Skip to content

[MM][Perf] Merge Q/K split to simplify AscendApplyRotaryEmb for better performance #5799

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
wangxiyuan merged 2 commits into from
Jan 13, 2026
Merged

[MM][Perf] Merge Q/K split to simplify AscendApplyRotaryEmb for better performance #5799

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
841f12b
Merge Q/K split to simplify AscendApplyRotaryEmb for better performance
shen-shanshan Jan 12, 2026
36e168f
fix
shen-shanshan Jan 12, 2026
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: 24 additions & 26 deletions vllm_ascend/ops/rotary_embedding.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -18,7 +18,6 @@
import math
from typing import Optional, Tuple

import einops
import torch
import torch_npu
from vllm.model_executor.layers.rotary_embedding import (
Expand All @@ -28,7 +27,8 @@

from vllm_ascend.platform import NPUPlatform
from vllm_ascend.utils import (AscendDeviceType, enable_custom_op,
get_ascend_device_type, has_rope, is_vl_model)
get_ascend_device_type, has_rope, is_vl_model,
vllm_version_is)

# Currently, rope ops used on npu requires detached cos && sin as inputs.
# However, RotaryEmbedding in vllm use cos_sin_cache as a whole variable.
Expand Down Expand Up @@ -580,37 +580,35 @@ def forward_oot(
cos: torch.Tensor,
sin: torch.Tensor,
) -> torch.Tensor:
head_dim = x.shape[-1]

origin_dtype = x.dtype
if self.enable_fp32_compute:
x = x.float()
cos = cos.float()
sin = sin.float()
if vllm_version_is('0.13.0'):
origin_shape = x.shape
origin_dtype = x.dtype
if len(origin_shape) == 3:
x = x.unsqueeze(0)
if self.enable_fp32_compute:
x = x.float()
cos = cos.float()
sin = sin.float()
else:
x, cos, sin, origin_shape, origin_dtype = self._pre_process(
x, cos, sin)

head_dim = x.shape[-1]
# cos, sin: [seq_len, head_dim // 2]
cos = torch.cat((cos, cos), dim=-1)
sin = torch.cat((sin, sin), dim=-1)
# cos, sin: [1, seq_len, 1, head_dim]
cos = cos.reshape(1, -1, 1, head_dim)
sin = sin.reshape(1, -1, 1, head_dim)

if len(x.shape) == 3:
# x: [seq_len, num_heads, head_size]
x = x.unsqueeze(0)
# x: [1, seq_len, num_heads, head_size]
output = torch_npu.npu_rotary_mul(x, cos, sin).squeeze(0)
output = torch_npu.npu_rotary_mul(x, cos, sin)

if vllm_version_is('0.13.0'):
if len(origin_shape) == 3:
output = output.squeeze(0)
if self.enable_fp32_compute:
output = output.to(origin_dtype)
else:
assert len(x.shape) == 4
# x: [2 * b, s, head, head_dim]
qk = einops.rearrange(
x, "(two b) s head head_dim -> b s two head head_dim", two=2)
# q, k: [b, s, head, head_dim]
q, k = qk[:, :, 0], qk[:, :, 1]
q = torch_npu.npu_rotary_mul(q, cos, sin)
k = torch_npu.npu_rotary_mul(k, cos, sin)
output = torch.cat([q, k], dim=0)

This comment was marked as resolved.

Sign in to view


if self.enable_fp32_compute:
output = output.to(origin_dtype)
output = self._post_process(output, origin_shape, origin_dtype)

return output