feat(model): use latest whisper with turbo

whisper-image/app/predictor.py

@@ -17,7 +17,8 @@
 app = flask.Flask(__name__)
 
 s3_client = boto3.client("s3")
-model_name = "medium.en"
+default_model_name = "turbo"
+
 
 @app.route("/ping", methods=["GET"])
 def ping():
@@ -32,10 +33,10 @@ def execution_parameters():
     status = 200
     return flask.Response(response="{}", status=status, mimetype="application/json")
 
+
 @app.route("/invocations", methods=["POST"])
 def transformation():
-    """Do an inference on a single batch of data. 
-    """
+    """Do an inference on a single batch of data."""
     content_type = flask.request.content_type
     request_data = flask.request.data
     logger.info(f"transformation: {content_type} {request_data}")
@@ -47,11 +48,14 @@ def transformation():
         input_dict = json.loads(data)
     else:
         return flask.Response(
-            response="The predictor only supports application/json content type", status=415, mimetype="text/plain"
+            response="The predictor only supports application/json content type",
+            status=415,
+            mimetype="text/plain",
         )
 
     bucket_name = input_dict["bucket_name"]
     object_key = input_dict["object_key"]
+    model_name = input_dict.get("model_name", default_model_name)
     fd, filename = tempfile.mkstemp()
     try:
         os.close(fd)
@@ -66,9 +70,6 @@ def transformation():
     finally:
         os.unlink(filename)
 
-    payload = {
-        **input_dict,
-        "result": result
-    }
+    payload = {**input_dict, "result": result}
     response = json.dumps(payload)
     return flask.Response(response=response, status=200, mimetype="application/json")

whisper-image/requirements.txt

@@ -1,4 +1,4 @@
 boto3==1.24.96
 Flask==3.0.1
-gunicorn==21.2.0 
-openai-whisper==20230918
+gunicorn==21.2.0
+openai-whisper==20240930