This model is derived from OpenCoder-1.5B-Base by applying additional context extension fine-tuning. The repository context is composed using the Duplication composer, more details on which, along with others, can be found in the On Pretraining for Project-Level Code Completion paper (arxiv). Specifically, Section A.1 of the Appendix describes the context composition method, and Table 3 provides a comparison with other composers from the same collection.

We publish this checkpoint to support the reproducibility and accessibility of our research results.

Quickstart

python

import torch
from transformers import AutoModelForCausalLM, AutoTokenizer

model_name = "JetBrains-Research/OpenCoder-1.5B-Duplication"
tokenizer_name = "infly/OpenCoder-1.5B-Base"

model = AutoModelForCausalLM.from_pretrained(model_name,
                                             torch_dtype=torch.bfloat16,
                                             device_map="auto",
                                             trust_remote_code=True)
tokenizer = AutoTokenizer.from_pretrained(tokenizer_name, trust_remote_code=True)

inputs = tokenizer("# write a quick sort algorithm", return_tensors="pt")
outputs = model.generate(**inputs.to(model.device), max_new_tokens=256)

result = tokenizer.decode(outputs[0], skip_special_tokens=True)
print(result)