ブルーアーカイブ
Qwen2.5-VL-3B GRPO 训练分析
Reproducing GRPO training for Qwen2.5-VL-3B with EasyR1 and analyzing its performance, stability, and resource usage on Geometry3K.
无标签
使用 Conda 配置 EasyR1 训练环境
Configure EasyR1 with Conda for Qwen2.5-VL-3B GRPO training on four RTX 3090 GPUs.
无标签
Qwen2.5-7B-Instruct 部署
A deployment guide for Qwen2.5-7B-Instruct on a multi-GPU Linux server using vLLM
无标签
RLHF algorithms, from PPO to DAPO
RLHF algorithms including PPO, DPO, GRPO, DAPO
