SEBIS/code_trans_t5_large_commit_generation_multitask_finetune

This model is trained on tokenized git commit: it works best with tokenized commit . It can be used on unparsed and untokenized commit changes . It has a total of approximately 220M parameters and was trained using the encoder-decoder architecture . The model was then fine-tuned on a single TPU Pod V2-8 for 3,000 steps in total, using sequence length 512 (batch size 256) The model could be used to generate the git commit message for the . commit changes or be fine-tuneed on other relevant tasks . The code-trans model was trained on 13 supervised tasks in the software development domain and 7 unsupervised datasets,