clarifai 11.2.3rc3__tar.gz → 11.2.3rc5__tar.gz

{clarifai-11.2.3rc3/clarifai.egg-info → clarifai-11.2.3rc5}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.2
 Name: clarifai
-Version: 11.2.3rc3
+Version: 11.2.3rc5
 Summary: Clarifai Python SDK
 Home-page: https://github.com/Clarifai/clarifai-python
 Author: Clarifai

clarifai-11.2.3rc5/clarifai/__init__.py

@@ -0,0 +1 @@
+__version__ = "11.2.3rc5"

{clarifai-11.2.3rc3 → clarifai-11.2.3rc5}/clarifai/runners/models/model_class.py

@@ -71,6 +71,37 @@ class ModelClass(ABC):
     output.data.text.raw = signatures_to_json(signatures)
     return resp
 
+  def _is_data_set(self, data_msg):
+      # Singular message fields
+      singular_fields = [
+          "image", "video", "metadata", "geo", "text", "audio", "ndarray"
+      ]
+      for field in singular_fields:
+          if data_msg.HasField(field):
+              return True
+
+      # Repeated fields
+      repeated_fields = [
+          "concepts", "colors", "clusters", "embeddings", "regions", "frames",
+          "tracks", "time_segments", "hits", "heatmaps", "parts"
+      ]
+      for field in repeated_fields:
+          if getattr(data_msg, field):  # checks if the list is not empty
+              return True
+
+      # Scalar fields (proto3 default: 0 for numbers, empty for strings/bytes, False for bool)
+      if (
+          data_msg.int_value != 0 or
+          data_msg.float_value != 0.0 or
+          data_msg.bytes_value != b"" or
+          data_msg.bool_value is True or
+          data_msg.string_value != ""
+      ):
+          return True
+
+      return False
+
+
   def _convert_input_data_to_new_format(
       self, data: resources_pb2.Data,
       input_fields: List[resources_pb2.ModelTypeField]) -> resources_pb2.Data:
@@ -78,97 +109,88 @@ class ModelClass(ABC):
     new_data = resources_pb2.Data()
     for field in input_fields:
       part_data = self._convert_field(data, field)
-      part = new_data.parts.add()
-      part.id = field.name
-      part.data.CopyFrom(part_data)
+      if self._is_data_set(part_data):
+        # if the field is set, add it to the new data part
+        part = new_data.parts.add()
+        part.id = field.name
+        part.data.CopyFrom(part_data)
+      else:
+        if field.required:
+          raise ValueError(f"Field {field.name} is required but not set")
     return new_data
 
   def _convert_field(self, old_data: resources_pb2.Data,
                      field: resources_pb2.ModelTypeField) -> resources_pb2.Data:
     data_type = field.type
+    new_data = resources_pb2.Data()
     if data_type == resources_pb2.ModelTypeField.DataType.STR:
-      new_data = resources_pb2.Data()
-      new_data.string_value = old_data.text.raw
+      if old_data.HasField('text'):
+        new_data.string_value = old_data.text.raw
       return new_data
     elif data_type == resources_pb2.ModelTypeField.DataType.IMAGE:
-      new_data = resources_pb2.Data()
-      new_data.image.CopyFrom(old_data.image)
+      if old_data.HasField('image'):
+        new_data.image.CopyFrom(old_data.image)
       return new_data
     elif data_type == resources_pb2.ModelTypeField.DataType.VIDEO:
-      new_data = resources_pb2.Data()
-      new_data.video.CopyFrom(old_data.video)
+      if old_data.HasField('video'):
+        new_data.video.CopyFrom(old_data.video)
       return new_data
     elif data_type == resources_pb2.ModelTypeField.DataType.BOOL:
-      new_data = resources_pb2.Data()
-      new_data.bool_value = old_data.bool_value
+      if old_data.HasField('bool_value'):
+        new_data.bool_value = old_data.bool_value
       return new_data
     elif data_type == resources_pb2.ModelTypeField.DataType.INT:
-      new_data = resources_pb2.Data()
-      new_data.int_value = old_data.int_value
+      if old_data.HasField('int_value'):
+        new_data.int_value = old_data.int_value
       return new_data
     elif data_type == resources_pb2.ModelTypeField.DataType.FLOAT:
-      new_data = resources_pb2.Data()
-      new_data.float_value = old_data.float_value
+      if old_data.HasField('float_value'):
+        new_data.float_value = old_data.float_value
       return new_data
     elif data_type == resources_pb2.ModelTypeField.DataType.BYTES:
-      new_data = resources_pb2.Data()
-      new_data.bytes_value = old_data.bytes_value
+      if old_data.HasField('bytes_value'):
+        new_data.bytes_value = old_data.bytes_value
       return new_data
     elif data_type == resources_pb2.ModelTypeField.DataType.NDARRAY:
-      new_data = resources_pb2.Data()
-      new_data.ndarray.CopyFrom(old_data.ndarray)
-      return new_data
-    elif data_type == resources_pb2.ModelTypeField.DataType.JSON_DATA:
-      new_data = resources_pb2.Data()
-      struct_dict = old_data.text.raw
-      new_data.string_value = json.dumps(struct_dict)
+      if old_data.HasField('ndarray'):
+        new_data.ndarray.CopyFrom(old_data.ndarray)
       return new_data
     elif data_type == resources_pb2.ModelTypeField.DataType.TEXT:
-      new_data = resources_pb2.Data()
-      new_data.text.CopyFrom(old_data.text)
+      if old_data.HasField('text'):
+        new_data.text.CopyFrom(old_data.text)
+      return new_data
+    elif data_type == resources_pb2.ModelTypeField.DataType.AUDIO:
+      if old_data.HasField('audio'):
+        new_data.audio.CopyFrom(old_data.audio)
       return new_data
     elif data_type == resources_pb2.ModelTypeField.DataType.CONCEPT:
-      new_data = resources_pb2.Data()
-      new_data.concepts.extend(old_data.concepts)
+      if old_data.concepts:
+        new_data.concepts.extend(old_data.concepts)
       return new_data
     elif data_type == resources_pb2.ModelTypeField.DataType.REGION:
-      new_data = resources_pb2.Data()
-      new_data.regions.extend(old_data.regions)
+      if old_data.regions:
+        new_data.regions.extend(old_data.regions)
       return new_data
     elif data_type == resources_pb2.ModelTypeField.DataType.FRAME:
-      new_data = resources_pb2.Data()
-      new_data.frames.extend(old_data.frames)
-      return new_data
-    elif data_type == resources_pb2.ModelTypeField.DataType.AUDIO:
-      new_data = resources_pb2.Data()
-      new_data.audio.CopyFrom(old_data.audio)
+      if old_data.frames:
+        new_data.frames.extend(old_data.frames)
       return new_data
     elif data_type == resources_pb2.ModelTypeField.DataType.LIST:
       new_data = resources_pb2.Data()
       if not field.type_args:
         raise ValueError("LIST type requires type_args")
       element_field = field.type_args[0]
-      element_data = self._convert_field(old_data, element_field)
-      part = new_data.parts.add()
-      part.data.CopyFrom(element_data)
-      return new_data
-    elif data_type == resources_pb2.ModelTypeField.DataType.TUPLE:
-      new_data = resources_pb2.Data()
-      for element_field in field.type_args:
+      if element_field in (resources_pb2.ModelTypeField.DataType.CONCEPT, 
+                           resources_pb2.ModelTypeField.DataType.REGION,
+                           resources_pb2.ModelTypeField.DataType.FRAME):
+        # convert to new format
         element_data = self._convert_field(old_data, element_field)
-        part = new_data.parts.add()
-        part.data.CopyFrom(element_data)
-      return new_data
-    elif data_type == resources_pb2.ModelTypeField.DataType.NAMED_FIELDS:
-      new_data = resources_pb2.Data()
-      for named_field in field.type_args:
-        part_data = self._convert_field(old_data, named_field)
-        part = new_data.parts.add()
-        part.id = named_field.name
-        part.data.CopyFrom(part_data)
-      return new_data
+        # part = new_data.parts.add()
+        # part.data.CopyFrom(element_data)
+      return element_data
     else:
-      raise ValueError(f"Unsupported data type: {data_type}")
+      return new_data
+      # raise ValueError(f"Unsupported data type: {data_type}")
 
   def is_old_format(self, data: resources_pb2.Data) -> bool:
     """Check if the Data proto is in the old format (without parts)."""
@@ -195,6 +217,17 @@ class ModelClass(ABC):
 
     return False
 
+  def _convert_output_data_to_old_format(self, data: resources_pb2.Data) -> resources_pb2.Data:
+    """Convert output data to old format."""
+    old_data = resources_pb2.Data()
+    part_data = data.parts[0].data
+    # Handle text.raw specially (common case for text outputs)
+    old_data = part_data
+    if old_data.string_value:
+      old_data.text.raw = old_data.string_value
+
+    return old_data
+
   def _batch_predict(self, method, inputs: List[Dict[str, Any]]) -> List[Any]:
     """Batch predict method for multiple inputs."""
     outputs = []
@@ -228,7 +261,8 @@ class ModelClass(ABC):
       python_param_types = method_info.python_param_types
       for input in request.inputs:
         # check if input is in old format
-        if self.is_old_format(input.data):
+        is_convert = self.is_old_format(input.data)
+        if is_convert:
           # convert to new format
           new_data = self._convert_input_data_to_new_format(input.data, signature.input_fields)
           input.data.CopyFrom(new_data)
@@ -238,11 +272,15 @@ class ModelClass(ABC):
       if len(inputs) == 1:
         inputs = inputs[0]
         output = method(**inputs)
-        outputs.append(self._convert_output_to_proto(output, signature.output_fields))
+        outputs.append(
+            self._convert_output_to_proto(
+                output, signature.output_fields, convert_old_format=is_convert))
       else:
         outputs = self._batch_predict(method, inputs)
         outputs = [
-            self._convert_output_to_proto(output, signature.output_fields) for output in outputs
+            self._convert_output_to_proto(
+                output, signature.output_fields, convert_old_format=is_convert)
+            for output in outputs
         ]
 
       return service_pb2.MultiOutputResponse(
@@ -269,7 +307,8 @@ class ModelClass(ABC):
       python_param_types = method_info.python_param_types
       for input in request.inputs:
         # check if input is in old format
-        if self.is_old_format(input.data):
+        is_convert = self.is_old_format(input.data)
+        if is_convert:
           # convert to new format
           new_data = self._convert_input_data_to_new_format(input.data, signature.input_fields)
           input.data.CopyFrom(new_data)
@@ -279,7 +318,11 @@ class ModelClass(ABC):
         inputs = inputs[0]
         for output in method(**inputs):
           resp = service_pb2.MultiOutputResponse()
-          self._convert_output_to_proto(output, signature.output_fields, proto=resp.outputs.add())
+          self._convert_output_to_proto(
+              output,
+              signature.output_fields,
+              proto=resp.outputs.add(),
+              convert_old_format=is_convert)
           resp.status.code = status_code_pb2.SUCCESS
           yield resp
       else:
@@ -287,7 +330,10 @@ class ModelClass(ABC):
           resp = service_pb2.MultiOutputResponse()
           for output in outputs:
             self._convert_output_to_proto(
-                output, signature.output_fields, proto=resp.outputs.add())
+                output,
+                signature.output_fields,
+                proto=resp.outputs.add(),
+                convert_old_format=is_convert)
           resp.status.code = status_code_pb2.SUCCESS
           yield resp
     except Exception as e:
@@ -322,7 +368,8 @@ class ModelClass(ABC):
 
       for input in request.inputs:
         # check if input is in old format
-        if self.is_old_format(input.data):
+        is_convert = self.is_old_format(input.data)
+        if is_convert:
           # convert to new format
           new_data = self._convert_input_data_to_new_format(input.data, signature.input_fields)
           input.data.CopyFrom(new_data)
@@ -349,7 +396,11 @@ class ModelClass(ABC):
 
       for output in method(**kwargs):
         resp = service_pb2.MultiOutputResponse()
-        self._convert_output_to_proto(output, signature.output_fields, proto=resp.outputs.add())
+        self._convert_output_to_proto(
+            output,
+            signature.output_fields,
+            proto=resp.outputs.add(),
+            convert_old_format=is_convert)
         resp.status.code = status_code_pb2.SUCCESS
         yield resp
     except Exception as e:
@@ -387,10 +438,15 @@ class ModelClass(ABC):
   def _convert_output_to_proto(self,
                                output: Any,
                                variables_signature: List[resources_pb2.ModelTypeField],
-                               proto=None) -> resources_pb2.Output:
+                               proto=None,
+                               convert_old_format=False) -> resources_pb2.Output:
     if proto is None:
       proto = resources_pb2.Output()
     serialize({'return': output}, variables_signature, proto.data, is_output=True)
+    if convert_old_format:
+      # convert to old format
+      data = self._convert_output_data_to_old_format(proto.data)
+      proto.data.CopyFrom(data)
     proto.status.code = status_code_pb2.SUCCESS
     if hasattr(self, "_prompt_tokens") and self._prompt_tokens is not None:
       proto.prompt_tokens = self._prompt_tokens

{clarifai-11.2.3rc3 → clarifai-11.2.3rc5}/clarifai/runners/utils/method_signatures.py

@@ -279,11 +279,6 @@ def serialize(kwargs, signatures, proto=None, is_output=False):
       raise TypeError('Got a single return value, but expected multiple outputs {%s}' %
                       ', '.join(sig.name for sig in signatures))
     raise TypeError('Got unexpected key: %s' % ', '.join(unknown))
-  inline_first_value = False
-  if (is_output and len(signatures) == 1 and signatures[0].name == 'return' and
-      len(kwargs) == 1 and 'return' in kwargs):
-    # if there is only one output, flatten it and return directly
-    inline_first_value = True
   for sig_i, sig in enumerate(signatures):
     if sig.name not in kwargs:
       if sig.required:
@@ -293,16 +288,10 @@ def serialize(kwargs, signatures, proto=None, is_output=False):
     serializer = serializer_from_signature(sig)
     # TODO determine if any (esp the first) var can go in the proto without parts
     # and whether to put this in the signature or dynamically determine it
-    if inline_first_value and sig_i == 0 and id(data) not in _ZERO_VALUE_IDS:
-      # inlined first value; note data must not be empty or 0 to inline, since that
-      # will correspond to the missing value case (which uses function defaults).
-      # empty values are put explicitly in parts.
-      serializer.serialize(proto, data)
-    else:
-      # add the part to the proto
-      part = proto.parts.add()
-      part.id = sig.name
-      serializer.serialize(part.data, data)
+    # add the part to the proto
+    part = proto.parts.add()
+    part.id = sig.name
+    serializer.serialize(part.data, data)
   return proto
 
 

{clarifai-11.2.3rc3 → clarifai-11.2.3rc5/clarifai.egg-info}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.2
 Name: clarifai
-Version: 11.2.3rc3
+Version: 11.2.3rc5
 Summary: Clarifai Python SDK
 Home-page: https://github.com/Clarifai/clarifai-python
 Author: Clarifai

clarifai-11.2.3rc3/clarifai/__init__.py

@@ -1 +0,0 @@
-__version__ = "11.2.3rc3"