kv_cache.cc
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1750,7 +1750,7 @@ class PagedAttentionKVCacheObj : public AttentionKVCacheObj {
for (int64_t seq_id = 0; seq_id < cur_batch_size_; ++seq_id) {
total_seq_length += cur_append_lengths_[seq_id];
}
CHECK_EQ(total_seq_length, qkv_data->shape[0]);
CHECK_LE(total_seq_length, qkv_data->shape[0]);
// Sync the copy stream and the compute stream.
ComputeStreamWaitForCopyStream();
// The auxiliary data structure on device must have been synchronized.
Expand All @@ -1762,12 +1762,21 @@ class PagedAttentionKVCacheObj : public AttentionKVCacheObj {
qkv_data->dtype);
NDArray v_data = temp_attn_v_device_.CreateView({total_seq_length, num_kv_heads_, head_dim_},
qkv_data->dtype);

NDArray qkv_data_view = qkv_data;
NDArray o_data_view = o_data;
if (total_seq_length != qkv_data->shape[0]) {
qkv_data_view = qkv_data.CreateView(
{total_seq_length, qkv_data->shape[1], qkv_data->shape[2]}, qkv_data->dtype);
o_data_view =
o_data.CreateView({total_seq_length, num_qo_heads_, head_dim_}, qkv_data->dtype);
}
// Part 2. Split fused qkv and apply rotary embedding to q/k data.
if (!rope_ext_factors_.defined()) {
f_split_rotary_(qkv_data, q_rope_position_map_view_, q_data, k_data, v_data,
f_split_rotary_(qkv_data_view, q_rope_position_map_view_, q_data, k_data, v_data,
static_cast<int>(rope_mode_ == RoPEMode::kNormal));
} else {
f_split_rotary_(qkv_data, q_rope_position_map_view_, q_data, k_data, v_data,
f_split_rotary_(qkv_data_view, q_rope_position_map_view_, q_data, k_data, v_data,
rope_ext_factors_.value());
}

Expand All @@ -1776,7 +1785,7 @@ class PagedAttentionKVCacheObj : public AttentionKVCacheObj {
f_transpose_append_(pages_[local_layer_id], k_data, v_data, append_position_map_view_);
}
// Part 4: perform attention
AttentionInternal(layer_id, q_data, k_data, v_data, o_data, attn_score_scaling_factor);
AttentionInternal(layer_id, q_data, k_data, v_data, o_data_view, attn_score_scaling_factor);
// Part 5. Append k/v data to kv-cache if flag "append_before_attn" is not set.
if (!append_before_attn_) {
f_transpose_append_(pages_[local_layer_id], k_data, v_data, append_position_map_view_);
Expand Down