Open sanshuiii opened 9 months ago
I also meet with the same problem. Here are the results I reproduced.
But I at least understand what your problem is. You are using 4 GPUs, but in default .yaml it is 1 GPU. That means you multiply the batch size by 4 but you do not change the learning rate.
I am wondering if we are using the wrong configs? Since at our second trial, the 2nd stage acc even decreases. The configs and scripts used are as follows:
stage 1
stage 2