[tune] Support all serializable objects in config (#2287)

* wip

* order

* lint

python/ray/tune/examples/tune_mnist_ray.py

@@ -36,6 +36,7 @@ import ray
 from ray.tune import grid_search, run_experiments, register_trainable
 
 from tensorflow.examples.tutorials.mnist import input_data
+import numpy as np
 
 import tensorflow as tf
 
@@ -227,6 +228,9 @@ if __name__ == '__main__':
         },
         'config': {
             'activation': grid_search(['relu', 'elu', 'tanh']),
+            # You can pass any serializable object as well
+            'foo': grid_search([np.array([1, 2]),
+                                np.array([2, 3])]),
         },
     }
 

python/ray/tune/logger.py

@@ -89,12 +89,12 @@ class _JsonLogger(Logger):
     def _init(self):
         config_out = os.path.join(self.logdir, "params.json")
         with open(config_out, "w") as f:
-            json.dump(self.config, f, sort_keys=True, cls=_CustomEncoder)
+            json.dump(self.config, f, sort_keys=True, cls=_SafeFallbackEncoder)
         local_file = os.path.join(self.logdir, "result.json")
         self.local_out = open(local_file, "w")
 
     def on_result(self, result):
-        json.dump(result._asdict(), self, cls=_CustomEncoder)
+        json.dump(result._asdict(), self, cls=_SafeFallbackEncoder)
         self.write("\n")
 
     def write(self, b):
@@ -150,9 +150,9 @@ class _VisKitLogger(Logger):
         self._file.close()
 
 
-class _CustomEncoder(json.JSONEncoder):
+class _SafeFallbackEncoder(json.JSONEncoder):
     def __init__(self, nan_str="null", **kwargs):
-        super(_CustomEncoder, self).__init__(**kwargs)
+        super(_SafeFallbackEncoder, self).__init__(**kwargs)
         self.nan_str = nan_str
 
     def iterencode(self, o, _one_shot=False):
@@ -171,12 +171,15 @@ class _CustomEncoder(json.JSONEncoder):
         return _iterencode(o, 0)
 
     def default(self, value):
-        if np.isnan(value):
-            return None
-        if np.issubdtype(value, float):
-            return float(value)
-        if np.issubdtype(value, int):
-            return int(value)
+        try:
+            if np.isnan(value):
+                return None
+            if np.issubdtype(value, float):
+                return float(value)
+            if np.issubdtype(value, int):
+                return int(value)
+        except Exception:
+            return str(value)  # give up, just stringify it (ok for logs)
 
 
 def pretty_print(result):
@@ -186,5 +189,5 @@ def pretty_print(result):
         if v is not None:
             out[k] = v
 
-    cleaned = json.dumps(out, cls=_CustomEncoder)
+    cleaned = json.dumps(out, cls=_SafeFallbackEncoder)
     return yaml.safe_dump(json.loads(cleaned), default_flow_style=False)

python/ray/tune/variant_generator.py

@@ -1,3 +1,7 @@
+from __future__ import absolute_import
+from __future__ import division
+from __future__ import print_function
+
 import copy
 import json
 import numpy
@@ -6,6 +10,7 @@ import random
 import types
 
 from ray.tune import TuneError
+from ray.tune.logger import _SafeFallbackEncoder
 from ray.tune.trial import Trial
 from ray.tune.config_parser import make_parser, json_to_resources
 
@@ -17,7 +22,7 @@ def to_argv(config):
         if isinstance(v, str):
             argv.append(v)
         else:
-            argv.append(json.dumps(v))
+            argv.append(json.dumps(v, cls=_SafeFallbackEncoder))
     return argv