Skip to content

Remove RoPE fallback vllm custom ops #13668

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
jinyouzhi wants to merge 2 commits into
base: main
Choose a base branch
from
Open

Remove RoPE fallback vllm custom ops #13668

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
9b2b9ae
Remove RoPE fallback vllm custom ops
jinyouzhi Nov 20, 2025
83e8f60
Merge branch 'main' into rope_clean
jinyouzhi Nov 21, 2025
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
54 changes: 14 additions & 40 deletions python/sglang/srt/layers/rotary_embedding.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -113,18 +113,6 @@ def __init__(
if not _is_cuda:
cache = cache.to(dtype)

if (
(not (_is_cuda or _is_npu) or self.head_size not in [64, 128, 256, 512])
and not (_is_cpu and _is_cpu_amx_available)
and not (_is_xpu)
):
from vllm._custom_ops import rotary_embedding

self.use_fallback_kernel = True
self.fallback_rotary_embedding = rotary_embedding
else:
self.use_fallback_kernel = False

self.cos_sin_cache: torch.Tensor
self.register_buffer("cos_sin_cache", cache, persistent=False)

Expand Down Expand Up @@ -276,34 +264,20 @@ def forward_cuda(
offsets: Optional[torch.Tensor] = None,
fused_set_kv_buffer_arg: Optional[FusedSetKVBufferArg] = None,
) -> Tuple[torch.Tensor, torch.Tensor]:
if not self.use_fallback_kernel:
apply_rope_with_cos_sin_cache_inplace(
positions=positions,
query=query,
key=key,
head_size=self.head_size,
cos_sin_cache=self.cos_sin_cache,
is_neox=self.is_neox_style,
# Compatible with old sgl-kernel
**(
dict(fused_set_kv_buffer_arg=fused_set_kv_buffer_arg)
if fused_set_kv_buffer_arg is not None
else {}
),
)
else:
assert (
fused_set_kv_buffer_arg is None
), "save kv cache is not supported for vllm_rotary_embedding."
self.cos_sin_cache = self.cos_sin_cache.to(query.device, dtype=query.dtype)
self.fallback_rotary_embedding(
positions,
query,
key,
self.head_size,
self.cos_sin_cache,
self.is_neox_style,
)
apply_rope_with_cos_sin_cache_inplace(
positions=positions,
query=query,
key=key,
head_size=self.head_size,
cos_sin_cache=self.cos_sin_cache,
is_neox=self.is_neox_style,
# Compatible with old sgl-kernel
**(
dict(fused_set_kv_buffer_arg=fused_set_kv_buffer_arg)
if fused_set_kv_buffer_arg is not None
else {}
),
)
return query, key

def extra_repr(self) -> str:
Expand Down
Loading