[Core][Refactor][Diffusion] refactor request scheduler's admission wait policy (#5843)

Signed-off-by: Huang, Zeyu <11222265+fhfuih@users.noreply.github.com>
Co-authored-by: Hongsheng Liu <liuhongsheng4@huawei.com>
This commit is contained in:
Zeyu Huang | 黃澤宇andHongsheng Liu authored and GitHub committed 2026-08-10 15:06:27 +08:00
1 parent c588208cc6
commit 33dd9a5a3e
13 files changed
+253 -54

No files matched your search

+10
View File
@@ -31,6 +31,16 @@ class DiffusionRequestStatus(enum.IntEnum):
return status >= DiffusionRequestStatus.FINISHED_COMPLETED
@dataclass(frozen=True)
class _AdmissionWaitDecision:
"""Internal scheduler policy for delaying the next admission wave."""
should_wait: bool
deadline: float | None = None
stable_window_s: float = 0.0
max_batch: int = 1
@dataclass(frozen=True, eq=True)
class StepBatchSamplingParamsKey:
"""Denoise step level Batch-compatibility key derived from ``OmniDiffusionSamplingParams``.