We provide a sample code of
RNN text generation, which you can adjust to an arithmetic version.
1.
Data (30 pts) We provide an example of arithmetic data on google
drive:
https://drive.google.com/file/d/1sGVtqG4J5w6kJtsXuLWVEqbD6TY1U2ui/view?
u_lp=sharing_Lij h
ps://drive.ggle.com/file/d/lsOVIqG4J5w6kJtsXuLWVEqbD6TY1U2ul/view?usp=sharingl
1.
Load the data and split them into train and validation (10 pts). You
can use pandas, Dataset and Dataloader. 2. Generate your own version
of data (10 pts), At least three two-digit addition and subtraction
problems, and if you'd like, you can also try multiplication and
division. 3. Tokenize the text (10 pts). You can design your own
tokenizer or use any API (if available).
2.
Generation Models (30 pts)
1.
Model design (10 pts). You can use RNN, GRU or LSTM... whatever. (at
least one sequential model) 2. Train(finetune) the model (10 pts). 3.
Evaluate your model when you are training. (10 pts)
3.
Analysis (40 pts)
1.
Model analysis (10 pts) Include your model design and the process of
loss reduction and the results of validation and testing (if
available). 2. Dataset analysis(15 pts) What are the characteristics
of the datasets you have generated, and what is your method or
understanding of it? What adjustments did you make to the dataset?
How did they impact your results? (At least 2 variations, 5pts per
variation)
3.
Discussion (15 pts) During the training process, what impact does
different learning rates have? What impact does different batch sizes
have? What are the characteristics of the model you are using, and
why do you think it is suitable(not suitable) for this task?
4.
Bonus (10 pts) 1. Compare the performance of multiple models and
provide a brief analysis. (10 pts) Submission • Structure
o
Your should submit a .zip file with the name {
studerit_io_GAz_proje0t2. (eg. F1234567_GAI_Project2a ). It should be
unzipped into a directory with the same name, and the directory
structure should be:
fstudent_idl_GAI_Project2a
main.py (.ipynb) // the code you use to run the language models &
generate figur // include other scripts if you have more
requirements.txt // pip freeze > requirements.txt report.pdf //
your report file, be sure it is .pdf
o
TA will not run your code for this project, but please make sure that
you hand in the code that train the model(s) and executes the
generation. • Make it readable with comments, lest we would need to
refer to it under any circumstances. • If your code does not look
like it can reproduce the results described in your report, we would
consider a grade discount/ask you for a demo.
Submission
Deadline:
o
Note that the deadline is 9:00 in the morning. o Late submission
within 1 week will get a 10% discount, and 3 weeks will get a 30%
discount. o Submissions later than that will not be graded and you
get a 0 as a result.
