AI Research & Engineering: RecSys, Search, NLP, Generative AI and Beyond

Tag DeepSeek-Coder

DeepSeek-Coder 详解:从 file-level 到 repo-level,代码模型训练范式的关键演进(DeepSeek 系列第 3 篇)

DeepSeek-Coder(arXiv:2401.14196)详解:从 file-level 升级到 repo-level 训练 + 拓扑排序、FIM 双模混合 (50% PSM + NTP)、16K 长上下文,让 6.7B 模型在 HumanEval / MBPP 上追平 CodeLlama-34B。

Loading

© 2026 Yudong‘s Blog — Powered by WordPress

Theme by Anders NorenUp ↑