Relax serve requirements (#1187)

flash/core/serve/flash_components.py

@@ -2,6 +2,7 @@
 from typing import Any, Callable, Mapping
 
 import torch
+from torch import Tensor
 
 from flash.core.data.batch import _ServeInputProcessor
 from flash.core.data.data_module import DataModule
@@ -41,6 +42,8 @@ def serialize(self, outputs) -> Any:  # pragma: no cover
                 result = self._output(output)
                 if isinstance(result, Mapping):
                     result = result[DataKeys.PREDS]
+                if isinstance(result, Tensor):
+                    result = result.tolist()
                 results.append(result)
         if len(results) == 1:
             return results[0]

flash_examples/serve/image_classification/inference_server.py

@@ -16,4 +16,4 @@
 model = ImageClassifier.load_from_checkpoint(
     "https://flash-weights.s3.amazonaws.com/0.7.0/image_classification_model.pt"
 )
-model.serve()
+model.serve(output="labels")

requirements/serve.txt

@@ -4,16 +4,10 @@ pyyaml
 cytoolz
 graphviz
 tqdm
-# until 1.0 release fastapi docs recommend pinning to MINOR releases.
-# https://fastapi.tiangolo.com/deployment/#fastapi-versions
-fastapi>=0.65.2,<0.66.0
-# to have full feature control of fastapi, manually install optional
-# dependencies rather than installing fastapi[all]
-# https://fastapi.tiangolo.com/#optional-dependencies
-pydantic>1.8.1,<2.0.0
+fastapi>=0.65.2
+pydantic>1.8.1
 starlette==0.14.2
-uvicorn[standard]>=0.12.0,<0.14.0
+uvicorn[standard]>=0.12.0
 aiofiles
 jinja2
-importlib-metadata>=0.12,<3;python_version<"3.8"
 torchvision