TorchDynamo Update 4: LazyTensor & nvFuser Experiments (original) (raw)

jansel February 25, 2022, 5:09am 3

@cicirori we have not done so yet, but are thinking about an XLA backend for TorchDynamo, so we could compare with LTC+XLA and see if these same results hold.

1 Like

TorchDynamo Update 5: Improved Capture & Bigger Graphs

TorchDynamo Update 6: Training support with AOTAutograd

TorchDynamo Update 10: Integrating with PyTorch/XLA for Inference and Training

TorchDynamo Update 9: Making DDP Work with TorchDynamo

TorchDynamo Update 7: Inference with FX2TRT

TorchDynamo Update 11: Making FSDP and Dynamo Work Together