[executorch][serialization] Data serialization interface

Pull Request resolved: #7194

Introduce data serialization interface.
ghstack-source-id: 257358159
@exported-using-ghexport

Differential Revision: [D65947145](https://our.internmc.facebook.com/intern/diff/D65947145/)

Author: lucylq

exir/_serialize/data_serializer.py

@@ -0,0 +1,77 @@
+from abc import ABC, abstractmethod
+from dataclasses import dataclass
+from typing import Dict, List, Sequence
+
+from executorch.exir._serialize._cord import Cord
+
+from executorch.exir.schema import ScalarType
+
+
+@dataclass
+class TensorLayout:
+    """
+    Tensor layout information for externally-serialized tensors.
+    """
+    scalar_type: ScalarType
+    dim_sizes: List[int]
+    dim_order: List[bytes]
+
+
+@dataclass
+class SerializationInfo:
+    # A sequence of tensor data buffers.
+    tensor_buffers: Sequence[bytes]
+
+    # A map from tensor name (fqn) to tensor index inside `tensor_buffers`.
+    fqn_to_buffer_index: Dict[str, int]
+
+    # A map from tensor name (fqn) to TensorLayout.
+    fqn_to_tensor_layout: Dict[str, TensorLayout]
+
+
+class DataSerializer(ABC):
+    """Serializes and deserializes FQN-tagged tensor data.
+
+    This base class enables serialization into different formats. See
+    executorch/extension/flat_tensor/ for an example.
+    """
+
+    @abstractmethod
+    def __init__(self) -> None:
+        """
+        This initializer may be overridden in derived classes to hold
+        the data required for serialization, eg. configurations.
+        """
+        pass
+
+    @abstractmethod
+    def serialize_tensors(
+        self,
+        serialization_info: SerializationInfo,
+    ) -> Cord:
+        """
+        Serializes a list of tensors emitted by ExecuTorch into a binary blob.
+
+        Args:
+            serialization_info: the tensor buffers and tensor layout
+            information required for serialization.
+
+        Returns:
+            A binary blob that contains the serialized data.
+        """
+        raise NotImplementedError("serialize_data")
+
+    @abstractmethod
+    def deserialize_tensors(self, blob: Cord) -> SerializationInfo:
+        """
+        Deserializes a blob into a list of tensors. Reverses the effect of
+        serialize_tensors.
+
+        Args:
+            blob: A binary blob that contains the serialized data.
+
+        Returns:
+            SerializationInfo: tensor buffers and tensor layout information
+            deserialized from `blob`.
+        """
+        raise NotImplementedError("deserialize_data")