Meta-Eval/EvalHarness
3
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

in-flight label: '2 gen · 94 wait' instead of '2/96'

Browse Source 
The slash form read as a fraction and kept inviting 'why is the
denominator growing' -- it is the pipeline depth (tokenizing +
gate-queued + generating), bounded by the global semaphore, while the
gate-admitted count is the real server load.

Co-Authored-By: Claude <noreply@anthropic.com>
This commit is contained in:
sora 2026-09-15 02:52:13 +00:00
parent d83c2cc1df
commit 8cc5f6a9c6
1 changed files with 3 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
evalharness/progress/rich_terminal.py
View File

@ -215,7 +215,9 @@ class RichTerminalProgress:
# 'the gate is not working' # 'the gate is not working'
if self.admitted is None: if self.admitted is None:
return str(self.inflight) return str(self.inflight)
return f'{self.admitted}/{self.inflight}' # 'N gen · M wait': N admitted by the gate (real server load),
# M held by the pipeline (tokenizing or gate-queued)
return f'{self.admitted} gen · {max(0, self.inflight - self.admitted)} wait'
def set_admitted(self, n: int): def set_admitted(self, n: int):
"""Pooled runs: requests actually admitted by the gate.""" """Pooled runs: requests actually admitted by the gate."""