Fix links

week2/week2-NER.ipynb

@@ -411,7 +411,7 @@
     "Now, let us specify the layers of the neural network. First, we need to perform some preparatory steps: \n",
     " \n",
     "- Create embeddings matrix with [tf.Variable](https://www.tensorflow.org/api_docs/python/tf/Variable). Specify its name (*embeddings_matrix*), type  (*tf.float32*), and initialize with random values.\n",
-    "- Create forward and backward LSTM cells. TensorFlow provides a number of RNN cells ready for you. We suggest that you use *BasicLSTMCell*, but you can also experiment with other types, e.g. GRU cells. [This](http://colah.github.io/posts/2015-08-Understanding-LSTMs/) blogpost could be interesting if you want to learn more about the differences.\n",
+    "- Create forward and backward LSTM cells. TensorFlow provides a number of RNN cells ready for you. We suggest that you use *LSTMCell*, but you can also experiment with other types, e.g. GRU cells. [This](http://colah.github.io/posts/2015-08-Understanding-LSTMs/) blogpost could be interesting if you want to learn more about the differences.\n",
     "- Wrap your cells with [DropoutWrapper](https://www.tensorflow.org/api_docs/python/tf/contrib/rnn/DropoutWrapper). Dropout is an important regularization technique for neural networks. Specify all keep probabilities using the dropout placeholder that we created before.\n",
     " \n",
     "After that, you can build the computation graph that transforms an input_batch:\n",
@@ -515,7 +515,7 @@
    },
    "source": [
     "During training we do not need predictions of the network, but we need a loss function. We will use [cross-entropy loss](http://ml-cheatsheet.readthedocs.io/en/latest/loss_functions.html#cross-entropy), efficiently implemented in TF as \n",
-    "[cross entropy with logits](https://www.tensorflow.org/api_docs/python/tf/nn/softmax_cross_entropy_with_logits). Note that it should be applied to logits of the model (not to softmax probabilities!). Also note,  that we do not want to take into account loss terms coming from `<PAD>` tokens. So we need to mask them out, before computing [mean](https://www.tensorflow.org/api_docs/python/tf/reduce_mean)."
+    "[cross entropy with logits](https://www.tensorflow.org/api_docs/python/tf/nn/softmax_cross_entropy_with_logits_v2). Note that it should be applied to logits of the model (not to softmax probabilities!). Also note,  that we do not want to take into account loss terms coming from `<PAD>` tokens. So we need to mask them out, before computing [mean](https://www.tensorflow.org/api_docs/python/tf/reduce_mean)."
    ]
   },
   {
@@ -529,7 +529,7 @@
     "def compute_loss(self, n_tags, PAD_index):\n",
     "    \"\"\"Computes masked cross-entopy loss with logits.\"\"\"\n",
     "    \n",
-    "    # Create cross entropy function function (tf.nn.softmax_cross_entropy_with_logits)\n",
+    "    # Create cross entropy function function (tf.nn.softmax_cross_entropy_with_logits_v2)\n",
     "    ground_truth_tags_one_hot = tf.one_hot(self.ground_truth_tags, n_tags)\n",
     "    loss_tensor =  ######### YOUR CODE HERE #############\n",
     "    \n",
@@ -557,7 +557,7 @@
    "source": [
     "The last thing to specify is how we want to optimize the loss. \n",
     "We suggest that you use [Adam](https://www.tensorflow.org/api_docs/python/tf/train/AdamOptimizer) optimizer with a learning rate from the corresponding placeholder. \n",
-    "You will also need to apply [clipping](https://www.tensorflow.org/api_guides/python/train#Gradient_Clipping) to eliminate exploding gradients. It can be easily done with [clip_by_norm](https://www.tensorflow.org/api_docs/python/tf/clip_by_norm) function. "
+    "You will also need to apply clipping to eliminate exploding gradients. It can be easily done with [clip_by_norm](https://www.tensorflow.org/api_docs/python/tf/clip_by_norm) function. "
    ]
   },
   {