CS224n: Natural Language Processing with Deep Learning

These past weeks I've spent several weeks on the CS224n course from Stanford University. Here are my solutions to the assignments. These solutions are for the 2017 version of the course.

Installation

1. Install Anaconda
2. go to assignmentX where X is either 1, 2, 3 using a Terminal:

   $ cd \path\to\assignment1

3. create a python 2.7 environnment using

   $ conda env -n cs224n python=2.7 anaconda

4. activate your environment using (add source before activate if you're working with Linux/Mac)

   $ activate cs231n

5. install the dependencies using requirements.txt

   $ pip install -r requirements.txt

6. Don't forget to deactivate your environment when you're done (add source before deactivate if you're on Linux/Mac)

   $ deactivate cs224n

Issues

If you're working on windows you won't be able to do assignment 2 because the code is for Python 2.7 and TensorFlow is only available with Python 3.5 on windows (at the time I'm writting these lines). So you will need to create another environment for TensorFlow using:

1. create a python 3.5 environnment using

   $ conda env -n tensorflow python=3.5 anaconda

2. activate your environment using (add source before activate if you're working with Linux/Mac)

   $ activate tensorflow

3. install the dependencies using requirements.txt

   $ pip install -r requirements.txt

4. install TensorFlow (you can also install tensorflow-gpu for the gpu support)

   $ pip install tensorflow

then you need to convert all the python files from Python 2 to Python 3. To do so you can simply use 2to3 which is a script included in anaconda to convert Python file from version 2 to version 3 automatically. Simply do:

Assignment 3

If you're using my files you will need to use Python 3 (on Windows TensorFlow is not compatible with Python 2). Yet, if you want to convert the files from assignment 3 to Python 3, beside using 2to3, you will need to:

replace line 105 from data_util.py to:

with open(os.path.join(path, "features.pkl"), "wb") as f:

(Just add a b to avoid) "write() argument must be str not bytes" error Also replace line 113 by:

with open(os.path.join(path, "features.pkl"), "rb") as f:

You will also need to replace: tf.nn.rnn_cell.RNNCell to tf.contrib.rnn.core_rnn_cell.RNNCell in files q2_gru_cell.py and q2_rnn_cell.py.

$ 2to3 --output-dir=python3-version/assignment2 -W -n assignment2

Note: If you cloned my repository you won't need to transform the code from Python 2.7 to Python 3.5 as I've already did it

More information

I wrote several blog posts accessible from my website if you want to understand in detail how the code works.