Edit model card

CodeGen (CodeGen-NL 16B)

Sharded version of codegen

This model was sharded using torch.float16. Use the code below to load this model, configure the device_map for your GPU/CPU split.

First pull the model.

git clone https://huggingface.co/abacaj/codegen-16B-nl-sharded
cd codegen-16B-nl-sharded
git-lfs install
git pull
def load_model_sharded():
    config = AutoConfig.from_pretrained("abacaj/codegen-16B-nl-sharded")
    tokenizer = AutoTokenizer.from_pretrained("abacaj/codegen-16B-nl-sharded")

    with init_empty_weights():
        model = AutoModelForCausalLM.from_config(config)

    device_map = infer_auto_device_map(
            0: "20GiB",
            "cpu": "110GiB",

    model = load_checkpoint_and_dispatch(

    return model, tokenizer

Model description

CodeGen is a family of autoregressive language models for program synthesis from the paper: A Conversational Paradigm for Program Synthesis 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, 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-NL 16B in the paper, where "NL" means it is pre-trained on the Pile and "16B" refers to the number of trainable parameters.

Training data

This checkpoint (CodeGen-NL 16B) was pre-trained on the Pile, a large-scale curated dataset created by EleutherAI. Parts of the dataset include code data.

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 for more details.

Evaluation results

We evaluate our models on two code generation benchmark: HumanEval and MTPB. Please refer to the paper 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:

from transformers import AutoTokenizer, AutoModelForCausalLM
tokenizer = AutoTokenizer.from_pretrained("Salesforce/codegen-16B-nl")
model = AutoModelForCausalLM.from_pretrained("Salesforce/codegen-16B-nl")

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

  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},
Downloads last month