Update attention.py

mesh_tensorflow/transformer/attention.py

@@ -94,16 +94,20 @@ def attention(q,
           "Dividing attention z-loss loss by num_microbatches={}".format(
               context.num_microbatches))
       z_loss /= context.num_microbatches
-    if context.train:
-      mtf.scalar_summary("attention_z_loss", z_loss)
+    mtf.scalar_summary("attention_z_loss", z_loss)
     z_loss *= z_loss_coeff
     context.losses.append(mtf.cast(z_loss, v.dtype))
 
   weights = mtf.softmax(logits, memory_length_dim, extra_logit=extra_logit)
   weights = mtf.cast(weights, v.dtype)
-  weights = mtf.dropout(
-      weights, context.train, 1.0 - dropout_rate,
-      noise_shape=weights.shape - dropout_broadcast_dims)
+  if context:
+    weights = mtf.dropout(
+        weights, context.train, 1.0 - dropout_rate,
+        noise_shape=weights.shape - dropout_broadcast_dims)
+  else:
+    weights = mtf.dropout(
+        weights, False, 1.0 - dropout_rate,
+        noise_shape=weights.shape - dropout_broadcast_dims)
   outputs_shape = q.shape - key_dim + value_dim
   outputs = mtf.einsum([weights, v], outputs_shape)
   outputs = mtf.reshape(outputs, orig_q_shape - key_dim + value_dim)