Total

[kirillepam]

Main part:

• Max 10 points. Repository exists, has public access, doesn’t contain unnecessary files or trained models and it can be cloned using the “git clone” command (Important! Ensure that your project has public access, otherwise it won’t be possible to check your homework)

  Great job! 10/10

• Max 10 points. requirements.txt includes all the necessary libraries and doesn’t have any conflicts or unnecessary libraries. All these libraries can be simply installed using “pip install -r requirements.txt”

  Good job! It is very important to specify the exact library version. If you rely on a library to perform a certain task, but then the library deprecates a function which your package requires, at that point anyone installing your package would download the newest version of the dependency, one that no longer has access to the specific function upon which your package relies. Also, some results on inference might vary from version to version. 8/10

• Max 20 points. Docker image can be built using dockerfile from the training directory. It should have all the necessary data and code for the training. The running container should automatically start the training process, save the trained model (in the container) and show the quality calculated on the test sample.

  Good job! But you forget estimate model quality on test set. 15/20

• Max 20 points. Docker image can be built using dockerfile from the inference directory. It should have all the necessary data, a prepared model (from the previous step) and the code for the inference. The running container should automatically run the trained model on the inference data and save the result in the additional file.

  Great job! 20/20

Bonus points:

• Max 10 points. Include tests for the main parts of your code. Examples: (Is your model can be trained with default parameters?).

  No any unit test, sorry. 0/10

• Max 10 points. Include exceptions for the problematic parts of your code. Example: (running inference without any trained model should definitely return an error)

  Good job! Some exceptions are better not to catch. 8/10

• Max 10 points. The code in the repository follows pep8 style guide (PEP 8 – Style Guide for Python Code | peps.python.org: check indentations, the style of names and comments)

  Good job! 10/10

• Max 5 points. Add logging that shows the quality of your model after each training, size of used dataset before training and inference and the time spent on training and inference

  Good job! You could add some datasets info. 4/5

• Max 5 points. README.md file is well-structured. It has the instructions of how to run the training and the inference

  Great job! 5/5

Total:

10/10 + 8/10 + 15/20 + 20/20 + 0/10 + 8/10 + 10/10 + 4/5 + 5/5 = 80

[bobrarity]

@kirillepam appreciate the detailed feedback, thanks)

[bobrarity]

@kirillepam Could you please grade it on Learn platform too, thanks in advance)

[kirillepam]

Unfortunately, you missed the deadline. I need some help of course supervisor