ML4T Software Setup
The repository has been made private for the Fall 2017 semester, and so the links to the repository below will no longer be visible for you. A zip file containing the grading script and any template code or data will be linked off of each assignment’s individual wiki page. A zip file containing the grading and util modules, as well as the data, is available here: File:ML4T_2021Sum.zip. The instructions on running the test scripts provided below still apply.
Most of the projects in this class will be graded automatically. As of the summer 2017 semester, we are providing the grading scripts with the template code for each of the projects, so that students can test their code to make sure they are API compatible.
- Your code MUST run properly on Gradescope. If you would like to develop on your personal machine and are comfortable installing libraries by hand, you can follow the instructions here: ML4T_Local_Environment. Note that these instructions are from an earlier version of the class, but should work reasonably well.
- We use a specific, static dataset for this course, which is provided as part of the repository detailed below. If you download your own data from Yahoo (or elsewhere), you will get wrong answers on assignments.
- We reserve the right to modify the grading script while maintaining API compatibility with what is described on the project pages. This includes modifying or withholding test cases, changing point values to match the given rubric, and changing timeout limits to accommodate grading deadlines. The scripts are provided as a convenience to help students avoid common pitfalls or mistakes, and are intended to be used as a sanity check. Passing all tests does not guarantee full credit on the assignment, and should be considered a necessary but not sufficient condition for completing an assignment.
- Using github.gatech.edu to back up your work is a very good idea which we encourage, however, make sure that you do not make your solutions to the assignments public. It’s easy to accidentally do this, so please be careful:
- Do not put your solutions in a public repository. Repositories on github.com are public by default. The Georgia Tech GitHub, github.gatech.edu, provides the same interface and allows for free private repos for students.
Getting code templates
As of Spring 2018, code for each of the individual assignments is provided in zip files, linked to on the individual project page. The data, grading module, and util.py, which are common across all assignments, are available here File:ML4T_2021Sum.zip (same file as above).
Running the grading scripts
The above zip files contain the grading scripts, data, and util.py for all assignments. Some project pages will also have a link to a zip file containing a directory with some template code, which you should extract in the same directory that contains the data/ and grading/directories, and util.py, (ML4T_2020Fall/). To complete the assignments you’ll need to modify the templates according to the assignment description.
To test your code, you’ll need to set up your PYTHONPATH to include the grading module and the utility module util.py, which are both one directory up from the project directories. Here’s an example of how to run the grading script for the optional (deprecated) assignment Assess Portfolio (note, grade_anlysis.py is included in the template zip file for Assess Portfolio):
PYTHONPATH=../:. python grade_analysis.py
which assumes you’re typing from the folder ML4T_2021SUM/assess_portfolio/. This will print out a lot of information, and will also produce two text files: points.txt and comments.txt. It will probably be helpful to scan through all of the output printed out in order to trace errors to your code, while comments.txt will contain a succinct summary of which test cases failed and the specific errors (without the backtrace). Here’s an example of the contents of comments.txt for the first assignment using the unchanged template:
<pre>--- Summary --- Tests passed: 0 out of 3 --- Details --- Test #0: failed Test case description: Wiki example 1 IncorrectOutput: One or more stats were incorrect. Inputs: start_date: 2010-01-01 00:00:00 end_date: 2010-12-31 00:00:00 symbols: ['GOOG', 'AAPL', 'GLD', 'XOM'] allocs: [0.2, 0.3, 0.4, 0.1] start_val: 1000000 Wrong values: cum_ret: 0.25 (expected: 0.255646784534) avg_daily_ret: 0.001 (expected: 0.000957366234238) sharpe_ratio: 2.1 (expected: 1.51819243641) Test #1: failed Test case description: Wiki example 2 ...