OpenCoder-LLM/opc-annealing-corpus

opc-annealing-corpus是OpenCoder数据集中的一个附加组件，用于退火阶段。它包含三个主要部分：algorithmic_corpus、synthetic_code_snippet和synthetic_qa。algorithmic_corpus是从The Stack v2中采样的算法相关代码，synthetic_code_snippet是通过重写algorithmic_corpus生成的高质量代码片段，synthetic_qa是通过改编algorithmic_corpus生成的高质量问答对。这些数据在OpenCoder的退火阶段被使用，并通过消融实验验证了其有效性。

The opc-annealing-corpus is an additional component incorporated into OpenCoder during the annealing phase. It consists of three main parts: algorithmic_corpus, synthetic_code_snippet, and synthetic_qa. The algorithmic_corpus is algorithm-related code sampled from The Stack v2, synthetic_code_snippet is high-quality code snippets generated by rewriting the algorithmic_corpus, and synthetic_qa is high-quality Q&A pairs generated by adapting the algorithmic_corpus. These data are used in the annealing phase of OpenCoder, and their effectiveness has been validated through ablation experiments.