Some questions about shuffle in GroupSampler

[Fengzeheng]

mmdetection/mmdet/datasets/samplers/group_sampler.py

Lines 23 to 41 in c14dd6c

	def __iter__(self):
	indices = []
	for i, size in enumerate(self.group_sizes):
	if size == 0:
	continue
	indice = np.where(self.flag == i)[0]
	assert len(indice) == size
	np.random.shuffle(indice)
	num_extra = int(np.ceil(size / self.samples_per_gpu)
	) * self.samples_per_gpu - len(indice)
	indice = np.concatenate(
	[indice, np.random.choice(indice, num_extra)])
	indices.append(indice)
	indices = np.concatenate(indices)
	indices = [
	indices[i * self.samples_per_gpu:(i + 1) * self.samples_per_gpu]
	for i in np.random.permutation(
	range(len(indices) // self.samples_per_gpu))
	]

1. If there is no random seed associated with epoch, what is the function of shuffle here？
2. The random seed for numpy seems to have been set in the training script by set_random_seed(), here is whether the same indice is generated in each epoch.