[Not to Land (Hopefully)]Revert PT Pin to 0912

.ci/docker/ci_commit_pins/pytorch.txt

@@ -1 +1 @@
-d1b87e26e5c4343f5b56bb1e6f89b479b389bfac
+aec9b2ab77389967ef39bb9c10662fd0fe3e185a

examples/apple/mps/scripts/mps_example.py

@@ -166,7 +166,7 @@ def get_model_config(args):
 
     # pre-autograd export. eventually this will become torch.export
     with torch.no_grad():
-        model = torch.export.export_for_training(model, example_inputs).module()
+        model = torch._export.capture_pre_autograd_graph(model, example_inputs)
         edge: EdgeProgramManager = export_to_edge(
             model,
             example_inputs,

examples/models/phi-3-mini/export_phi-3-mini.py

@@ -15,13 +15,13 @@
 from executorch.backends.xnnpack.partition.xnnpack_partitioner import XnnpackPartitioner
 from executorch.backends.xnnpack.utils.configs import get_xnnpack_edge_compile_config
 from executorch.exir import to_edge
+from torch._export import capture_pre_autograd_graph
 from torch.ao.quantization.quantize_pt2e import convert_pt2e, prepare_pt2e
 
 from torch.ao.quantization.quantizer.xnnpack_quantizer import (
     get_symmetric_quantization_config,
     XNNPACKQuantizer,
 )
-from torch.export import export_for_training
 
 from transformers import Phi3ForCausalLM
 
@@ -64,9 +64,9 @@ def export(args) -> None:
         xnnpack_quantizer = XNNPACKQuantizer()
         xnnpack_quantizer.set_global(xnnpack_quant_config)
 
-        model = export_for_training(
+        model = capture_pre_autograd_graph(
             model, example_inputs, dynamic_shapes=dynamic_shapes
-        ).module()
+        )
         model = prepare_pt2e(model, xnnpack_quantizer)  # pyre-fixme[6]
         model(*example_inputs)
         model = convert_pt2e(model)

exir/tests/test_passes.py

@@ -1431,10 +1431,10 @@ def quantize_model(
             m_eager: torch.nn.Module, example_inputs: Tuple[torch.Tensor]
         ) -> Tuple[EdgeProgramManager, int, int]:
             # program capture
-            m = torch.export.export_for_training(
+            m = torch._export.capture_pre_autograd_graph(
                 m_eager,
                 example_inputs,
-            ).module()
+            )
 
             quantizer = XNNPACKQuantizer()
             quantization_config = get_symmetric_quantization_config()

install_requirements.py

@@ -106,14 +106,14 @@ def python_is_compatible():
 # NOTE: If a newly-fetched version of the executorch repo changes the value of
 # NIGHTLY_VERSION, you should re-run this script to install the necessary
 # package versions.
-NIGHTLY_VERSION = "dev20241007"
+NIGHTLY_VERSION = "dev20240912"
 
 # The pip repository that hosts nightly torch packages.
 TORCH_NIGHTLY_URL = "https://download.pytorch.org/whl/nightly/cpu"
 
 # pip packages needed by exir.
 EXIR_REQUIREMENTS = [
-    f"torch==2.6.0.{NIGHTLY_VERSION}",
+    f"torch==2.5.0.{NIGHTLY_VERSION}",
     f"torchvision==0.20.0.{NIGHTLY_VERSION}",  # For testing.
     "typing-extensions",
 ]