Vllm-Omni

The 30-Second Watchdog That Ate a Four-Hour Render: Long-Form MiniMax H3 on Two RTX PRO 6000

A 15s arcade-to-live-action video edit on H3 failed four different ways before it succeeded: cudnn graph crashes on long sequences, a 30s async-output watchdog that throws away finished renders, flashinfer kernels that require tcgen05 MMA my RTX PRO 6000s don’t have, and a finisher script that reported failure after succeeding. The winning recipe: patch the watchdog via bind-mount, CUDNN_ATTN, 540p reference, 1h45m end to end.

One Server, Every Mode: MiniMax H3 Goes From Static Frames to Full Video Editing on Two RTX PRO 6000

H3 ships as two checkpoints: FL2VA (text + first/last frame) and Ref2VA (references including video). Serving both DiTs from one process took three failed hypotheses, a basename-sensitive model resolver, a 66 GB minimal partition download, and one multipart field spelled plural. Now one systemd unit serves text-to-video, keyframe animation, and video editing at 56 GiB per GPU, validated end to end.