Skip to content

GitLab

"git@developer.sourcefind.cn:OpenDAS/autoawq.git" did not exist on "abdc726ca30f9b24837cbb6a6ee92ad441a4a10a"
Commit 8e26a612 authored by letaoqin's avatar letaoqin
Browse files

Merge branch 'mha-train-develop' into mha-train-bias

parents 18b1aa04 927cfb23
Hide whitespace changes
Inline Side-by-side
Showing with 8 additions and 8 deletions
include/ck/tensor_operation/gpu/device/impl/device_batched_mha_fwd_xdl_cshuffle_v1.hpp
View file @ 8e26a612
...@@ -113,8 +113,8 @@ __global__ void ...@@ -113,8 +113,8 @@ __global__ void
p_b_grid + b_batch_offset, p_b_grid + b_batch_offset,
p_b1_grid + b1_batch_offset, p_b1_grid + b1_batch_offset,
p_c_grid + c_batch_offset, p_c_grid + c_batch_offset,
nullptr ? nullptr : p_z_grid + z_batch_offset, p_z_grid == nullptr ? nullptr : p_z_grid + z_batch_offset,
nullptr ? nullptr : p_lse_grid + lse_batch_offset, p_lse_grid == nullptr ? nullptr : p_lse_grid + lse_batch_offset,
p_shared, p_shared,
a_element_op, a_element_op,
b_element_op, b_element_op,
...@@ -142,8 +142,8 @@ __global__ void ...@@ -142,8 +142,8 @@ __global__ void
p_b_grid + b_batch_offset, p_b_grid + b_batch_offset,
p_b1_grid + b1_batch_offset, p_b1_grid + b1_batch_offset,
p_c_grid + c_batch_offset, p_c_grid + c_batch_offset,
nullptr ? nullptr : p_z_grid + z_batch_offset, p_z_grid == nullptr ? nullptr : p_z_grid + z_batch_offset,
nullptr ? nullptr : p_lse_grid + lse_batch_offset, p_lse_grid == nullptr ? nullptr : p_lse_grid + lse_batch_offset,
p_shared, p_shared,
a_element_op, a_element_op,
b_element_op, b_element_op,
......
include/ck/tensor_operation/gpu/device/impl/device_batched_mha_fwd_xdl_cshuffle_v2.hpp
View file @ 8e26a612
...@@ -117,8 +117,8 @@ __global__ void ...@@ -117,8 +117,8 @@ __global__ void
p_b_grid + b_batch_offset, p_b_grid + b_batch_offset,
p_b1_grid + b1_batch_offset, p_b1_grid + b1_batch_offset,
p_c_grid + c_batch_offset, p_c_grid + c_batch_offset,
nullptr ? nullptr : p_z_grid + z_batch_offset, p_z_grid == nullptr ? nullptr : p_z_grid + z_batch_offset,
nullptr ? nullptr : p_lse_grid + lse_batch_offset, p_lse_grid == nullptr ? nullptr : p_lse_grid + lse_batch_offset,
p_shared, p_shared,
a_element_op, a_element_op,
b_element_op, b_element_op,
...@@ -148,8 +148,8 @@ __global__ void ...@@ -148,8 +148,8 @@ __global__ void
p_b_grid + b_batch_offset, p_b_grid + b_batch_offset,
p_b1_grid + b1_batch_offset, p_b1_grid + b1_batch_offset,
p_c_grid + c_batch_offset, p_c_grid + c_batch_offset,
nullptr ? nullptr : p_z_grid + z_batch_offset, p_z_grid == nullptr ? nullptr : p_z_grid + z_batch_offset,
nullptr ? nullptr : p_lse_grid + lse_batch_offset, p_lse_grid == nullptr ? nullptr : p_lse_grid + lse_batch_offset,
p_shared, p_shared,
a_element_op, a_element_op,
b_element_op, b_element_op,
......
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment