[serge] integration failure triage - 2026-10-01

Automated integration-failure triage for the daily CI window 2026-09-11 → 2026-10-01.

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).

The Task column links Serge's own page for the run — the steps it went through, what each tool cost, and the full task text. means it is not public: the host is VPN-internal and the page needs a Serge login. A task links once Serge has accepted it, so the column fills in on the first refresh after dispatch.

Dispatched failure groups

Model	Error	Occurrences	PR	Task
minicpmv4_7 (regressed by PR #48979)	mixed — other (10)	10	no fix	task
qwen3_omni_moe (regressed by PR #48714)	mixed — other (2)	2	no fix	task
generation	cuda_runtime — other (12)	12	no fix	task
blt	import_or_config — other (4)	4	no fix	task
cwm	OOM — other (2)	2	no fix	task
efficientnet	output_mismatch — tensor values differ (2)	2	#47840	task
fuyu	other — other (2)	2	#49244	task
pvt_v2	output_mismatch — tensor values differ (2)	2	#49243	task
deepseek_vl_hybrid	OOM — other (1)	1	⚠️ task failed	task
exaone4	other — other (1)	1	⚠️ task failed	task

14 group(s) skipped — stopped failing before this run (deepseek_v32, cohere2_vision, cohere2_vision, llama4 +10 more). Not dispatched and not awaiting a human.

Outcome recap

Why each group that opened no PR ended without one, and what it cost — surfaced here from the Serge dashboard. Failing tests links straight to each test's dashboard view, since these are the groups a human has to pick up.

Group	Reason	LLM	Tokens (in / out)	Failing tests
minicpmv4_7 (regressed by PR #48979)	[reproduced] GPU reproduce + classify: this is an ENVIRONMENT issue (The tracebacks show the Hub repository 'openbmb/MiniCPM-V-4_7' cannot be found, which is an external checkpoint availability issue, not a bug in the library or test code.) — no source patch can fix it, so no investigation and no PR.	Qwen/Qwen3.8-27B:novita	— / —	test_small_model_logits · test_small_model_video_generation · test_small_model_vision_generation · +2 more
qwen3_omni_moe (regressed by PR #48714)	[not_reproduced] GPU reproduce: the targeted tests did NOT fail at the base commit (not_reproduced) — the failure is stale, already fixed, flaky, or environment-specific. Skipping this group; no investigation, no PR.	Qwen/Qwen3.8-27B:novita	— / —	test_small_model_integration_test_batch_audio_matches_single · test_small_model_integration_test_w_audio
generation	[not_reproduced] GPU reproduce: the targeted tests did NOT fail at the base commit (not_reproduced) — the failure is stale, already fixed, flaky, or environment-specific. Skipping this group; no investigation, no PR.	Qwen/Qwen3.8-27B:novita	— / —	test_validate_assistant · test_matches_immediate_check_at_max_length · test_matches_immediate_check_on_a_batch · +3 more
blt	All 4 blt integration tests crash with AttributeError: 'BltConfig' object has no attribute 'num_hidden_layers' raised from src/transformers/integrations/heterogeneity/configuration_utils.py:312 (the HeterogeneousConfigMixin.__getattribute__ fallback return super().__getattribute__(key)).	Qwen/Qwen3.8-27B:novita	1,030,710 / 68,881	test_model_logits · test_model_logits_bf16
cwm	[not_reproduced] GPU reproduce: the targeted tests did NOT fail at the base commit (not_reproduced) — the failure is stale, already fixed, flaky, or environment-specific. Skipping this group; no investigation, no PR.	Qwen/Qwen3.8-27B:novita	— / —	test_cwm_generation_20_tokens · test_cwm_sliding_window_long_sequence
deepseek_vl_hybrid	review aborted (server restarted while running)	Qwen/Qwen3.8-27B:novita	— / —	test_model_text_generation_with_multi_image
exaone4	review aborted (server restarted while running)	Qwen/Qwen3.8-27B:novita	— / —	test_model_generation_beyond_sliding_window

Generated 2026-10-02T01:03:29+00:00.