[RLlib] Exploration API (+EpsilonGreedy sub-class). (#6974)

python/ray/tune/logger.py

@@ -214,7 +214,7 @@ class TBXLogger(Logger):
                         full_attr, value, global_step=step)
                 # In case TensorboardX still doesn't think it's a valid value
                 # (e.g. `[[]]`), warn and move on.
-                except ValueError:
+                except (ValueError, TypeError):
                     logger.warning(
                         "You are trying to log an invalid value ({}={}) "
                         "via {}!".format(full_attr, value,

python/ray/tune/utils/util.py

@@ -164,7 +164,7 @@ def deep_update(original, new_dict, new_keys_allowed, whitelist):
         if k not in original:
             if not new_keys_allowed:
                 raise Exception("Unknown config parameter `{}` ".format(k))
-        if isinstance(original.get(k), dict):
+        if isinstance(original.get(k), dict) and isinstance(value, dict):
             if k in whitelist:
                 deep_update(original[k], value, True, [])
             else: