clarifai 11.1.7__py3-none-any.whl → 11.1.7rc2__py3-none-any.whl

clarifai/runners/utils/data_handler.py

@@ -1,231 +1,292 @@
-from typing import Dict, List, Tuple, Union
+import io
+from typing import Any
 
 import numpy as np
 from clarifai_grpc.grpc.api import resources_pb2
+from clarifai_grpc.grpc.api.resources_pb2 import Audio as AudioProto
+from clarifai_grpc.grpc.api.resources_pb2 import Image as ImageProto
+from clarifai_grpc.grpc.api.resources_pb2 import Text as TextProto
+from clarifai_grpc.grpc.api.resources_pb2 import Video as VideoProto
 from clarifai_grpc.grpc.api.status import status_code_pb2, status_pb2
-from PIL import Image
+from google.protobuf.json_format import MessageToDict, ParseDict
+from google.protobuf.struct_pb2 import Struct
+from PIL import Image as PILImage
+
+
+def metadata_to_dict(data: resources_pb2.Data) -> dict:
+  return MessageToDict(data.metadata)
+
+
+def dict_to_metadata(data: resources_pb2.Data, metadata_dict: dict):
+  struct = Struct()
+  ParseDict(metadata_dict, struct)
+  data.metadata.CopyFrom(struct)
+
+
+def kwargs_to_proto(*args, **kwargs) -> resources_pb2.Data:
+  """Converts the kwargs to a Clarifai protobuf Data message."""
+
+  kwargs = dict(kwargs)
+  if any(k.startswith("_arg_") for k in kwargs.keys()):
+    raise ValueError("Keys starting with '_arg_' are reserved for positional arguments")
+  for arg_i, arg in enumerate(args):
+    kwargs[f"_arg_{arg_i}"] = arg
+
+  def _handle_list(target_data, value_list, part_name):
+    """Handles list values by processing each item into a new part."""
+    if isinstance(value_list[0], dict):
+      raise ValueError("List of dictionaries is not supported")
+
+    for item in value_list:
+      new_part = target_data.parts.add()
+      _process_value(new_part.data, item, part_name)
+
+  def _process_value(target_data, value, part_name):
+    """Processes individual values and sets the appropriate proto field."""
+    if isinstance(value, Text):
+      target_data.text.CopyFrom(value.to_proto())
+    elif isinstance(value, Image):
+      target_data.image.CopyFrom(value.to_proto())
+    elif isinstance(value, Audio):
+      target_data.audio.CopyFrom(value.to_proto())
+    elif isinstance(value, Video):
+      target_data.video.CopyFrom(value.to_proto())
+    elif isinstance(value, str):
+      target_data.text.raw = value
+    elif isinstance(value, bytes):
+      target_data.bytes_value = value
+    elif isinstance(value, int):
+      target_data.int_value = value
+    elif isinstance(value, float):
+      target_data.float_value = value
+    elif isinstance(value, bool):
+      target_data.bool_value = value
+    elif isinstance(value, np.ndarray):
+      ndarray_proto = resources_pb2.NDArray(
+          buffer=value.tobytes(), shape=value.shape, dtype=str(value.dtype))
+      target_data.ndarray.CopyFrom(ndarray_proto)
+    elif isinstance(value, PILImage.Image):
+      image = Image.from_pil(value)
+      target_data.image.CopyFrom(image.to_proto())
+    else:
+      raise TypeError(f"Unsupported type {type(value)} for part '{part_name}'")
+
+  data_proto = resources_pb2.Data()
+  for part_name, part_value in kwargs.items():
+    part = data_proto.parts.add()
+    part.id = part_name
+
+    if isinstance(part_value, list):
+      _handle_list(part.data, part_value, part_name)
+    elif isinstance(part_value, dict):
+      dict_to_metadata(part.data, part_value)
+    else:
+      _process_value(part.data, part_value, part_name)
+  return data_proto
+
+
+def proto_to_kwargs(data: resources_pb2.Data) -> dict:
+  """Converts the Clarifai protobuf Data message to a dictionary."""
+
+  def process_part(part, allow_metadata: bool = True) -> object:
+    if part.HasField("text"):
+      return Text.from_proto(part.text).text
+    elif part.HasField("image"):
+      return Image(part.image)
+    elif part.HasField("audio"):
+      return Audio(part.audio)
+    elif part.HasField("video"):
+      return Video(part.video)
+    elif part.bytes_value != b'':
+      return part.bytes_value
+    elif part.int_value != 0:
+      return part.int_value
+    elif part.float_value != 0.0:
+      return part.float_value
+    elif part.bool_value is not False:
+      return part.bool_value
+    elif part.HasField("ndarray"):
+      ndarray = part.ndarray
+      return np.frombuffer(ndarray.buffer, dtype=np.dtype(ndarray.dtype)).reshape(ndarray.shape)
+    elif part.HasField("metadata"):
+      if not allow_metadata:
+        raise ValueError("Metadata in list is not supported")
+      return metadata_to_dict(part)
+    elif part.parts:
+      return [process_part(p.data,) for p in part.parts]
+    else:
+      raise ValueError(f"Unknown part data: {part}")
+
+  kwargs = {}
+  part_names = [part.id for part in data.parts]
+  assert "return" not in part_names, "The key 'return' is reserved"
+  for part in data.parts:
+    part_name = part.id
+    part_data = part.data
+    kwargs[part_name] = process_part(part_data)
+  args = [kwargs.pop(f"_arg_{i}") for i in range(len(kwargs)) if f"_arg_{i}" in kwargs]
+  return args, kwargs
+
+
+class Output:
+
+  def __init__(self, *args, **kwargs: Any):
+
+    if not kwargs:
+      raise ValueError("Output must have at least one  key-value pair")
+    if isinstance(kwargs, dict):
+      kwargs = kwargs
+    else:
+      raise ValueError("Output must be a dictionary")
+    self.parts = kwargs
+    self.args = args
+
+  def __repr__(self):
+    args_str = ', '.join(repr(arg) for arg in self.args)
+    kwargs_str = ', '.join(f"{k}={v!r}" for k, v in self.parts.items())
+    parts = []
+    if args_str:
+      parts.append(args_str)
+    if kwargs_str:
+      parts.append(kwargs_str)
+    return f"Output({', '.join(parts)})"
+
+  def to_proto(self) -> resources_pb2.Output:
+    """Converts the Output instance to a Clarifai protobuf Output message."""
+    data_proto = kwargs_to_proto(*self.args, **self.parts)
+
+    return resources_pb2.Output(
+        data=data_proto, status=status_pb2.Status(code=status_code_pb2.SUCCESS))
+
+
+class Text:
+
+  def __init__(self, text: str):
+    self.text = text
+
+  def to_proto(self) -> TextProto:
+    return TextProto(raw=self.text)
 
-from clarifai.client.auth.helper import ClarifaiAuthHelper
+  @classmethod
+  def from_proto(cls, proto: TextProto) -> "Text":
+    return cls(proto.raw)
 
-from .data_utils import bytes_to_image, image_to_bytes
 
+class Image:
 
-class BaseDataHandler:
+  def __init__(self, proto_image: ImageProto):
+    self.proto = proto_image
 
-  def __init__(self,
-               proto: Union[resources_pb2.Input, resources_pb2.Output],
-               auth: ClarifaiAuthHelper = None):
-    self._proto = proto
-    self._auth = auth
+  @property
+  def url(self) -> str:
+    return self.proto.url
 
-  #
-  def to_python(self):
-    return dict(text=self.text, image=self.image, audio=self.audio)
+  @url.setter
+  def url(self, value: str):
+    self.proto.url = value
 
-  # ---------------- Start get/setters ---------------- #
-  # Proto
   @property
-  def proto(self):
-    return self._proto
+  def bytes(self) -> bytes:
+    return self.proto.base64
 
-  # Status
-  @property
-  def status(self) -> status_pb2.Status:
-    return self._proto.status
+  @bytes.setter
+  def bytes(self, value: bytes):
+    self.proto.base64 = value
 
-  def set_status(self, code: str, description: str = ""):
-    self._proto.status.code = code
-    self._proto.status.description = description
+  def __repr__(self) -> str:
+    attrs = []
+    if self.url:
+      attrs.append(f"url={self.url!r}")
+    if self.bytes:
+      attrs.append(f"bytes=<{len(self.bytes)} bytes>")
+    return f"Image({', '.join(attrs)})"
+
+  @classmethod
+  def from_url(cls, url: str) -> "Image":
+    proto_image = ImageProto(url=url)
+    return cls(proto_image)
+
+  @classmethod
+  def from_pil(cls, pil_image: PILImage.Image) -> "Image":
+    with io.BytesIO() as output:
+      pil_image.save(output, format="PNG")
+      image_bytes = output.getvalue()
+    proto_image = ImageProto(base64=image_bytes)
+    return cls(proto_image)
+
+  def to_pil(self) -> PILImage.Image:
+    return PILImage.open(io.BytesIO(self.proto.base64))
+
+  def to_numpy(self) -> np.ndarray:
+    # below is very slow, need to find a better way
+    # return np.array(self.to_pil())
+    pass
+
+  def to_proto(self) -> ImageProto:
+    return self.proto
+
+
+class Audio:
+
+  def __init__(self, proto_audio: AudioProto):
+    self.proto = proto_audio
 
-  # Text
-  @property
-  def text(self) -> Union[None, str]:
-    data = self._proto.data.text
-    text = None
-    if data.ByteSize():
-      if data.raw:
-        text = data.raw
-      else:
-        raise NotImplementedError
-    return text
-
-  def set_text(self, text: str):
-    self._proto.data.text.raw = text
-
-  # Image
-  @property
-  def image(self, format: str = "np") -> Union[None, Image.Image, np.ndarray]:
-    data = self._proto.data.image
-    image = None
-    if data.ByteSize():
-      data: resources_pb2.Image = data
-      if data.base64:
-        image = data.base64
-      elif data.url:
-        raise NotImplementedError
-      image = bytes_to_image(image)
-      image = image if not format == "np" else np.asarray(image).astype("uint8")
-    return image
-
-  def set_image(self, image: Union[Image.Image, np.ndarray]):
-    if isinstance(image, np.ndarray):
-      image = Image.fromarray(image)
-    self._proto.data.image.base64 = image_to_bytes(image)
-
-  # Audio
-  @property
-  def audio(self) -> bytes:
-    data = self._proto.data.audio
-    audio = None
-    if data.ByteSize():
-      if data.base64:
-        audio = data.base64
-    return audio
-
-  def set_audio(self, audio: bytes):
-    self._proto.data.audio.base64 = audio
-
-  # Bboxes
-  @property
-  def bboxes(self, real_coord: bool = False, image_width: int = None,
-             image_height: int = None) -> Tuple[List, List, List]:
-    if real_coord:
-      assert (image_height or image_width
-             ), "image_height and image_width are required when when return real coordinates"
-    xyxy = []
-    scores = []
-    concepts = []
-    for _, each in enumerate(self._proto.data.regions):
-      box = each.region_info
-      score = each.value
-      concept = each.data.concepts[0].id
-      x1 = box.left_col
-      y1 = box.top_row
-      x2 = box.right_col
-      y2 = box.bottom_row
-      if real_coord:
-        x1 = x1 * image_width
-        y1 = y1 * image_height
-        x2 = x2 * image_width
-        y2 = y2 * image_height
-      xyxy.append([x1, y1, x2, y2])
-      scores.append(score)
-      concepts.append(concept)
-
-    return xyxy, scores, concepts
-
-  def set_bboxes(self,
-                 boxes: list,
-                 scores: list,
-                 concepts: list,
-                 real_coord: bool = False,
-                 image_width: int = None,
-                 image_height: int = None):
-    if real_coord:
-      assert (image_height and
-              image_width), "image_height and image_width are required when `real_coord` is set"
-      bboxes = [[x[1] / image_height, x[0] / image_width, x[3] / image_height, x[2] / image_width]
-                for x in boxes]  # normalize the bboxes to [0,1] and [y1 x1 y2 x2]
-      bboxes = np.clip(bboxes, 0, 1.0)
-
-    regions = []
-    for ith, bbox in enumerate(bboxes):
-      score = scores[ith]
-      concept = concepts[ith]
-      if any([each > 1.0 for each in bbox]):
-        assert ValueError(
-            "Box coordinates is not normalized between [0, 1]. Please set format_box to True and provide image_height and image_width to normalize"
-        )
-      region = resources_pb2.RegionInfo(bounding_box=resources_pb2.BoundingBox(
-          top_row=bbox[0],  # y_min
-          left_col=bbox[1],  # x_min
-          bottom_row=bbox[2],  # y_max
-          right_col=bbox[3],  # x_max
-      ))
-      data = resources_pb2.Data(concepts=resources_pb2.Concept(id=concept, value=score))
-      regions.append(resources_pb2.Region(region_info=region, data=data))
-
-    self._proto.data.regions = regions
-
-  # Concepts
   @property
-  def concepts(self) -> Dict[str, float]:
-    con_scores = {}
-    for each in self.proto.data.concepts:
-      con_scores.update({each.id: each.value})
-    return con_scores
-
-  def set_concepts(self, concept_score_pairs: Dict[str, float]):
-    concepts = []
-    for concept, score in concept_score_pairs.items():
-      con_score = resources_pb2.Concept(id=concept, name=concept, value=score)
-      concepts.append(con_score)
-    if concepts:
-      self._proto.data.ClearField("concepts")
-      for each in concepts:
-        self._proto.data.concepts.append(each)
-
-  # Embeddings
+  def url(self) -> str:
+    return self.proto.url
+
+  @url.setter
+  def url(self, value: str):
+    self.proto.url = value
+
   @property
-  def embeddings(self) -> List[List[float]]:
-    return [each.vector for each in self.proto.data.embeddings]
+  def bytes(self) -> bytes:
+    return self.proto.base64
 
-  def set_embeddings(self, list_vectors: List[List[float]]):
-    if list_vectors[0]:
-      self._proto.data.ClearField("embeddings")
-    for vec in list_vectors:
-      self._proto.data.embeddings.append(
-          resources_pb2.Embedding(vector=vec, num_dimensions=len(vec)))
+  @bytes.setter
+  def bytes(self, value: bytes):
+    self.proto.base64 = value
 
-  # ---------------- End get/setters ---------------- #
+  def __repr__(self) -> str:
+    attrs = []
+    if self.url:
+      attrs.append(f"url={self.url!r}")
+    if self.bytes:
+      attrs.append(f"bytes=<{len(self.bytes)} bytes>")
+    return f"Audio({', '.join(attrs)})"
 
-  # Constructors
-  @classmethod
-  def from_proto(cls, proto):
-    clss = cls(proto=proto)
-    return clss
+  def to_proto(self) -> AudioProto:
+    return self.proto
 
-  @classmethod
-  def from_data(
-      cls,
-      status_code: int = status_code_pb2.SUCCESS,
-      status_description: str = "",
-      text: str = None,
-      image: Union[Image.Image, np.ndarray] = None,
-      audio: bytes = None,
-      boxes: dict = None,
-      concepts: Dict[str, float] = {},
-      embeddings: List[List[float]] = [],
-  ) -> 'OutputDataHandler':
-    clss = cls(proto=resources_pb2.Output())
-    if isinstance(image, Image.Image) or isinstance(image, np.ndarray):
-      clss.set_image(image)
-    if text:
-      clss.set_text(text)
-    if audio:
-      clss.set_audio(audio)
-    if boxes:
-      clss.set_bboxes(**boxes)
-    if concepts:
-      clss.set_concepts(concepts)
-    if embeddings:
-      clss.set_embeddings(embeddings)
-
-    clss.set_status(code=status_code, description=status_description)
-    return clss
-
-
-class InputDataHandler(BaseDataHandler):
-
-  def __init__(self,
-               proto: resources_pb2.Input = resources_pb2.Input(),
-               auth: ClarifaiAuthHelper = None):
-    super().__init__(proto=proto, auth=auth)
-
-
-class OutputDataHandler(BaseDataHandler):
-
-  def __init__(self,
-               proto: resources_pb2.Output = resources_pb2.Output(),
-               auth: ClarifaiAuthHelper = None):
-    super().__init__(proto=proto, auth=auth)
+
+class Video:
+
+  def __init__(self, proto_video: VideoProto):
+    self.proto = proto_video
+
+  @property
+  def url(self) -> str:
+    return self.proto.url
+
+  @url.setter
+  def url(self, value: str):
+    self.proto.url = value
+
+  @property
+  def bytes(self) -> bytes:
+    return self.proto.base64
+
+  @bytes.setter
+  def bytes(self, value: bytes):
+    self.proto.base64 = value
+
+  def __repr__(self) -> str:
+    attrs = []
+    if self.url:
+      attrs.append(f"url={self.url!r}")
+    if self.bytes:
+      attrs.append(f"bytes=<{len(self.bytes)} bytes>")
+    return f"Video({', '.join(attrs)})"
+
+  def to_proto(self) -> VideoProto:
+    return self.proto