diff --git a/README.md b/README.md
index b31cd5a..96031d1 100644
--- a/README.md
+++ b/README.md
@@ -1,6 +1,6 @@
 # CS224u: Natural Language Understanding
 
-Code for [the Stanford course](http://web.stanford.edu/class/cs224u/)
+Code for [the Stanford course](http://web.stanford.edu/class/cs224u/). The code is written to run under Python 3.7; [setup.ipynb](setup.ipynb) provides  additional details.
 
 # Instructors
 
diff --git a/remove_solutions.py b/remove_solutions.py
deleted file mode 100644
index 1c71838..0000000
--- a/remove_solutions.py
+++ /dev/null
@@ -1,40 +0,0 @@
-"""Removes the answers from homework notebooks.
-"""
-import json
-import re
-import sys
-
-__author__ = 'Chris Potts'
-
-
-def main(src_filename):
-    output_filename = src_filename.replace('_solved', '')
-    doc = None
-    with open(src_filename, 'rt') as f:
-        doc = json.load(f)
-        new_cells = []
-        for i, cell in enumerate(doc['cells']):
-            if cell['cell_type'] == 'code':
-                cell_start = cell['source'][0]
-                if not re.search(r"^#+\s*SOLUTION", cell_start):
-                    removing = False
-                    new_source = []
-                    for line in cell['source']:
-                        if "# <<<<<<<<<< TO BE COMPLETED" in line:
-                            removing = True
-                        if not removing:
-                            new_source.append(line)
-                        if "# >>>>>>>>>>" in line:
-                            removing = False
-                        cell['source'] = new_source
-                    new_cells.append(cell)
-            else:
-                new_cells.append(cell)
-        doc['cells'] = new_cells
-
-    with open(output_filename, 'wt') as output:
-        json.dump(doc, output)
-
-if __name__ == '__main__':
-
-    main(sys.argv[1])
diff --git a/requirements.txt b/requirements.txt
index 7788d9d..b547a59 100644
--- a/requirements.txt
+++ b/requirements.txt
@@ -3,22 +3,38 @@
 # Only people working in their own virtual environments should install
 # them via this script.
 #
-# numpy==1.14.0
-# scipy==1.0.0
-# matplotlib==2.1.2
-# scikit-learn >= 0.19.1
-# nltk==3.2.5
-# python-dateutil==2.6.1
-# pytest==3.0.3
+# numpy>=1.15.0
+# scipy>=1.1.0
+# matplotlib>=3.0.0
+# scikit-learn >= 0.20.0
+# nltk>=3.4.0
+# python-dateutil>=2.7.0
+# pytest==4.0.0
 # jupyter==1.0.0
 #
 ######################################################################
 
 mittens>=0.1
 
+######################################################################
 # For NLTK, the data distibution needs to be downloaded separately:
+#
 # https://www.nltk.org/data.html
 
-# Tensorflow is also required. It should be installed via the
-# instructions provided by the TensorFlow team, which depend on your
-# computing environment: https://www.tensorflow.org/install/
+######################################################################
+# PyTorch is required. For installation instructions, see
+#
+# If you're using Anaconda, we suggest
+#
+# conda install -c pytorch pytorch
+
+######################################################################
+# Tensorflow is required. It should be installed via the instructions
+# provided by the TensorFlow team, which depend on your computing
+# environment:
+#
+# https://www.tensorflow.org/install/
+#
+# If you're using Anaconda, we suggest
+#
+# conda install -c conda-forge tensorflow