[ET Device Support] Schema changes: device info on Tensor

Pull Request resolved: #17533

This diff adds device placement information to the ExecuTorch schema to support representing tensor-level device type information, which will be the basic requirement for the following tensor_parser updates.

This is part of the Phase 1 implementation to make ET device type work E2E without user-specified device placement.

Design doc: https://docs.google.com/document/d/1lwd9BlohmwkN5EEvRulO_b-XnZBwv1nMb5l2K3jfuwA/edit?tab=t.0#heading=h.o6anuvkix4bu
ghstack-source-id: 354940210
@exported-using-ghexport

Differential Revision: [D93635657](https://our.internmc.facebook.com/intern/diff/D93635657/)

Author: Gasoonjia

exir/_serialize/generated/executorch_flatbuffer/DeviceType.py

@@ -0,0 +1,7 @@
+# automatically generated by the FlatBuffers compiler, do not modify
+
+# namespace: executorch_flatbuffer
+
+class DeviceType(object):
+    CPU = 0
+    CUDA = 1

exir/_serialize/generated/executorch_flatbuffer/ExtraTensorInfo.py

@@ -51,8 +51,22 @@ def Location(self):
             return self._tab.Get(flatbuffers.number_types.Int8Flags, o + self._tab.Pos)
         return 0
 
+    # ExtraTensorInfo
+    def DeviceType(self):
+        o = flatbuffers.number_types.UOffsetTFlags.py_type(self._tab.Offset(10))
+        if o != 0:
+            return self._tab.Get(flatbuffers.number_types.Int8Flags, o + self._tab.Pos)
+        return 0
+
+    # ExtraTensorInfo
+    def DeviceIndex(self):
+        o = flatbuffers.number_types.UOffsetTFlags.py_type(self._tab.Offset(12))
+        if o != 0:
+            return self._tab.Get(flatbuffers.number_types.Int8Flags, o + self._tab.Pos)
+        return 0
+
 def ExtraTensorInfoStart(builder: flatbuffers.Builder):
-    builder.StartObject(3)
+    builder.StartObject(5)
 
 def Start(builder: flatbuffers.Builder):
     ExtraTensorInfoStart(builder)
@@ -75,6 +89,18 @@ def ExtraTensorInfoAddLocation(builder: flatbuffers.Builder, location: int):
 def AddLocation(builder: flatbuffers.Builder, location: int):
     ExtraTensorInfoAddLocation(builder, location)
 
+def ExtraTensorInfoAddDeviceType(builder: flatbuffers.Builder, deviceType: int):
+    builder.PrependInt8Slot(3, deviceType, 0)
+
+def AddDeviceType(builder: flatbuffers.Builder, deviceType: int):
+    ExtraTensorInfoAddDeviceType(builder, deviceType)
+
+def ExtraTensorInfoAddDeviceIndex(builder: flatbuffers.Builder, deviceIndex: int):
+    builder.PrependInt8Slot(4, deviceIndex, 0)
+
+def AddDeviceIndex(builder: flatbuffers.Builder, deviceIndex: int):
+    ExtraTensorInfoAddDeviceIndex(builder, deviceIndex)
+
 def ExtraTensorInfoEnd(builder: flatbuffers.Builder) -> int:
     return builder.EndObject()
 
@@ -89,6 +115,8 @@ def __init__(self):
         self.mutableDataSegmentsIdx = 0  # type: int
         self.fullyQualifiedName = None  # type: str
         self.location = 0  # type: int
+        self.deviceType = 0  # type: int
+        self.deviceIndex = 0  # type: int
 
     @classmethod
     def InitFromBuf(cls, buf, pos):
@@ -111,7 +139,9 @@ def __eq__(self, other):
         return type(self) == type(other) and \
             self.mutableDataSegmentsIdx == other.mutableDataSegmentsIdx and \
             self.fullyQualifiedName == other.fullyQualifiedName and \
-            self.location == other.location
+            self.location == other.location and \
+            self.deviceType == other.deviceType and \
+            self.deviceIndex == other.deviceIndex
 
     # ExtraTensorInfoT
     def _UnPack(self, extraTensorInfo):
@@ -120,6 +150,8 @@ def _UnPack(self, extraTensorInfo):
         self.mutableDataSegmentsIdx = extraTensorInfo.MutableDataSegmentsIdx()
         self.fullyQualifiedName = extraTensorInfo.FullyQualifiedName()
         self.location = extraTensorInfo.Location()
+        self.deviceType = extraTensorInfo.DeviceType()
+        self.deviceIndex = extraTensorInfo.DeviceIndex()
 
     # ExtraTensorInfoT
     def Pack(self, builder):
@@ -130,5 +162,7 @@ def Pack(self, builder):
         if self.fullyQualifiedName is not None:
             ExtraTensorInfoAddFullyQualifiedName(builder, fullyQualifiedName)
         ExtraTensorInfoAddLocation(builder, self.location)
+        ExtraTensorInfoAddDeviceType(builder, self.deviceType)
+        ExtraTensorInfoAddDeviceIndex(builder, self.deviceIndex)
         extraTensorInfo = ExtraTensorInfoEnd(builder)
         return extraTensorInfo

exir/_serialize/generated/executorch_flatbuffer/__init__.py

@@ -13,6 +13,7 @@
 from . import DataLocation
 from . import DataSegment
 from . import DelegateCall
+from . import DeviceType
 from . import Double
 from . import DoubleList
 from . import EValue
@@ -56,6 +57,7 @@
     "DataLocation",
     "DataSegment",
     "DelegateCall",
+    "DeviceType",
     "Double",
     "DoubleList",
     "EValue",

exir/schema.py

@@ -48,6 +48,11 @@ class TensorDataLocation(IntEnum):
     EXTERNAL = 1
 
 
+class DeviceType(IntEnum):
+    CPU = 0
+    CUDA = 1
+
+
 @dataclass
 class ExtraTensorInfo:
     """
@@ -57,6 +62,8 @@ class ExtraTensorInfo:
     mutable_data_segments_idx: int = 0
     fully_qualified_name: Optional[str] = None
     location: TensorDataLocation = TensorDataLocation.SEGMENT
+    device_type: DeviceType = DeviceType.CPU
+    device_index: int = 0
 
 
 @dataclass

schema/program.fbs

@@ -62,6 +62,13 @@ enum TensorDataLocation : byte {
   EXTERNAL = 1,
 }
 
+// Device type enum indicating where a tensor resides or should be allocated.
+// Please keep this in sync with executorch/runtime/core/portable_type/device.h
+enum DeviceType : byte {
+  CPU = 0,
+  CUDA = 1,
+}
+
 // Table to put additional information about tensors in that is not applicable
 // to the vast majority of tensors in the vast majority of programs.
 table ExtraTensorInfo {
@@ -80,6 +87,14 @@ table ExtraTensorInfo {
   //   must be non-empty, and is used as a key to find the tensor's external
   //   data. Tensor.data_buffer_idx is ignored.
   location: TensorDataLocation;
+
+  // [Optional] The device type where this tensor resides or should be allocated.
+  // Defaults to CPU for backward compatibility with existing PTE files.
+  device_type: DeviceType = CPU;
+
+  // [Optional] The device index for multi-device scenarios (e.g., cuda:0, cuda:1).
+  // Defaults to 0 (the first device of the given type).
+  device_index: byte = 0;
 }
 
 table Tensor {