SEBIS / code_trans_t5_small_code_documentation_generation_go_multitask_finetune

CodeTrans model for code documentation generation go

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

Model description

This CodeTrans model is based on the t5-small 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 code documentation generation task for the go function/method.

Intended uses & limitations

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

How to use

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

from transformers import AutoTokenizer, AutoModelWithLMHead, SummarizationPipeline

pipeline = SummarizationPipeline(
    model=AutoModelWithLMHead.from_pretrained("SEBIS/code_trans_t5_small_code_documentation_generation_go_multitask_finetune"),
    tokenizer=AutoTokenizer.from_pretrained("SEBIS/code_trans_t5_small_code_documentation_generation_go_multitask_finetune", skip_special_tokens=True),
    device=0
)

tokenized_code = "func ( pr * Progress ) needSnapshotAbort ( ) bool { return pr . State == ProgressStateSnapshot && pr . Match >= pr . PendingSnapshot   }"
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 half million 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 2000 steps in total, using sequence length 512 (batch size 256), using only the dataset only containing go code.

Evaluation results

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

Test results :

Created by Ahmed Elnaggar | LinkedIn and Wei Ding | LinkedIn