Update wording in transformer notebook

Author: ajweeks

workshops/transformer/transformer.ipynb

@@ -337,7 +337,7 @@
             "cell_type": "markdown",
             "metadata": {},
             "source": [
-                "#### **Step 2:** Feed tokens to model\n",
+                "#### **Step 2:** Map tokens to logits\n",
                 "\n",
                 "\n",
                 "After running a feed-forward pass on our model with an input with shape `[batch, seq_len]`, we get an output of shape `[batch, seq_len, vocab_size]`. The `[i, j, :]`-th element of our output is a vector of logits representing our prediction for the `j+1`-th token in the `i`-th sequence.\n"

workshops/transformer/transformer_hard.ipynb

@@ -337,7 +337,7 @@
             "cell_type": "markdown",
             "metadata": {},
             "source": [
-                "#### **Step 2:** Feed tokens to model\n",
+                "#### **Step 2:** Map tokens to logits\n",
                 "\n",
                 "\n",
                 "After running a feed-forward pass on our model with an input with shape `[batch, seq_len]`, we get an output of shape `[batch, seq_len, vocab_size]`. The `[i, j, :]`-th element of our output is a vector of logits representing our prediction for the `j+1`-th token in the `i`-th sequence.\n"

workshops/transformer/transformer_normal.ipynb

@@ -337,7 +337,7 @@
             "cell_type": "markdown",
             "metadata": {},
             "source": [
-                "#### **Step 2:** Feed tokens to model\n",
+                "#### **Step 2:** Map tokens to logits\n",
                 "\n",
                 "\n",
                 "After running a feed-forward pass on our model with an input with shape `[batch, seq_len]`, we get an output of shape `[batch, seq_len, vocab_size]`. The `[i, j, :]`-th element of our output is a vector of logits representing our prediction for the `j+1`-th token in the `i`-th sequence.\n"