| --- |
| license: bsd-3-clause |
| --- |
| # CodeGen (CodeGen-Mono 350M) |
| This is a clone of CodeGen project which is optimized to run on CPU by using the ONNX optimisations. |
| The reason we created ONNX version of the original version is, we wanted to make it possible for ICortex kernel users |
| to easily generate code for their use case without using a GPU. |
| Original model can be found [here](https://huggingface.co/Salesforce/codegen-350M-mono). |
|
|
| ## Model description |
|
|
| CodeGen is a family of autoregressive language models for **program synthesis** from the paper: [A Conversational Paradigm for Program Synthesis](https://arxiv.org/abs/2203.13474) by Erik Nijkamp, Bo Pang, Hiroaki Hayashi, Lifu Tu, Huan Wang, Yingbo Zhou, Silvio Savarese, Caiming Xiong. The models are originally released in [this repository](https://github.com/salesforce/CodeGen), under 3 pre-training data variants (`NL`, `Multi`, `Mono`) and 4 model size variants (`350M`, `2B`, `6B`, `16B`). |
|
|
| The checkpoint included in this repository is denoted as **CodeGen-Mono 350M** in the paper, where "Mono" means the model is initialized with *CodeGen-Multi 350M* and further pre-trained on a Python programming language dataset, and "350M" refers to the number of trainable parameters. |
|
|
| ## Training data |
|
|
| This checkpoint (CodeGen-Mono 350M) was firstly initialized with *CodeGen-Multi 350M*, and then pre-trained on BigPython dataset. The data consists of 71.7B tokens of Python programming language. See Section 2.1 of the [paper](https://arxiv.org/abs/2203.13474) for more details. |
|
|
| ## Training procedure |
|
|
| CodeGen was trained using cross-entropy loss to maximize the likelihood of sequential inputs. |
| The family of models are trained using multiple TPU-v4-512 by Google, leveraging data and model parallelism. |
| See Section 2.3 of the [paper](https://arxiv.org/abs/2203.13474) for more details. |
|
|
| ## Evaluation results |
|
|
| We evaluate our models on two code generation benchmark: HumanEval and MTPB. Please refer to the [paper](https://arxiv.org/abs/2203.13474) for more details. |
|
|
|
|
| ## Intended Use and Limitations |
|
|
| As an autoregressive language model, CodeGen is capable of extracting features from given natural language and programming language texts, and calculating the likelihood of them. |
| However, the model is intended for and best at **program synthesis**, that is, generating executable code given English prompts, where the prompts should be in the form of a comment string. The model can complete partially-generated code as well. |
|
|
| ## How to use |
|
|
| This model can be easily loaded using the `AutoModelForCausalLM` functionality: |
|
|
| ```python |
| from transformers import AutoTokenizer, AutoModelForCausalLM |
| tokenizer = AutoTokenizer.from_pretrained("Salesforce/codegen-350M-mono") |
| model = AutoModelForCausalLM.from_pretrained("Salesforce/codegen-350M-mono") |
| |
| text = "def hello_world():" |
| input_ids = tokenizer(text, return_tensors="pt").input_ids |
| |
| generated_ids = model.generate(input_ids, max_length=128) |
| print(tokenizer.decode(generated_ids[0], skip_special_tokens=True)) |
| ``` |
|
|
| ## BibTeX entry and citation info |
|
|
| ```bibtex |
| @article{Nijkamp2022ACP, |
| title={A Conversational Paradigm for Program Synthesis}, |
| author={Nijkamp, Erik and Pang, Bo and Hayashi, Hiroaki and Tu, Lifu and Wang, Huan and Zhou, Yingbo and Savarese, Silvio and Xiong, Caiming}, |
| journal={arXiv preprint}, |
| year={2022} |
| } |
| ``` |
