You can either load the model and further fine-tune it for a target task (as done for the puzzle solver), or you can experiment with mask-filling directly with this model as in the following example:

python

from transformers import AutoTokenizer, AutoModelWithLMHead, pipeline

tokenizer = AutoTokenizer.from_pretrained("tals/roberta_python")
model = AutoModelWithLMHead.from_pretrained("tals/roberta_python")

demo = pipeline("fill-mask", model=model, tokenizer=tokenizer)

code = """sum= 0
for i in range(<mask>):
    sum += i
assert sum == 6
"""
demo(code)

BibTeX entry and citation info

bibtex

@inproceedings{
      schuster2021programming,
      title={Programming Puzzles},
      author={Tal Schuster and Ashwin Kalyan and Alex Polozov and Adam Tauman Kalai},
      booktitle={Thirty-fifth Conference on Neural Information Processing Systems Datasets and Benchmarks Track (Round 1)},
      year={2021},
      url={https://openreview.net/forum?id=fe_hCc4RBrg}
}