- You can run locally all programming assignments of all courses in this one environment
- docker promise you will get the same environment with only one step
- All assignments tested (besides 3 assignments in firs two weeks of Course4, but it should work)
0. Install docker and docker-compose (if you haven't)
Both docker ce version and docker-compose for your os.
1. Follow step B and C in this thread
https://www.coursera.org/learn/neural-networks-deep-learning/discussions/all/threads/29j3DW9WEeiqiRKZ5Tzn-A/replies/rQbsl29XEei0dhKA653RhA, but we don't need ...-condaenv.txt, because I have had thos into one txt
2. Activate environment
In project directory, docker-compose up, depends on your network speed it may take few hours to download the 5G image for the first time.
3. Copy the last line output of console to browser
you can see like http://0.0.0.0:8888/?token=b2c061a4..., everytime the token will change, so we can't avoid this step.
(Course2 mostly) 4. fix error related to matplotlib
matplotlib scatter: TypeError: unhashable type: 'numpy.ndarray' occurs when running assignments locally, not the problem of this project, see https://stackoverflow.com/questions/49840380/matplotlib-scatter-typeerror-unhashable-type-numpy-ndarray. Anyway, search code for plot and append .ravel().tolist() to Y or train_Y .... plot used mostly in Course2, and less in other course.
(Course5 Week2) 5. fix error of encoding
add encoding='utf-8' to open in read_glove_vecs used in both word2vector and emoji assignments.
If deeplearning.ai change their environments (install/remove/update packages), and you encounter problems when running pythonbooks, using this docker image as base to write the changes to a new dockerfile could be good way to solve it.