You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
0.69.4: run the #4250 serve batched-prefill test where the Qwen3.5 model exists (it is a model-absent SKIP in ci.yml cuda-unit) #4251
#4250's gpu_every_serve_route_prefills_through_the_batched_prefill (aprender-serve lib, cuda-gated) is selected by ci.yml's derived cuda-only step. That step counts a missing model file as a legitimate skip, and the runners do not carry /home/noah/models/Qwen3.5-0.8B-Q4_K_M.gguf. So in CI the test is dark: it passes without running.
done_when: a cuda-nightly step, on a runner with the 0.8B GGUF (APR_*_MODEL override pattern as PERF-053), runs cargo test -p aprender-serve --features cuda --lib --release gpu_every_serve_route_prefills_through_the_batched_prefill -- --exact --nocapture and FAILS on the model-absent SKIP line. Show the RED once with APR_QWEN35_SESSION_PREFILL=per-token.
This is a workflow edit, so it needs the workflow quorum. Epic #4249. Refs #4250, #3596.
Branch PMAT-4251-serve-prefill-nightly @bfcf8dc82 (off rc/0.69.3-rc.2). Heads-up: the done_when command's --exact with the bare fn name matches 0 tests; the path is api::openai_handlers::qwen35_serve_tests::gpu_every_serve_route_prefills_through_the_batched_prefill. On lambda RTX 4090: default GREEN (batched prefill logged on every route), APR_QWEN35_SESSION_PREFILL=per-token RED (exit 101, panic at qwen35_chat_backend_tests.rs:448). The step fails on model-absent, nonzero exit, any SKIP: line, or anything but 1 passed. Workflow edit, so it needs the quorum before fold.
P0 to P1 (2026-10-14 P0 triage): P0 now means it blocks the next release or carries the operator's own words on it; this issue does neither, so it stays on the 0.71.0 milestone at P1.
Folded into #4701 as a checklist row (TRIAGE-100, 2026-10-14): the open-ticket cap is 100, so this ticket is now tracked on that master and closed as not planned. Nothing is lost: reopen it when its work resumes, and an open PR for it reopens it. Pushed branch: PMAT-4251-serve-prefill-nightly.
Activity
fix(cuda-nightly): #4251's --exact filter needs the full test path, o…
noahgift commentedon Sep 24, 2026
Branch
PMAT-4251-serve-prefill-nightly@bfcf8dc82 (off rc/0.69.3-rc.2). Heads-up: the done_when command's--exactwith the bare fn name matches 0 tests; the path isapi::openai_handlers::qwen35_serve_tests::gpu_every_serve_route_prefills_through_the_batched_prefill. On lambda RTX 4090: default GREEN (batched prefill logged on every route),APR_QWEN35_SESSION_PREFILL=per-tokenRED (exit 101, panic at qwen35_chat_backend_tests.rs:448). The step fails on model-absent, nonzero exit, anySKIP:line, or anything but1 passed. Workflow edit, so it needs the quorum before fold.ci(cuda-nightly): run #4250's serve batched-prefill test on gx10; a S…
fix(cuda-nightly): #4251's --exact filter needs the full test path, o…
ci(cuda-nightly): run #4250's serve batched-prefill test on gx10; a S…
fix(cuda-nightly): #4251's --exact filter needs the full test path, o…
ci: carry the 0.69.x workflow-only backports onto car (86d3775, f6da455…
noahgift commentedon Oct 14, 2026
P0 to P1 (2026-10-14 P0 triage): P0 now means it blocks the next release or carries the operator's own words on it; this issue does neither, so it stays on the 0.71.0 milestone at P1.
noahgift commentedon Oct 14, 2026
Folded into #4701 as a checklist row (TRIAGE-100, 2026-10-14): the open-ticket cap is 100, so this ticket is now tracked on that master and closed as not planned. Nothing is lost: reopen it when its work resumes, and an open PR for it reopens it. Pushed branch:
PMAT-4251-serve-prefill-nightly.Agent: aprender-clerk