C3W3 -- how can I do the assignment locally? need `data` dir

Course Q&A TensorFlow Developer Professional Natural Language Processing in TensorFlow
1

The provided coursera jupyter environment for C3W3 is prohibitively slow, as it does not have a GPU configuration. I know this, because I ran tf.config.list_physical_devices('GPU') and got an empty list response. As such, training is taking on the order of 20-30min for a full 20 epochs, and at least 5 min for 5 epochs to gauge whether my solution is meeting the threshold criterion.

While I have a strong GPU locally that I could leverage, it seems that this particular assignment is currently hosted on github without the data directory. I was able to scrape out training_cleaned.csv from my Jupyter notebook, but could not download glove.6B.100d.txt. I was however able to print its lines out to the notebook console, but Iā€™m not confident on the data it spat out.

Anyway, guidance appreciated.

2

There are 2 options for you:

  1. On top right of coursera notebook environment, click on Lab Files and then Download all files to get the entire assignment.
  2. Download zip from stanford and use glove.6B.100d.txt inside it.
1 Like
3

Thanks, I figured it out. You guys should update the assignment to include instructions on how to get the dataset.

4

I have tried this one, but when I try to submit the assignment, it required me to have pkl file with ipynb file. May I ask if I can upload the pkl file from my local environment since I cannot connect to the online notebook? Thanks

5

If the grader needs a pkl file, you can upload it from anywhere.