Skip to content

Eliminate c10::optional usage in vllm_flash_attn #38

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 1 commit into from
Jan 17, 2025
Merged

Eliminate c10::optional usage in vllm_flash_attn #38

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
14 changes: 7 additions & 7 deletions csrc/flash_attn/flash_api.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -346,14 +346,14 @@ mha_fwd_sparse(at::Tensor &q, // batch_size x seqlen_q x num_heads x hea
const at::Tensor &block_offset,
const at::Tensor &column_count,
const at::Tensor &column_index,
const c10::optional<at::Tensor> &out_, // batch_size x seqlen_q x num_heads x head_size
const c10::optional<at::Tensor> &alibi_slopes_, // num_heads or batch_size x num_heads
const std::optional<at::Tensor> &out_, // batch_size x seqlen_q x num_heads x head_size
const std::optional<at::Tensor> &alibi_slopes_, // num_heads or batch_size x num_heads
const double p_dropout,
const double softmax_scale,
bool is_causal,
const double softcap,
const bool return_softmax,
c10::optional<at::Generator> gen_) {
std::optional<at::Generator> gen_) {

auto dprops = at::cuda::getCurrentDeviceProperties();
// bool is_sm75 = dprops->major == 7 && dprops->minor == 5;
Expand Down Expand Up @@ -515,11 +515,11 @@ mha_varlen_fwd_sparse(at::Tensor &q, // total_q x num_heads x head_size, total_
const at::Tensor &block_offset,
const at::Tensor &column_count,
const at::Tensor &column_index,
const c10::optional<at::Tensor> &out_, // total_q x num_heads x head_size, total_k := \sum_{i=0}^{b} s_i
const std::optional<at::Tensor> &out_, // total_q x num_heads x head_size, total_k := \sum_{i=0}^{b} s_i
const at::Tensor &cu_seqlens_q, // b+1
const at::Tensor &cu_seqlens_k, // b+1
const c10::optional<at::Tensor> &seqused_k, // b. If given, only this many elements of each batch element's keys are used.
const c10::optional<at::Tensor> &alibi_slopes_, // num_heads or b x num_heads
const std::optional<at::Tensor> &seqused_k, // b. If given, only this many elements of each batch element's keys are used.
const std::optional<at::Tensor> &alibi_slopes_, // num_heads or b x num_heads
int64_t max_seqlen_q,
const int64_t max_seqlen_k,
const double p_dropout,
Expand All @@ -528,7 +528,7 @@ mha_varlen_fwd_sparse(at::Tensor &q, // total_q x num_heads x head_size, total_
bool is_causal,
const double softcap,
const bool return_softmax,
c10::optional<at::Generator> gen_) {
std::optional<at::Generator> gen_) {

auto dprops = at::cuda::getCurrentDeviceProperties();
// bool is_sm75 = dprops->major == 7 && dprops->minor == 5;
Expand Down