when I open option(a) for assignment 2 it runs with error:

[CanoeFZH]

diff --git a/assignment2/assignment2.py b/assignment2/assignment2.py
index 28e45bd..942fa2e 100644
--- a/assignment2/assignment2.py
+++ b/assignment2/assignment2.py
@@ -150,8 +150,8 @@ def train(epochs):

             # Options:
             # (a)
-            # embed_to_hid_weights_gradient = np.dot(
-            #    np.transpose(back_propagated_deriv_1), embedding_layer_state)
+            embed_to_hid_weights_gradient = np.dot(
+                    np.transpose(back_propagated_deriv_1), embedding_layer_state)

             # (b)
             # embed_to_hid_weights_gradient = np.dot(embedding_layer_state,
@@ -169,7 +169,7 @@ def train(epochs):

             # Options
             # (a)
-            # hid_bias_gradient = np.sum(back_propagated_deriv_1, axis=1)
+            hid_bias_gradient = np.sum(back_propagated_deriv_1, axis=1)

             # (b)
             # hid_bias_gradient = np.sum(back_propagated_deriv_1, axis=0)
@@ -188,8 +188,8 @@ def train(epochs):

             # Options
             # (a)
-            # back_propagated_deriv_2 = np.dot(embed_to_hid_weights,
-            #                                     back_propagated_deriv_1)
+            back_propagated_deriv_2 = np.dot(embed_to_hid_weights,
+                    back_propagated_deriv_1)

             # (b)
             # back_propagated_deriv_2 = np.dot(back_propagated_deriv_1,
@@ -385,11 +385,14 @@ def fprop(input_batch, word_embedding_weights, embed_to_hid_weights,

     # Combine three inputs, turn them into 1D vector
     ib_tmp = np.reshape(input_batch, (1,-1), order="F").ravel()
+    print (ib_tmp.shape)
     # Take rows from weight matrices, row number equal to word index.
     # Then transpose.
     wew_tmp = np.transpose(word_embedding_weights[ib_tmp])
+    print (wew_tmp.shape)
     # Combine three nearby columns
     embedding_layer_state = np.reshape(wew_tmp, (numhid1 * numwords,-1), order="F")
+    print(embed_to_hid_weights.shape)

     ## COMPUTE STATE OF HIDDEN LAYER
     ## Compute inputs to hidden units.
@@ -404,7 +407,7 @@ def fprop(input_batch, word_embedding_weights, embed_to_hid_weights,

     # Options
     # (a)
-    # hidden_layer_state = 1 / (1 + np.exp(inputs_to_hidden_units))
+    hidden_layer_state = 1 / (1 + np.exp(inputs_to_hidden_units))

     # (b)
     # hidden_layer_state = 1 / (1 - np.exp(-inputs_to_hidden_units))
@@ -423,8 +426,8 @@ def fprop(input_batch, word_embedding_weights, embed_to_hid_weights,

     # Options
     # (a)
-    # inputs_to_softmax = np.dot(np.transpose(hid_to_output_weights),
-    #                             hidden_layer_state)
+    inputs_to_softmax = np.dot(np.transpose(hid_to_output_weights),
+            hidden_layer_state)
     # output_bias_tmp = np.tile(output_bias,(1,batchsize))
     # inputs_to_softmax = inputs_to_softmax + output_bias_tmp

with error

→ python assignment2.py
('Epoch', 1)
(300,)
(50, 300)
(150, 200)
Traceback (most recent call last):
  File "assignment2.py", line 564, in <module>
    model = train(10)
  File "assignment2.py", line 154, in train
    np.transpose(back_propagated_deriv_1), embedding_layer_state)
ValueError: shapes (100,200) and (150,100) not aligned: 200 (dim 1) != 150 (dim 0)