How can I remove data from Dataset

index = [1,2,3,....]
train_tensor.shape = torch.size(128,30,100)
train_label.shape=torch.size(128,1,1) 
train_set = dataset(train_tensor,train_label) 

Now , I want to remove data form Train_set using index?
I had tried this function:

def tensor_remove(tensors,index):
    # print("tensors:",tensors.shape[0])
    sub_tensors = torch.empty([0,30,100])
    if index == []:
        return tensors
    if tensors.shape[1] == 1:
        sub_tensors = torch.empty([0,1,1])
    last_idx = -1
    for idx in index:
        if idx != last_idx+1:
            sub_tensors = torch.cat((sub_tensors,tensors[last_idx+1:idx]),0)
        if idx == index[-1]:
            sub_tensors = torch.cat((sub_tensors,tensors[idx+1:]),0)
        last_idx = idx
    # print("sub_tensor:",sub_tensors.shape[0])
    return sub_tensors

It words!But when I running train.py ,it rasing CUDA out of memerys! so ,I guess maybe the function caused that problem! Please help me! thanks!