FastAPI model changes

Signed-off-by: shubh chaurasia <[email protected]>

Author: shubh0508

mlflow/pyfunc/scoring_server/__init__.py

@@ -20,6 +20,12 @@
 import pandas as pd
 import sys
 import traceback
+from pydantic import BaseModel
+from fastapi import FastAPI, APIRouter, Request, HTTPException, Response, Header, status
+from typing import List, Optional, Dict
+import uvicorn
+import asyncio
+import json
 
 # NB: We need to be careful what we import form mlflow here. Scoring server is used from within
 # model's conda environment. The version of mlflow doing the serving (outside) and the version of
@@ -65,13 +71,24 @@
 
 CONTENT_TYPE_FORMAT_RECORDS_ORIENTED = "pandas-records"
 CONTENT_TYPE_FORMAT_SPLIT_ORIENTED = "pandas-split"
+CONTENT_TYPE_RAW_JSON = "raw-json"
 
-FORMATS = [CONTENT_TYPE_FORMAT_RECORDS_ORIENTED, CONTENT_TYPE_FORMAT_SPLIT_ORIENTED]
+FORMATS = [CONTENT_TYPE_FORMAT_RECORDS_ORIENTED, CONTENT_TYPE_FORMAT_SPLIT_ORIENTED, CONTENT_TYPE_RAW_JSON]
 
 PREDICTIONS_WRAPPER_ATTR_NAME_ENV_KEY = "PREDICTIONS_WRAPPER_ATTR_NAME"
 
 _logger = logging.getLogger(__name__)
 
+class RequestData(BaseModel):
+    columns: List[str] = []
+    data: list = []
+
+    def is_valid(self):
+        return True
+
+    def get_dataframe(self):
+        df = pd.DataFrame(data = self.data, columns = self.columns)
+        return df
 
 def infer_and_parse_json_input(json_input, schema: Schema = None):
     """
@@ -205,38 +222,38 @@ def _handle_serving_error(error_message, error_code, include_traceback=True):
         e = MlflowException(message=error_message, error_code=error_code)
     reraise(MlflowException, e)
 
-
 def init(model: PyFuncModel):
 
     """
     Initialize the server. Loads pyfunc model from the path.
     """
-    app = flask.Flask(__name__)
+    fast_app = FastAPI(title= __name__, version= "v1")
+    fast_app.include_router(APIRouter())
     input_schema = model.metadata.get_input_schema()
 
-    @app.route("/ping", methods=["GET"])
+    @fast_app.get("/ping")
     def ping():  # pylint: disable=unused-variable
         """
         Determine if the container is working and healthy.
         We declare it healthy if we can load the model successfully.
         """
-        health = model is not None
-        status = 200 if health else 404
-        return flask.Response(response="\n", status=status, mimetype="application/json")
+        if model is None:
+            raise HTTPException(status_code=404, detail="Model not loaded properly")
+        return {"message": "OK"}
 
-    @app.route("/invocations", methods=["POST"])
-    @catch_mlflow_exception
-    def transformation():  # pylint: disable=unused-variable
+    @fast_app.post("/invocations")
+    def transformation(request_data: RequestData, content_type: Optional[str] = Header(None)):  # pylint: disable=unused-variable
         """
         Do an inference on a single batch of data. In this sample server,
         we take data as CSV or json, convert it to a Pandas DataFrame or Numpy,
         generate predictions and convert them back to json.
         """
+        # data = _dataframe_from_json(request_data.json())
 
         # Content-Type can include other attributes like CHARSET
         # Content-type RFC: https://datatracker.ietf.org/doc/html/rfc2045#section-5.1
         # TODO: Suport ";" in quoted parameter values
-        type_parts = flask.request.content_type.split(";")
+        type_parts = content_type.split(";")
         type_parts = list(map(str.strip, type_parts))
         mime_type = type_parts[0]
         parameter_value_pairs = type_parts[1:]
@@ -247,27 +264,31 @@ def transformation():  # pylint: disable=unused-variable
 
         charset = parameter_values.get("charset", "utf-8").lower()
         if charset != "utf-8":
-            return flask.Response(
-                response="The scoring server only supports UTF-8",
-                status=415,
-                mimetype="text/plain",
+            return Response(
+                content="The scoring server only supports UTF-8",
+                status_code=415,
+                media_type="text/plain"
             )
 
         content_format = parameter_values.get("format")
 
         # Convert from CSV to pandas
         if mime_type == CONTENT_TYPE_CSV and not content_format:
-            data = flask.request.data.decode("utf-8")
+            data = request_data.json()
             csv_input = StringIO(data)
             data = parse_csv_input(csv_input=csv_input)
+        elif mime_type == CONTENT_TYPE_JSON and content_format == CONTENT_TYPE_RAW_JSON:
+            if len(request_data.data) != 0:
+                data = dict(zip(request_data.columns, request_data.data[0]))
+            else:
+                data = {}
         elif mime_type == CONTENT_TYPE_JSON and not content_format:
-            json_str = flask.request.data.decode("utf-8")
-            data = infer_and_parse_json_input(json_str, input_schema)
+            data = infer_and_parse_json_input(request_data.json(), input_schema)
         elif (
             mime_type == CONTENT_TYPE_JSON and content_format == CONTENT_TYPE_FORMAT_SPLIT_ORIENTED
         ):
             data = parse_json_input(
-                json_input=StringIO(flask.request.data.decode("utf-8")),
+                json_input=StringIO(request_data.json()),
                 orient="split",
                 schema=input_schema,
             )
@@ -276,29 +297,25 @@ def transformation():  # pylint: disable=unused-variable
             and content_format == CONTENT_TYPE_FORMAT_RECORDS_ORIENTED
         ):
             data = parse_json_input(
-                json_input=StringIO(flask.request.data.decode("utf-8")),
+                json_input=StringIO(request_data.json()),
                 orient="records",
                 schema=input_schema,
             )
         elif mime_type == CONTENT_TYPE_JSON_SPLIT_NUMPY and not content_format:
-            data = parse_split_oriented_json_input_to_numpy(flask.request.data.decode("utf-8"))
+            data = parse_split_oriented_json_input_to_numpy(request_data.json())
         else:
-            return flask.Response(
-                response=(
-                    "This predictor only supports the following content types and formats:"
+            return Response(
+                content="This predictor only supports the following content types and formats:"
                     " Types: {supported_content_types}; Formats: {formats}."
                     " Got '{received_content_type}'.".format(
                         supported_content_types=CONTENT_TYPES,
                         formats=FORMATS,
-                        received_content_type=flask.request.content_type,
-                    )
-                ),
-                status=415,
-                mimetype="text/plain",
+                        received_content_type=content_type,
+                    ),
+                status_code=415,
+                media_type="text/plain"
             )
 
-        # Do the prediction
-
         try:
             raw_predictions = model.predict(data)
         except MlflowException as e:
@@ -314,11 +331,10 @@ def transformation():  # pylint: disable=unused-variable
                 ),
                 error_code=BAD_REQUEST,
             )
-        result = StringIO()
-        predictions_to_json(raw_predictions, result)
-        return flask.Response(response=result.getvalue(), status=200, mimetype="application/json")
+        predictions = _get_jsonable_obj(raw_predictions, pandas_orient="records")
+        return predictions
 
-    return app
+    return fast_app
 
 
 def _predict(model_uri, input_path, output_path, content_type, json_format):
@@ -342,8 +358,8 @@ def _predict(model_uri, input_path, output_path, content_type, json_format):
 
 def _serve(model_uri, port, host):
     pyfunc_model = load_model(model_uri)
-    init(pyfunc_model).run(port=port, host=host)
-
+    fast_app = init(pyfunc_model)
+    uvicorn.run(fast_app, host=host, port=port, log_level="info")
 
 def get_cmd(
     model_uri: str, port: int = None, host: int = None, nworkers: int = None
@@ -362,8 +378,7 @@ def get_cmd(
             args.append(f"-w {nworkers}")
 
         command = (
-            f"gunicorn {' '.join(args)} ${{GUNICORN_CMD_ARGS}}"
-            " -- mlflow.pyfunc.scoring_server.wsgi:app"
+            "gunicorn mlflow.pyfunc.scoring_server.wsgi:app --worker-class uvicorn.workers.UvicornWorker"
         )
     else:
         args = []

setup.py

@@ -68,6 +68,8 @@ def package_files(directory):
     "alembic<=1.4.1",
     # Required
     "docker>=4.0.0",
+    "fastapi",
+    "uvicorn",
     "Flask",
     "gunicorn; platform_system != 'Windows'",
     "numpy",