[TRTLLM-13304][test] Add VisualGen multi-node E2E and external-launch unit coverage - #17271

Open

yingguo-trt wants to merge 22 commits into

NVIDIA:mainfrom

yingguo-trt:yiguo/TRTLLM-13304-visualgen-multinode-qa

Commits on Aug 6, 2026

[TRTLLM-13304][test] Add VisualGen multi-node QA coverage
yingguo-trt
committed
2a8c5d2
[TRTLLM-13304][test] Skip VisualGen multinode SLURM case under trtllm-llmapi-launch
yingguo-trt
committed
51c20ae
[TRTLLM-13304][ci] Add dedicated L0 stage for VisualGen multinode SLURM test
yingguo-trt
committed
a2b802a
[TRTLLM-13304][test] Cover VisualGen external-launch init and rank-0 wiring
yingguo-trt
committed
714f085
[TRTLLM-13304][fix] Fix NameError in VisualGen multinode SLURM rank path
yingguo-trt
committed
5fd4dce
[TRTLLM-13304][fix] Fix undefined MODULES_AVAILABLE and missing media deps in multinode case
yingguo-trt
committed
d611f87
[TRTLLM-13304][fix] Fix multinode case timeout layering
yingguo-trt
committed
a1e6374
[TRTLLM-13304][test] Simplify multinode case to single variant and tidy unit tests
yingguo-trt
committed
5e56754
[TRTLLM-13304][test] Re-register VisualGen multinode case in QA multinode list
yingguo-trt
committed
8c9c233
[TRTLLM-13304][fix] Clean up leaked threads in VisualGen multinode case
yingguo-trt
committed
ea4a80a
[TRTLLM-13304][fix] Align multinode LPIPS golden with the multi-GPU baseline
yingguo-trt
committed
8002016
[TRTLLM-13304][test] Register VisualGen multinode case in the QA multinode list
yingguo-trt
committed
af11a89
[TRTLLM-13304][fix] Capture rank-0 worker thread before shutdown
yingguo-trt
committed
87a5efe
[TRTLLM-13304][fix] Install media deps on rank 0 inside the srun step
yingguo-trt
committed
ea1ae89
[TRTLLM-13304][test] Drop VisualGen multinode case from the QA multinode list
yingguo-trt
committed
a5bd4ea
[TRTLLM-13304][chore] Trim test-db YAML header to the two-line SPDX form
yingguo-trt
committed
8cb0f94
[TRTLLM-13304][test] Rename test-db list to the multi_nodes convention
yingguo-trt
committed
0361758
[TRTLLM-13304][test] Drop redundant ring rank-linearization test
yingguo-trt
committed
4b42a01
[TRTLLM-13304][test] Address review feedback on the multinode case
yingguo-trt
committed
473bb40
[TRTLLM-13304][test] Narrow the PR to the multinode E2E and fix two guards
yingguo-trt
committed
18b6dff
[TRTLLM-13304][ci] Disable PMIx bootstrap for the VisualGen multinode step
yingguo-trt
committed
4cb0d10
[TRTLLM-13304][ci] Revert the SLURM_JOB_ID marker suffix
yingguo-trt
committed
53ce5b2

Navigation Menu

Uh oh!

[TRTLLM-13304][test] Add VisualGen multi-node E2E and external-launch unit coverage - #17271

[TRTLLM-13304][test] Add VisualGen multi-node E2E and external-launch unit coverage#17271
yingguo-trt wants to merge 22 commits into
NVIDIA:mainfrom
yingguo-trt:yiguo/TRTLLM-13304-visualgen-multinode-qa

Commits on Aug 6, 2026

[TRTLLM-13304][test] Add VisualGen multi-node QA coverage

[TRTLLM-13304][test] Skip VisualGen multinode SLURM case under trtllm-llmapi-launch

[TRTLLM-13304][ci] Add dedicated L0 stage for VisualGen multinode SLURM test

[TRTLLM-13304][test] Cover VisualGen external-launch init and rank-0 wiring

[TRTLLM-13304][fix] Fix NameError in VisualGen multinode SLURM rank path

[TRTLLM-13304][fix] Fix undefined MODULES_AVAILABLE and missing media deps in multinode case

[TRTLLM-13304][fix] Fix multinode case timeout layering

[TRTLLM-13304][test] Simplify multinode case to single variant and tidy unit tests

[TRTLLM-13304][test] Re-register VisualGen multinode case in QA multinode list

[TRTLLM-13304][fix] Clean up leaked threads in VisualGen multinode case

[TRTLLM-13304][fix] Align multinode LPIPS golden with the multi-GPU baseline

[TRTLLM-13304][test] Register VisualGen multinode case in the QA multinode list

[TRTLLM-13304][fix] Capture rank-0 worker thread before shutdown

[TRTLLM-13304][fix] Install media deps on rank 0 inside the srun step

[TRTLLM-13304][test] Drop VisualGen multinode case from the QA multinode list

[TRTLLM-13304][chore] Trim test-db YAML header to the two-line SPDX form

[TRTLLM-13304][test] Rename test-db list to the multi_nodes convention

[TRTLLM-13304][test] Drop redundant ring rank-linearization test

[TRTLLM-13304][test] Address review feedback on the multinode case

[TRTLLM-13304][test] Narrow the PR to the multinode E2E and fix two guards

[TRTLLM-13304][ci] Disable PMIx bootstrap for the VisualGen multinode step

[TRTLLM-13304][ci] Revert the SLURM_JOB_ID marker suffix