Skip to content

Optimizing kernel performance #3452

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

Open
wants to merge 1 commit into
base: develop
Choose a base branch
from
Open

Optimizing kernel performance #3452

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
25 changes: 23 additions & 2 deletions fastdeploy/model_executor/layers/moe/triton_moe_kernels.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -21,6 +21,18 @@
)


@paddle_use_triton_v2()
def write_zeros_to_output(c_ptr, stride_cm, stride_cn, pid_n, N, offs_token,
token_mask, BLOCK_SIZE_M, BLOCK_SIZE_N,
compute_type):
accumulator = tl.zeros((BLOCK_SIZE_M, BLOCK_SIZE_N), dtype=compute_type)
offs_cn = pid_n * BLOCK_SIZE_N + tl.arange(0, BLOCK_SIZE_N)
c_ptrs = c_ptr + stride_cm * offs_token[:, None] + stride_cn * offs_cn[
None, :]
c_mask = token_mask[:, None] & (offs_cn[None, :] < N)
tl.store(c_ptrs, accumulator, mask=c_mask)


@paddle_use_triton_v2()
def fused_moe_kernel_paddle(
a_ptr,
Expand Down Expand Up @@ -108,11 +120,20 @@ def fused_moe_kernel_paddle(
offs_token = tl.load(sorted_token_ids_ptr + offs_token_id)
token_mask = offs_token < num_valid_tokens

off_experts = tl.load(expert_ids_ptr + pid_m)
if off_experts == -1:
# -----------------------------------------------------------
# Write back zeros to the output when the expert is not
# in the current expert parallel rank.
write_zeros_to_output(c_ptr, stride_cm, stride_cn, pid_n, N,
offs_token, token_mask, BLOCK_SIZE_M,
BLOCK_SIZE_N, compute_type)
return

offs_bn = (pid_n * BLOCK_SIZE_N + tl.arange(0, BLOCK_SIZE_N)) % N
offs_k = tl.arange(0, BLOCK_SIZE_K)
a_ptrs = a_ptr + (offs_token[:, None] // top_k * stride_am + offs_k[None, :] * stride_ak)

off_experts = tl.load(expert_ids_ptr + pid_m)
a_ptrs = a_ptr + (offs_token[:, None] // top_k * stride_am + offs_k[None, :] * stride_ak)
b_ptrs = b_ptr + off_experts * stride_be + (offs_k[:, None] * stride_bk + offs_bn[None, :] * stride_bn)

if use_int8_w8a16:
Expand Down
Loading