Hello, I'm currently using a multi-node execution method. When max_active is set to 256 on two nodes, it can run approximately 15 steps. When max_active is reduced to 64 on four nodes, it runs 75 steps. When max_active is set to 128 on two nodes, it can run all 80 steps. I'm wondering if there's any pattern here. I'm really struggling to figure out how to set the max_active value.
Hello, I'm currently using a multi-node execution method. When max_active is set to 256 on two nodes, it can run approximately 15 steps. When max_active is reduced to 64 on four nodes, it runs 75 steps. When max_active is set to 128 on two nodes, it can run all 80 steps. I'm wondering if there's any pattern here. I'm really struggling to figure out how to set the max_active value.