Skip to content

[KV Sharing] Raise error if using eagle with fast prefill #24350

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
Sep 6, 2025
Merged

[KV Sharing] Raise error if using eagle with fast prefill #24350

Show file tree
Hide file tree
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
18 changes: 18 additions & 0 deletions vllm/config/__init__.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -3665,6 +3665,24 @@ def __post_init__(self):
" Disabling `torch.compile`.")
self.compilation_config.level = CompilationLevel.NO_COMPILATION

if self.cache_config.kv_sharing_fast_prefill:
if not envs.VLLM_USE_V1:
raise NotImplementedError(
"Fast prefill optimization for KV sharing is not supported "
"in V0 currently.")

if self.speculative_config is not None and \
self.speculative_config.use_eagle():
raise NotImplementedError(
"Fast prefill optimization for KV sharing is not "
"compatible with EAGLE as EAGLE requires correct logits "
"for all tokens while fast prefill gives incorrect logits "
"for prompt tokens.")

logger.warning_once(
"--kv-sharing-fast-prefill requires changes on model side for "
"correctness and to realize prefill savings. ")

if ((not envs.VLLM_USE_V1) and self.lora_config is not None
and self.compilation_config.level
!= CompilationLevel.NO_COMPILATION):
Expand Down
7 changes: 0 additions & 7 deletions vllm/config/cache.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -145,19 +145,12 @@ def __post_init__(self) -> None:

self._verify_cache_dtype()
self._verify_prefix_caching()
self._verify_kv_sharing_fast_prefill()

def metrics_info(self):
# convert cache_config to dict(key: str, value: str) for prometheus
# metrics info
return {key: str(value) for key, value in self.__dict__.items()}

def _verify_kv_sharing_fast_prefill(self) -> None:
if self.kv_sharing_fast_prefill and not envs.VLLM_USE_V1:
raise NotImplementedError(
"Fast prefill optimization for KV sharing is not supported "
"in V0 currently.")

@model_validator(mode='after')
def _verify_args(self) -> Self:
if self.cpu_offload_gb < 0:
Expand Down