Issue evidence · Serve · vllm-project/vllm
GitHub issue: https://github.com/vllm-project/vllm/issues/61052
Not run
CPU KV offload faults with Xid 31 in cuMemcpyBatchAsync on an NVIDIA vGPU.
Nothing was run: the fault is specific to a vGPU-partitioned GPU, which Badgr's providers do not offer.
A physical-GPU run would only show the offload path working, which is not the reported failure.
Checked 2026-10-11 in Badgr’s local development environment.
← All issue evidence