CodeTrans model for program synthesis

Pretrained model on programming language lisp inspired DSL using the t5 small model architecture. It was first released in this repository.

Model description

This CodeTrans model is based on the t5-small model. It has its own SentencePiece vocabulary model. It used single-task training on Program Synthesis dataset.

Intended uses & limitations

The model could be used to generate lisp inspired DSL code given the human language description tasks.

How to use

Here is how to use this model to generate lisp inspired DSL code using Transformers SummarizationPipeline:

from transformers import AutoTokenizer, AutoModelWithLMHead, SummarizationPipeline

pipeline = SummarizationPipeline(
    model=AutoModelWithLMHead.from_pretrained("SEBIS/code_trans_t5_small_program_synthese"),
    tokenizer=AutoTokenizer.from_pretrained("SEBIS/code_trans_t5_small_program_synthese", skip_special_tokens=True),
    device=0
)

tokenized_code = "you are given an array of numbers a and a number b , compute the difference of elements in a and b"
pipeline([tokenized_code])

Run this example in colab notebook.

Training data

The supervised training tasks datasets can be downloaded on Link

Evaluation results

For the code documentation tasks, different models achieves the following results on different programming languages (in BLEU score):

Test results :

Language / Model	LISP
CodeTrans-ST-Small	89.43
CodeTrans-ST-Base	89.65
CodeTrans-TF-Small	90.30
CodeTrans-TF-Base	90.24
CodeTrans-TF-Large	90.21
CodeTrans-MT-Small	82.88
CodeTrans-MT-Base	86.99
CodeTrans-MT-Large	90.27
CodeTrans-MT-TF-Small	90.31
CodeTrans-MT-TF-Base	90.30
CodeTrans-MT-TF-Large	90.17
State of the art	85.80

Created by Ahmed Elnaggar | LinkedIn and Wei Ding | LinkedIn