Began converting A3C to KerasModel

deepchem/models/keras_model.py

@@ -224,11 +224,11 @@ def _create_training_ops(self, example_batch):
     if tf.executing_eagerly():
       return
     self._label_placeholders = [
-        tf.placeholder(dtype=tf.as_dtype(t), shape=x.shape)
+        tf.placeholder(dtype=tf.as_dtype(t), shape=(None,) + x.shape[1:])
         for x, t in zip(example_batch[1], self._label_dtypes)
     ]
     self._weights_placeholders = [
-        tf.placeholder(dtype=tf.as_dtype(t), shape=x.shape)
+        tf.placeholder(dtype=tf.as_dtype(t), shape=(None,) + x.shape[1:])
         for x, t in zip(example_batch[2], self._weights_dtypes)
     ]
     self._loss_tensor = self._loss_fn(
@@ -937,6 +937,7 @@ def restore(self, checkpoint=None):
       checkpoint will be chosen automatically.  Call get_checkpoints() to get a
       list of all available checkpoints.
     """
+    self._ensure_built()
     if checkpoint is None:
       checkpoint = tf.train.latest_checkpoint(self.model_dir)
     if checkpoint is None:
@@ -972,6 +973,7 @@ def __call__(self, outputs, labels, weights):
         shape = tuple(w.shape.as_list())
       else:
         shape = w.shape
+      shape = tuple(-1 if x is None else x for x in shape)
       w = tf.reshape(w, shape + (1,) * (len(losses.shape) - len(w.shape)))
     loss = losses * w
     return tf.reduce_mean(loss) + sum(self.model.losses)

deepchem/rl/__init__.py

@@ -63,6 +63,7 @@ def __init__(self,
         self._state_dtype = numpy.float32
     else:
       self._state_dtype = state_dtype
+    print(self._state_dtype)
 
   @property
   def state(self):
@@ -185,6 +186,13 @@ class Policy(object):
   or even on different computers.
   """
 
+  def create_model(self, **kwargs):
+    raise NotImplemented("Subclasses must implement this")
+
+  @property
+  def output_names(self):
+    raise NotImplemented("Subclasses must implement this")
+
   def create_layers(self, state, **kwargs):
     """Create the TensorGraph Layers that define the policy.