Shangdi Yu

[export] Log to torch_export_usage using Logger

3

Summary: We need to use logger so we can dual inject into both scuba table and hive table. For now we still leave the old scuba ingestion path, but it...

training ir torchao migration

5

Summary: Migrate capture_pre_autograd_graph to export_for_training. Differential Revision: D63859678

[AOTI Minifier] Exclude illegal graphs from minifier search

44

Summary: Some graphs produced by the minifier graph cutter cannot be used for AOTI/export (illegal graphs), these should be considered as graphs that don't fail in the minifier, so the...

AOTI minifier flatten inputs

2

Flatten the inputs to minifier so AOTI Minifier can handle unflattened inputs and kwargs. - changed the entry point of minifier to be `compile_fx_aot`, which takes in a graph module...

Fix autocast HOP pass for nested autocast

3

Test Plan: ``` buck2 run 'fbcode//mode/dev-nosan' fbcode//caffe2/test:test_export -- -r "test_predispatch_autocast" ``` Differential Revision: D65970066

[export] avoid name collision when inlining node

2

Summary: When we have both `set_grad` and `autocast` HOP, name collision might happen when we try to inline a node. For exmaple, for a GraphModule like this: ``` GraphModule( (submod_0):...

Remove capture_pre_autograd_graph references in quantization

7

Summary: As title We remove the deprecated API references in code, docs, and tests. We also removed two tests that specific to capture_pre_autograd_graph API. Test Plan: CI Differential Revision: D65351887

[AOTI Minifier] Save EP instead of graphs

2

Summary: `repro.py` can have nested graph modules, e.g. ``` class Repro(torch.nn.Module): def __init__(self) -> None: super().__init__() self.true_graph_0 = GraphModule() def forward(self): true_graph_0 = self.true_graph_0 return (true_graph_0,) ``` So dumping the...

Aoti minifier flatten 2

2

Fixes #ISSUE_NUMBER cc @voznesenskym @penguinwu @EikanWang @jgong5 @Guobing-Chen @XiaobingSuper @zhuhaozhe @blzheng @wenzhe-nrv @jiayisunx @ipiszy @yf225 @chenyang78 @kadeng @muchulee8 @ColinPeppler @amjames @desertfire @chauhang @aakhundov

Add model code stack trace to torch.profile

9

Stack from [ghstack](https://github.com/ezyang/ghstack) (oldest at bottom): * __->__ #166677 ```python python test/test_fx.py -k profiler ``` Insert `torch._C._profiler._RecordFunctionFast` to fx graph codegen. We post-process the profiler dump using `map_recorded_events_to_aten_ops_with_stack_trace` to add...