3 Commits

Author	SHA1	Message	Date
Brian Hirsh	9834358e0f	Get SchemaCheckMode to error on ops that return inputs directly. Expose as a dynamo backend, eager_debug (#99744) ... Talked to @zou3519 and @ezyang on what the right UX is: tentatively, adding a new dynamo backend is cheap and simple, so it seems worth doing. And longer term, we agreed (?) that it's worth seeing if we can get custom ops sanity asserts to run more automatically, instead of needing a separate backend. Side comment: that actually seems tough: the mode detects secret mutations by cloning every input to every op, running the op, and checking that the data matches between the real input and the cloned input. So I doubt we'll be able to make that behavior always-on? It would need some config at least. Pull Request resolved: https://github.com/pytorch/pytorch/pull/99744 Approved by: https://github.com/albanD, https://github.com/ezyang, https://github.com/zou3519	2023-04-27 20:12:42 +00:00
Jason Ansel	e071d72f3c	Tag dynamo backends as debug/experimental (#93878) ... Hides debug/experimental backends by default. Before: ``` torch._dynamo.list_backends() ['aot_eager', 'aot_eager_decomp_partition', 'aot_torchxla_trace_once', 'aot_torchxla_trivial', 'aot_ts', 'aot_ts_nvfuser', 'cudagraphs', 'dynamo_accuracy_minifier_backend', 'dynamo_minifier_backend', 'eager', 'inductor', 'ipex', 'nvprims_aten', 'nvprims_nvfuser', 'onnxrt', 'tensorrt', 'torchxla_trace_once', 'torchxla_trivial', 'ts', 'tvm'] ``` After: ``` torch._dynamo.list_backends() ['aot_ts_nvfuser', 'cudagraphs', 'inductor', 'ipex', 'nvprims_nvfuser', 'onnxrt', 'tensorrt', 'tvm'] ``` Fixes https://github.com/pytorch/pytorch/issues/93733 Pull Request resolved: https://github.com/pytorch/pytorch/pull/93878 Approved by: https://github.com/voznesenskym	2023-02-04 00:50:51 +00:00
Jason Ansel	60e8c766b5	Refactor dynamo training backends (#93409) ... This splits training.py into many files and moves them from `dynamo.optimizations.training` to `dynamo.backends.*`. Pull Request resolved: https://github.com/pytorch/pytorch/pull/93409 Approved by: https://github.com/ezyang	2023-02-03 03:07:15 +00:00