prime-rl 0.10.0发布，IPO成默认损失
Prime Intellect发布prime-rl 0.10.0和verifiers 0.4.0，并将新的INTELLECT Policy Optimization（IPO）设为异步强化学习的默认损失函数。此次更新加入DeepSeek V4 Flash、Qwen3.8 Flash Next和Nemotron-H等模型的优化支持，同时将评估功能整合进prime-rl。Prime Intellect称，在其Qwen3-235B-A22B监督微调对比中，prime-rl的速度是torchtitan的1.15倍，但这些基准数据来自厂商自身测试。
该版本把新的异步RL损失、面向性能的训练内核和统一评估工具整合在一起，对构建大模型强化学习工作流的实践者具有直接参考价值。
prime-rl 0.10.0: faster training, our new RL loss IPO, and one unified eval stack
Today we're releasing prime-rl v0.10.0 and verifiers v0.4.0. We added our new RL algorithm IPO as the default, tailored for async RL stability, as well as performance improvements and new model support.
Training throughput
The trainer's defaults favor memory, so large models fit on few GPUs. This release also documents the other end: a max-throughput recipe that keeps everything on GPU and compiles with fullgraph.
#Prime Intellect#prime-rl#verifiers#INTELLECT Policy Optimization#DeepSeek V4 Flash#Qwen3.8 Flash Next#Nemotron-H#GLM-5.3#Qwen3-235B-A22B#Qwen-VL#Kimi K2.5