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

Auto-stream threshold 8192→100000 (streaming strips all thinking params on GLM gateway; non-streaming works for 32k); EVALHARNESS_NO_AUTOSTREAM=1 to disable entirely

Browse Source 
Co-Authored-By: Claude <noreply@anthropic.com>
This commit is contained in:
sora 2026-09-11 09:50:03 +00:00
parent f5aaddb870
commit ed675e790f
1 changed files with 2 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

3
evalharness/model/adapter.py
View File

@ -205,7 +205,8 @@ class OpenAICompatible(ModelAdapter):
if stream:
out = await self._post_stream_perf(
f'{self.api_base}/chat/completions', payload, headers, t0)
elif int(payload.get('max_tokens') or 0) > 8192:
elif int(payload.get('max_tokens') or 0) > 100000 \
and not os.environ.get('EVALHARNESS_NO_AUTOSTREAM'):
# long generation: stream and aggregate (gateway-safe).
# Some gateways drop chat_template_kwargs on the STREAM
# path only (non-stream honors it) -- append the /no_think