Uses torch._inductor.compile instead of torch.compile in benchmark script to avoid segmentation by Dynamo #1540
Azure Pipelines / lightning-thunder (ipynb) (jupyter ubuntu22.04 | cuda 12.1 | torch 2.4.0)
succeeded
Dec 11, 2024 in 10m 12s
jupyter ubuntu22.04 | cuda 12.1 | torch 2.4.0 succeeded
Loading