learning an API representation for library migration with program synthesis
-
Prerequisite:
- python 3.8.5 (anaconda)
-
Python packages dependencies:
- click
- z3-solver
- tensorflow
- pytorch
- rpy2
- transformers
- fuzzywuzzy
- matplotlib
- numpy
- nltk
- lark-parser
- mkl
- tqdm
The DL benchmarks can be found at api-learning-representation/autotesting/benchmarks, and the
data wrangling benchmarks at api-learning-representation/synthesis/synthesizer/dplyr_to_pd/dplyr.
- Add the repository to the PYTHONPATH environment variable:
$ export PYTHONPATH=/path/to/api-learning-representation
- To run a specific benchmark (DL) do the following:
$ cd synthesis/synthesizer/tf_to_torch
$ python3 torch_synthesizer.py -b name
- Conversly, to run a data wrangling task execute the following code:
$ cd synthesis/synthesizer/dplyr_to_pandas
$ python3 pd_synthesizer.py -b name
