automatic-triton-programming

【开发与工程】Train specialized LLMs to generate optimized Triton GPU kernels using RL with dual rewards for correctness and syntax compliance. 8B model achieves parity with Claude-Sonnet and DeepSeek-R1 by combining supervised fine-tuning on curated code pairs with RL exploration beyond imitation learning ceilings.

独立 GitHub 快照 · 自动分类 · 以原作者文档为准