[MLA-1519] Don't mark action_probs as an output node. (#4613)

* remove action_probs from output nodes

* changelog

* pin cattrs upper version

* print pip freeze results

* add comment about cattrs version

.github/workflows/pytest.yml

         python -m pip install --progress-bar=off -r test_requirements.txt -c ${{ matrix.pip_constraints }}
         python -m pip install --progress-bar=off -e ./gym-unity -c ${{ matrix.pip_constraints }}
     - name: Save python dependencies
-      run: pip freeze > pip_versions-${{ matrix.python-version }}.txt
+      run: |
+        pip freeze > pip_versions-${{ matrix.python-version }}.txt
+        cat pip_versions-${{ matrix.python-version }}.txt
     - name: Run pytest
       run: pytest --cov=ml-agents --cov=ml-agents-envs --cov=gym-unity --cov-report html --junitxml=junit/test-results-${{ matrix.python-version }}.xml -p no:warnings
     - name: Upload pytest test results

com.unity.ml-agents/CHANGELOG.md

 #### com.unity.ml-agents (C#)
 - The Barracuda dependency was upgraded to 1.1.2 (#4571)
 #### ml-agents / ml-agents-envs / gym-unity (Python)
+- The `action_probs` node is no longer listed as an output in TensorFlow models (#4613).
 
 ### Bug Fixes
 #### com.unity.ml-agents (C#)

ml-agents/mlagents/trainers/tf/model_serialization.py

     ]
 )
 
-POSSIBLE_OUTPUT_NODES = frozenset(
-    ["action", "action_probs", "recurrent_out", "value_estimate"]
-)
+POSSIBLE_OUTPUT_NODES = frozenset(["action", "recurrent_out", "value_estimate"])
 
 MODEL_CONSTANTS = frozenset(
     [

ml-agents/setup.py

         # https://github.com/Unity-Technologies/ml-agents/blob/master/docs/Installation.md#windows-installing-pytorch
         'torch>=1.6.0,<1.8.0;platform_system!="Windows"',
         "tensorboard>=1.15",
-        "cattrs>=1.0.0",
+        # cattrs 1.1.0 dropped support for python 3.6.
+        "cattrs>=1.0.0,<1.1.0",
         "attrs>=19.3.0",
         'pypiwin32==223;platform_system=="Windows"',
     ],