Added ONNX compatible broadcast_like function (#152)

- Also simplified the hybrid_clip implementation.

rl_coach/architectures/mxnet_components/heads/ppo_head.py

@@ -11,7 +11,7 @@ from rl_coach.utils import eps
 from rl_coach.architectures.mxnet_components.heads.head import Head, HeadLoss, LossInputSchema,\
     NormalizedRSSInitializer
 from rl_coach.architectures.mxnet_components.heads.head import LOSS_OUT_TYPE_LOSS, LOSS_OUT_TYPE_REGULARIZATION
-from rl_coach.architectures.mxnet_components.utils import hybrid_clip
+from rl_coach.architectures.mxnet_components.utils import hybrid_clip, broadcast_like
 
 
 LOSS_OUT_TYPE_KL = 'kl_divergence'
@@ -146,7 +146,7 @@ class MultivariateNormalDist:
         sigma_b_inv = self.F.linalg.potri(self.F.linalg.potrf(alt_dist.sigma))
         term1a = mx.nd.batch_dot(sigma_b_inv, self.sigma)
         # sum of diagonal for batch of matrices
-        term1 = (self.F.eye(self.num_var).broadcast_like(term1a) * term1a).sum(axis=-1).sum(axis=-1)
+        term1 = (broadcast_like(self.F, self.F.eye(self.num_var), term1a) * term1a).sum(axis=-1).sum(axis=-1)
         mean_diff = (alt_dist.mean - self.mean).expand_dims(-1)
         mean_diff_t = (alt_dist.mean - self.mean).expand_dims(-2)
         term2 = self.F.batch_dot(self.F.batch_dot(mean_diff_t, sigma_b_inv), mean_diff).reshape_like(term1)
@@ -155,7 +155,6 @@ class MultivariateNormalDist:
         return 0.5 * (term1 + term2 - self.num_var + term3)
 
 
-
 class CategoricalDist:
     def __init__(self, n_classes: int, probs: nd_sym_type, F: ModuleType=mx.nd) -> None:
         """
@@ -284,7 +283,7 @@ class ContinuousPPOHead(nn.HybridBlock):
             of shape (batch_size, time_step, action_mean).
         """
         policy_means = self.dense(x)
-        policy_std = log_std.exp().expand_dims(0).broadcast_like(policy_means)
+        policy_std = broadcast_like(F, log_std.exp().expand_dims(0), policy_means)
         return policy_means, policy_std
 
 

rl_coach/architectures/mxnet_components/utils.py

@@ -253,13 +253,23 @@ def hybrid_clip(F: ModuleType, x: nd_sym_type, clip_lower: nd_sym_type, clip_upp
     :param clip_upper: upper bound used for clipping, should be of shape (1,)
     :return: clipped data
     """
-    x_clip_lower = clip_lower.broadcast_like(x)
-    x_clip_upper = clip_upper.broadcast_like(x)
-    x_clipped = F.stack(x, x_clip_lower, axis=0).max(axis=0)
-    x_clipped = F.stack(x_clipped, x_clip_upper, axis=0).min(axis=0)
+    x_clip_lower = broadcast_like(F, clip_lower, x)
+    x_clip_upper = broadcast_like(F, clip_upper, x)
+    x_clipped = F.minimum(F.maximum(x, x_clip_lower), x_clip_upper)
     return x_clipped
 
 
+def broadcast_like(F: ModuleType, x: nd_sym_type, y: nd_sym_type) -> nd_sym_type:
+    """
+    Implementation of broadcast_like using broadcast_add and broadcast_mul because ONNX doesn't support broadcast_like.
+    :param F: backend api, either `mxnet.nd` or `mxnet.sym` (if block has been hybridized).
+    :param x: input to be broadcast
+    :param y: tensor to broadcast x like
+    :return: broadcast x
+    """
+    return F.broadcast_mul(x, (y * 0) + 1)
+
+
 def get_mxnet_activation_name(activation_name: str):
     """
     Convert coach activation name to mxnet specific activation name

rl_coach/tests/architectures/mxnet_components/test_utils.py

@@ -141,7 +141,14 @@ def test_hybrid_clip():
     a = mx.nd.array((1,))
     b = mx.nd.array((2,))
     clipped = hybrid_clip(F=mx.nd, x=x, clip_lower=a, clip_upper=b)
-    assert (np.isclose(a= clipped.asnumpy(), b=(1, 1.5, 2))).all()
+    assert (np.isclose(a=clipped.asnumpy(), b=(1, 1.5, 2))).all()
+
+
+@pytest.mark.unit_test
+def test_broadcast_like():
+    x = nd.ones((1, 2)) * 10
+    y = nd.ones((100, 100, 2)) * 20
+    assert mx.test_utils.almost_equal(x.broadcast_like(y).asnumpy(), broadcast_like(nd, x, y).asnumpy())
 
 
 @pytest.mark.unit_test