a question about dataset augmentation

[XiaoyuShi97]

https://github.com/princeton-vl/RAFT/blob/aac9dd54726caf2cf81d8661b07663e220c5586d/core/datasets.py#L47 Hi. I am a bit confused about the datasets. Let's say the size of dataset is1k, batch size is 1 and training iteration is 4k and num_worker is 1. Then will this line be executed for 4 time? The point is that if so, then the augmentation pattern repeats for 4 times, since every time you reset the seed. Thanks!

[gddcx]

https://github.com/princeton-vl/RAFT/blob/aac9dd54726caf2cf81d8661b07663e220c5586d/core/datasets.py#L51 https://github.com/princeton-vl/RAFT/blob/aac9dd54726caf2cf81d8661b07663e220c5586d/core/datasets.py#L45 Line51 set self.init_seed=True
Line45 will be equal to False in the subsequent steps.