Welcome to QCon SF2017. I have added some files in this repo. Please download. Also please complete all the tasks below. If you could not, do not worry. I will discuss them in the workshop.
NOTE, NOTE, NOTE: Though I do not wish to change the contents of the repo or the tasks listed, please check this repo once again on/after 10 Nov 2017 for any changes. All the changes will be listed under the section "CHANGES AFTER 10NOV2017".
Task(a): Install any python editor of your choice. I prefer PyCharm Community Edition 2016.3 (though a later version is available, I had difficulties in running it on my iMac). You can find the relevant links here. https://www.jetbrains.com/pycharm/download/previous.html. I will use the editor to view and code and discuss. Task(b): Install Anaconda Navigator. This will help you install certain basic packages like Numpy, Pandas, Keras, etc. (https://docs.continuum.io/anaconda/install/). I may talk about these packages but nevertheless you need these packages for future ML experiments
Here are some tasks that you would need to complete to get the best benefit of the workshop. Please complete the tasks in the order as they appear. If you cannot, do not worry.
Task 1: Read intro stuff about Tensorflow here https://www.tensorflow.org/get_started/get_started. If you do not understand, do not worry. I will discuss in the workshop. Task 2: In your laptop, create a new directory QCon-NMT. You will download all the relevant files I have provided in github into this directory. Task 3: Go to the directory QCon-NMT Task 4: Go to https://github.com/tensorflow/nmt/tree/tf-1.2 and scroll down to the section "Installing the Tutorial". Follow the instructions in the section to download the tutorial. Task 5: Install Tensorflow as per the instructions in the page. Check that the installation was successful. Instructions to check the successful installation are provided in the same page. Task 6: Run the NMT tutorial as per the instructions in the page. In other words, go to the site https://github.com/tensorflow/nmt/tree/tf-1.2. Scroll down to the section "Hands-on-Let's train an NMT model". Follow the instructions to train the model. Depending upon the capability of your machine, it may take several hours (more than 3-6 hrs) to train. If you have difficulty in running, don't worry. We will try to solve it in the class. Meanwhile if you want to see the output file of the successful training of the model, check the file "QCon- NMT compilation output file".
NOTE: If the installation is unsuccessful for any reasons, ONE of the reasons, I found, could be because of cloning of the repo. Hence instead of running the script "git clone https://github.com/tensorflow/nmt/" as mentioned in the site, try "clone/download" option.
I am excited to discuss about Machine Translation using Tensorflow. As you can image, Machine Translation is a huge exercise and I will try to explain the concepts and discuss tips on how to build one. I am sure that with the concepts discussed in the workshop, you will be equipped to develop and enhance Machine Translation models.
CHANGES AFTER 10NOV2017:----