CodeTrans model for source code summarization csharp

Pretrained model on programming language csharp using the t5 large model architecture. It was first released in this repository. This model is trained on tokenized csharp code functions: it works best with tokenized csharp functions.

Model description

This CodeTrans model is based on the t5-large model. It has its own SentencePiece vocabulary model. It used multi-task training on 13 supervised tasks in the software development domain and 7 unsupervised datasets. It is then fine-tuned on the source code summarization task for the csharp code snippets.

Intended uses & limitations

The model could be used to generate the description for the csharp function or be fine-tuned on other csharp code tasks. It can be used on unparsed and untokenized csharp code. However, if the csharp code is tokenized, the performance should be better.

How to use

Here is how to use this model to generate csharp function documentation using Transformers SummarizationPipeline:

from transformers import AutoTokenizer, AutoModelWithLMHead, SummarizationPipeline

pipeline = SummarizationPipeline(
    model=AutoModelWithLMHead.from_pretrained("SEBIS/code_trans_t5_large_source_code_summarization_csharp_multitask_finetune"),
    tokenizer=AutoTokenizer.from_pretrained("SEBIS/code_trans_t5_large_source_code_summarization_csharp_multitask_finetune", skip_special_tokens=True),
    device=0
)

tokenized_code =  "public static DateTime ParseUnixDateTime ( double unixTime ) { var dt = new DateTime ( CODE_INTEGER , CODE_INTEGER , CODE_INTEGER , CODE_INTEGER , CODE_INTEGER , CODE_INTEGER , CODE_INTEGER , System . DateTimeKind . Utc ) ; dt = dt . AddSeconds ( unixTimeStamp ) . ToLocalTime ( ) ; return dt ; }"
pipeline([tokenized_code])

Run this example in colab notebook.

Training data

The supervised training tasks datasets can be downloaded on Link

Training procedure

Multi-task Pretraining

The model was trained on a single TPU Pod V3-8 for 260,000 steps in total, using sequence length 512 (batch size 4096). It has a total of approximately 220M parameters and was trained using the encoder-decoder architecture. The optimizer used is AdaFactor with inverse square root learning rate schedule for pre-training.

Fine-tuning

This model was then fine-tuned on a single TPU Pod V2-8 for 100 steps in total, using sequence length 512 (batch size 256), using only the dataset only containing csharp code.

Evaluation results

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

Test results :

Language / Model	Python	SQL	C#
CodeTrans-ST-Small	8.45	17.55	19.74
CodeTrans-ST-Base	9.12	15.00	18.65
CodeTrans-TF-Small	10.06	17.71	20.40
CodeTrans-TF-Base	10.94	17.66	21.12
CodeTrans-TF-Large	12.41	18.40	21.43
CodeTrans-MT-Small	13.11	19.15	22.39
CodeTrans-MT-Base	13.37	19.24	23.20
CodeTrans-MT-Large	13.24	19.40	23.57
CodeTrans-MT-TF-Small	12.10	18.25	22.03
CodeTrans-MT-TF-Base	10.64	16.91	21.40
CodeTrans-MT-TF-Large	12.14	19.98	21.10
CODE-NN	--	18.40	20.50

Created by Ahmed Elnaggar | LinkedIn and Wei Ding | LinkedIn