Harry Mellor

We have a test suite where some tests are regular Python tests and others test a Python API with a locally hosted server. So we want everything parallelised but the...

Stale

`vllm-nccl-cu12` was a workaround to pin the NCCL version when we upgraded to PyTorch 2.2. NCCL 2.19 (which was the new default with PyTorch 2.2) was using much more memory...

@Tostino provided the correct answer.