Increase min supported tensorflow to 1.14.0 (#4411)

.circleci/config.yml

       pip_constraints:
         type: string
         description: Constraints file that is passed to "pip install". We constraint older versions of libraries for older python runtime, in order to help ensure compatibility.
-      enforce_onnx_conversion:
-        type: integer
-        default: 0
-        description: Whether to raise an exception if ONNX models couldn't be saved.
     executor: << parameters.executor >>
     working_directory: ~/repo
 
-      TEST_ENFORCE_ONNX_CONVERSION: << parameters.enforce_onnx_conversion >>
+      # Whether to raise an exception if ONNX models couldn't be saved.
+      TEST_ENFORCE_ONNX_CONVERSION: 1
 
     steps:
       - checkout
           name: python_3.7.3
           executor: python373
           pyversion: 3.7.3
-          # Test python 3.7 with the newest supported versions
+          # Test python 3.7 with the newest supported versions of 1.x
-          # Make sure ONNX conversion passes here (recent version of tensorflow 1.x)
-          enforce_onnx_conversion: 1
-      - build_python:
-          name: python_3.7.3+tf2
-          executor: python373
-          pyversion: 3.7.3
-          # Test python 3.7 with the newest supported versions
-          pip_constraints: test_constraints_max_tf2_version.txt
       - build_python:
           name: python_3.8.2+tf2.2
           executor: python382

com.unity.ml-agents/CHANGELOG.md

 - Update Barracuda to 1.0.2.
 - Enabled C# formatting using `dotnet-format`.
 #### ml-agents / ml-agents-envs / gym-unity (Python)
+- Experimental PyTorch support has been added. Use `--torch` when running `mlagents-learn`, or add
+`framework: pytorch` to your trainer configuration (under the behavior name) to enable it.
+Note that PyTorch 1.6.0 or greater should be installed to use this feature; see
+[the PyTorch website](https://pytorch.org/) for installation instructions. (#4335)
+- The minimum supported version of TensorFlow was increased to 1.14.0. (#4411)
 
 ### Bug Fixes
 #### com.unity.ml-agents (C#)
 - The interaction between EnvManager and TrainerController was changed; EnvManager.advance() was split into to stages,
 and TrainerController now uses the results from the first stage to handle new behavior names. This change speeds up
 Python training by approximately 5-10%. (#4259)
-- Experimental PyTorch support has been added. Use `--torch` when running `mlagents-learn`, or add
-`framework: pytorch` to your trainer configuration (under the behavior name) to enable it.
-Note that PyTorch 1.6.0 or greater should be installed to use this feature; see
-[the PyTorch website](https://pytorch.org/) for installation instructions. (#4335)
 
 ### Minor Changes
 #### com.unity.ml-agents (C#)

ml-agents/setup.py

         "Pillow>=4.2.1",
         "protobuf>=3.6",
         "pyyaml>=3.1.0",
-        "tensorflow>=1.7,<3.0",
+        "tensorflow>=1.14,<3.0",
         "cattrs>=1.0.0",
         "attrs>=19.3.0",
         'pypiwin32==223;platform_system=="Windows"',

test_constraints_max_tf2_version.txt

 # For projects with upper bounds, we should periodically update this list to the latest release version
 grpcio>=1.23.0
 numpy>=1.17.2
-tensorflow==2.2.0rc3
+tensorflow==2.3.0
 h5py>=2.10.0

test_constraints_min_version.txt

 numpy==1.14.1
 Pillow==4.2.1
 protobuf==3.6
-tensorflow==1.7.0
+tensorflow==1.14.0
 h5py==2.9.0

test_requirements.txt

 # PyTorch tests are here for the time being, before they are used in the codebase.
 torch>=1.5.0
 
-# onnx doesn't currently have a wheel for 3.8
-tf2onnx>=1.5.5;python_version<'3.8'
+tf2onnx>=1.5.5