[serge] integration failure triage - 2026-08-12

[github-actions]

Automated integration-failure triage for the daily CI window 2026-08-06 → 2026-08-12.

This issue was generated by AI-assisted automation. The grouping, summaries, and recommended follow-up can be incomplete or misleading; verify the failures before acting.

Serge dispatched one task per failure group below — each opens or updates its own PR on a serge/fix/itf-<fingerprint> branch. This table is refreshed in place as Serge runs: a group links its #<pr> when opened, shows 🚫 no fix when Serge found no safe change, ⚠️ task failed on error, or (pending) while still running (a late PR links on the next nightly run).

Dispatched failure groups

Model	Error	Occurrences	PR
deepseek_vl	other — other (3)	3	🚫 no fix
got_ocr2	other — other (2)	2	#47937
emu3	output_mismatch — list output differs (1)	1	#47938

Not dispatched — environment / dependency

These groups were triaged but NOT handed to Serge: their failure mode is a property of the runner or the environment, so no minimal source patch can fix them. They need a human (runner capacity, a dependency pin).

Model	Error	Occurrences	Why not dispatched
mamba2	OOM — other (8)	8	runner ran out of device memory — needs runner capacity, not a patch
deepseek_v2	OOM — other (5)	5	runner ran out of device memory — needs runner capacity, not a patch
emu3	OOM — other (5)	5	runner ran out of device memory — needs runner capacity, not a patch
bamba	OOM — other (4)	4	runner ran out of device memory — needs runner capacity, not a patch
glm	OOM — other (4)	4	runner ran out of device memory — needs runner capacity, not a patch
llava	OOM — other (4)	4	runner ran out of device memory — needs runner capacity, not a patch
olmo	OOM — other (3)	3	runner ran out of device memory — needs runner capacity, not a patch
cohere2_vision	OOM — other (2)	2	runner ran out of device memory — needs runner capacity, not a patch
gemma4	OOM — other (2)	2	runner ran out of device memory — needs runner capacity, not a patch
glm4_moe	OOM — other (2)	2	runner ran out of device memory — needs runner capacity, not a patch
glm4_moe_lite	OOM — other (2)	2	runner ran out of device memory — needs runner capacity, not a patch
olmo3	OOM — other (2)	2	runner ran out of device memory — needs runner capacity, not a patch
pi0	OOM — other (2)	2	runner ran out of device memory — needs runner capacity, not a patch
qwen3_vl_moe	OOM — other (2)	2	runner ran out of device memory — needs runner capacity, not a patch
cwm	OOM — other (1)	1	runner ran out of device memory — needs runner capacity, not a patch
deepseek_vl_hybrid	OOM — other (1)	1	runner ran out of device memory — needs runner capacity, not a patch
exaone4	OOM — other (1)	1	runner ran out of device memory — needs runner capacity, not a patch
exaone4_5	OOM — other (1)	1	runner ran out of device memory — needs runner capacity, not a patch

Outcome recap

Why each group that opened no PR ended without one, and what it cost — surfaced here from the Serge dashboard.

Group	Reason	LLM	Tokens (in / out)
deepseek_vl	[not_fixed] GPU verification did not confirm the fix (not_fixed): the patch did not turn the targeted tests green.	moonshotai/Kimi-K2.7-Code	1,675,580 / 3,591

Generated 2026-08-12T23:12:08+00:00.