Moves WeightsDescr down so it won't need forward references

Forward references are not yet properly supported by the interactive
docs generation scripts. They would also probably cause trouble on
older versions of python.

Author: Tomaz-Vieira

bioimageio/spec/model/v0_4.py

@@ -238,89 +238,6 @@ def file(self):
 ]
 
 
-class WeightsDescr(Node):
-    keras_hdf5: Optional[KerasHdf5WeightsDescr] = None
-    onnx: Optional[OnnxWeightsDescr] = None
-    pytorch_state_dict: Optional[PytorchStateDictWeightsDescr] = None
-    tensorflow_js: Optional[TensorflowJsWeightsDescr] = None
-    tensorflow_saved_model_bundle: Optional[TensorflowSavedModelBundleWeightsDescr] = (
-        None
-    )
-    torchscript: Optional[TorchscriptWeightsDescr] = None
-
-    @model_validator(mode="after")
-    def check_one_entry(self) -> Self:
-        if all(
-            entry is None
-            for entry in [
-                self.keras_hdf5,
-                self.onnx,
-                self.pytorch_state_dict,
-                self.tensorflow_js,
-                self.tensorflow_saved_model_bundle,
-                self.torchscript,
-            ]
-        ):
-            raise ValueError("Missing weights entry")
-
-        return self
-
-    def __getitem__(
-        self,
-        key: WeightsFormat,
-    ):
-        if key == "keras_hdf5":
-            ret = self.keras_hdf5
-        elif key == "onnx":
-            ret = self.onnx
-        elif key == "pytorch_state_dict":
-            ret = self.pytorch_state_dict
-        elif key == "tensorflow_js":
-            ret = self.tensorflow_js
-        elif key == "tensorflow_saved_model_bundle":
-            ret = self.tensorflow_saved_model_bundle
-        elif key == "torchscript":
-            ret = self.torchscript
-        else:
-            raise KeyError(key)
-
-        if ret is None:
-            raise KeyError(key)
-
-        return ret
-
-    @property
-    def available_formats(self):
-        return {
-            **({} if self.keras_hdf5 is None else {"keras_hdf5": self.keras_hdf5}),
-            **({} if self.onnx is None else {"onnx": self.onnx}),
-            **(
-                {}
-                if self.pytorch_state_dict is None
-                else {"pytorch_state_dict": self.pytorch_state_dict}
-            ),
-            **(
-                {}
-                if self.tensorflow_js is None
-                else {"tensorflow_js": self.tensorflow_js}
-            ),
-            **(
-                {}
-                if self.tensorflow_saved_model_bundle is None
-                else {
-                    "tensorflow_saved_model_bundle": self.tensorflow_saved_model_bundle
-                }
-            ),
-            **({} if self.torchscript is None else {"torchscript": self.torchscript}),
-        }
-
-    @property
-    def missing_formats(self):
-        return {
-            wf for wf in get_args(WeightsFormat) if wf not in self.available_formats
-        }
-
-
 class WeightsEntryDescrBase(FileDescr):
     type: ClassVar[WeightsFormat]
     weights_format_name: ClassVar[str]  # human readable
@@ -544,6 +461,89 @@ def _tfv(cls, value: Any):
         return value
 
 
+class WeightsDescr(Node):
+    keras_hdf5: Optional[KerasHdf5WeightsDescr] = None
+    onnx: Optional[OnnxWeightsDescr] = None
+    pytorch_state_dict: Optional[PytorchStateDictWeightsDescr] = None
+    tensorflow_js: Optional[TensorflowJsWeightsDescr] = None
+    tensorflow_saved_model_bundle: Optional[TensorflowSavedModelBundleWeightsDescr] = (
+        None
+    )
+    torchscript: Optional[TorchscriptWeightsDescr] = None
+
+    @model_validator(mode="after")
+    def check_one_entry(self) -> Self:
+        if all(
+            entry is None
+            for entry in [
+                self.keras_hdf5,
+                self.onnx,
+                self.pytorch_state_dict,
+                self.tensorflow_js,
+                self.tensorflow_saved_model_bundle,
+                self.torchscript,
+            ]
+        ):
+            raise ValueError("Missing weights entry")
+
+        return self
+
+    def __getitem__(
+        self,
+        key: WeightsFormat,
+    ):
+        if key == "keras_hdf5":
+            ret = self.keras_hdf5
+        elif key == "onnx":
+            ret = self.onnx
+        elif key == "pytorch_state_dict":
+            ret = self.pytorch_state_dict
+        elif key == "tensorflow_js":
+            ret = self.tensorflow_js
+        elif key == "tensorflow_saved_model_bundle":
+            ret = self.tensorflow_saved_model_bundle
+        elif key == "torchscript":
+            ret = self.torchscript
+        else:
+            raise KeyError(key)
+
+        if ret is None:
+            raise KeyError(key)
+
+        return ret
+
+    @property
+    def available_formats(self):
+        return {
+            **({} if self.keras_hdf5 is None else {"keras_hdf5": self.keras_hdf5}),
+            **({} if self.onnx is None else {"onnx": self.onnx}),
+            **(
+                {}
+                if self.pytorch_state_dict is None
+                else {"pytorch_state_dict": self.pytorch_state_dict}
+            ),
+            **(
+                {}
+                if self.tensorflow_js is None
+                else {"tensorflow_js": self.tensorflow_js}
+            ),
+            **(
+                {}
+                if self.tensorflow_saved_model_bundle is None
+                else {
+                    "tensorflow_saved_model_bundle": self.tensorflow_saved_model_bundle
+                }
+            ),
+            **({} if self.torchscript is None else {"torchscript": self.torchscript}),
+        }
+
+    @property
+    def missing_formats(self):
+        return {
+            wf for wf in get_args(WeightsFormat) if wf not in self.available_formats
+        }
+
+
 class ParameterizedInputShape(Node):
     """A sequence of valid shapes given by `shape_k = min + k * step for k in {0, 1, ...}`."""