Make dense feature optional for bytes model (facebookresearch#429)

Summary:
Pull Request resolved: facebookresearch#429

^title

Reviewed By: ppuliu

Differential Revision: D14678130

fbshipit-source-id: 297d61cff25175bd51bfdf118abe0d22e7b70e25

pytext/models/doc_model.py

@@ -64,26 +64,29 @@ class ModelInput(Model.Config.ModelInput):
         inputs: ModelInput = ModelInput()
         embedding: WordEmbedding.Config = WordEmbedding.Config()
 
-    input_names = ["tokens", "tokens_lens"]
-    output_names = ["scores"]
-
     def arrange_model_inputs(self, tensor_dict):
         tokens, seq_lens = tensor_dict["tokens"]
         return (tokens, seq_lens)
 
     def arrange_targets(self, tensor_dict):
         return tensor_dict["labels"]
 
+    def get_export_input_names(self, tensorizers):
+        return ["tokens", "tokens_lens"]
+
+    def get_export_output_names(self, tensorizers):
+        return ["scores"]
+
     def vocab_to_export(self, tensorizers):
         return {"tokens": list(tensorizers["tokens"].vocab)}
 
     def caffe2_export(self, tensorizers, tensor_dict, path, export_onnx_path=None):
         exporter = ModelExporter(
             ModelExporter.Config(),
-            self.input_names,
+            self.get_export_input_names(tensorizers),
             self.arrange_model_inputs(tensor_dict),
             self.vocab_to_export(tensorizers),
-            self.output_names,
+            self.get_export_output_names(tensorizers),
         )
         return exporter.export_to_caffe2(self, path, export_onnx_path=export_onnx_path)
 

pytext/models/model.py

@@ -11,6 +11,7 @@
 from pytext.config.doc_classification import ModelInput
 from pytext.config.field_config import FeatureConfig
 from pytext.config.pytext_config import ConfigBase, ConfigBaseMeta
+from pytext.config.serialize import _is_optional
 from pytext.data import CommonMetadata
 from pytext.data.tensorizers import Tensorizer
 from pytext.models.module import create_module
@@ -25,6 +26,8 @@ class ModelInputMeta(ConfigBaseMeta):
     def __new__(metacls, typename, bases, namespace):
         annotations = namespace.get("__annotations__", {})
         for type in annotations.values():
+            if _is_optional(type):
+                type = type.__args__[0]
             if not issubclass(type, Tensorizer.Config):
                 raise TypeError(
                     "ModelInput configuration should only include tensorizers"

pytext/task/new_task.py

@@ -106,6 +106,7 @@ def from_config(cls, config: Config, unused_metadata=None, model_state=None):
         tensorizers = {
             name: create_component(ComponentType.TENSORIZER, tensorizer)
             for name, tensorizer in config.model.inputs._asdict().items()
+            if tensorizer
         }
         schema: Dict[str, Type] = {}
         for tensorizer in tensorizers.values():