Add eval mode to property encoder

Add eval mode to property encoder

Author: sfluegel05

chebai_graph/models/base.py

@@ -184,6 +184,7 @@ def forward(self, batch: dict) -> torch.Tensor:
             torch.Tensor: Predicted output.
         """
         graph_data = batch["features"][0]
+        graph_data.to(self.device)
         assert isinstance(graph_data, GraphData)
         a = self.gnn(batch)
         a = scatter_add(a, graph_data.batch, dim=0)

chebai_graph/preprocessing/datasets/__init__.py

@@ -12,12 +12,14 @@
     ChEBI50_WFGE_WGN_GraphProp,
     ChEBI50GraphData,
     ChEBI50GraphProperties,
+    ChEBI100GraphProperties,
 )
 from .pubchem import PubChemGraphProperties
 
 __all__ = [
     "ChEBI50GraphFGAugmentorReader",
     "ChEBI50GraphProperties",
+    "ChEBI100GraphProperties",
     "ChEBI50GraphData",
     "PubChemGraphProperties",
     "ChEBI50_Atom_WGNOnly_GraphProp",

chebai_graph/preprocessing/datasets/chebi.py

@@ -2,6 +2,7 @@
 from abc import ABC
 from collections.abc import Callable
 from pprint import pformat
+from typing import Optional
 
 import pandas as pd
 import torch
@@ -281,7 +282,9 @@ def _merge_props_into_base(self, row: pd.Series) -> GeomData:
             molecule_attr=molecule_attr,
         )
 
-    def load_processed_data_from_file(self, filename: str) -> list[dict]:
+    def load_processed_data(
+        self, kind: Optional[str] = None, filename: Optional[str] = None
+    ) -> list[dict]:
         """
         Load dataset and merge cached properties into base features.
 
@@ -291,7 +294,7 @@ def load_processed_data_from_file(self, filename: str) -> list[dict]:
         Returns:
             List of data entries, each a dictionary.
         """
-        base_data = super().load_processed_data_from_file(filename)
+        base_data = super().load_processed_data(kind, filename)
         base_df = pd.DataFrame(base_data)
 
         for property in self.properties:
@@ -379,7 +382,9 @@ def __init__(self, properties=None, transform=None, **kwargs):
             f"Data module uses these properties (ordered): {', '.join([str(p) for p in self.properties])}",
         )
 
-    def load_processed_data_from_file(self, filename: str) -> list[dict]:
+    def load_processed_data(
+        self, kind: Optional[str] = None, filename: Optional[str] = None
+    ) -> list[dict]:
         """
         Load dataset and merge cached properties into base features.
 
@@ -389,9 +394,8 @@ def load_processed_data_from_file(self, filename: str) -> list[dict]:
         Returns:
             List of data entries, each a dictionary.
         """
-        base_data = super().load_processed_data_from_file(filename)
+        base_data = super().load_processed_data(kind, filename)
         base_df = pd.DataFrame(base_data)
-
         props_categories = {
             "AllNodeTypeProperties": [],
             "FGNodeTypeProperties": [],
@@ -442,6 +446,7 @@ def load_processed_data_from_file(self, filename: str) -> list[dict]:
         )
 
         for property in self.properties:
+            rank_zero_info(f"Loading property {property.name}...")
             property_data = torch.load(
                 self.get_property_path(property), weights_only=False
             )

chebai_graph/preprocessing/property_encoder.py

@@ -16,9 +16,10 @@ class PropertyEncoder(abc.ABC):
         **kwargs: Additional keyword arguments.
     """
 
-    def __init__(self, property, **kwargs) -> None:
+    def __init__(self, property, eval=False, **kwargs) -> None:
         self.property = property
         self._encoding_length: int = 1
+        self.eval = eval  # if True, do not update cache (for index encoder)
 
     @property
     def name(self) -> str:
@@ -150,6 +151,10 @@ def encode(self, token: str | None) -> torch.Tensor:
             self._count_for_unk_token += 1
             return torch.tensor([self._unk_token_idx])
 
+        if self.eval and str(token) not in self.cache:
+            self._count_for_unk_token += 1
+            return torch.tensor([self._unk_token_idx])
+
         if str(token) not in self.cache:
             self.cache[str(token)] = len(self.cache)
         return torch.tensor([self.cache[str(token)] + self.offset])
@@ -213,6 +218,15 @@ def encode(self, token: str | None) -> torch.Tensor:
         Returns:
             One-hot encoded tensor of shape (1, encoding_length).
         """
+        if self.eval:
+            if token is None or str(token) not in self.cache:
+                self._count_for_unk_token += 1
+                return torch.zeros(self.get_encoding_length(), dtype=torch.int64)
+            index = self.cache[str(token)] + self.offset
+            return torch.nn.functional.one_hot(
+                torch.tensor(index), num_classes=self.get_encoding_length()
+            )
+
         if token not in self.tokens_dict:
             self._count_for_unk_token += 1
             return torch.zeros(1, self.get_encoding_length(), dtype=torch.int64)