Handled permutations error in augmentation

[saandeepa93]

Changes for issue. Made simple changes to augmentation module under dataloader/augmentation.py

warp = np.concatenate(np.random.permutation(splits)).ravel()

to

np.random.shuffle(splits)
warp = np.concatenate(splits).ravel()

[klay7w]

It perfectly solved my problem. Thank you^^

[evileleven]

Another question. Why just get the first channel (ret[i] = pat[0, warp]) in the 'permutation' function instead of the all channels (ret[i] = pat[:, warp])?

def permutation(x, max_segments=5, seg_mode="random"):
    # x: (batch_size, channels, timesteps)
    orig_steps = np.arange(x.shape[2])
    num_segs = np.random.randint(1, max_segments, size=(x.shape[0]))
    ret = np.zeros_like(x)
    for i, pat in enumerate(x):
        if num_segs[i] > 1:
            # print(num_segs[i])
            if seg_mode == "random":
                split_points = np.random.choice(x.shape[2] - 2, num_segs[i] - 1, replace=False)
                split_points.sort()
                splits = np.split(orig_steps, split_points)
            else:
                splits = np.array_split(orig_steps, num_segs[i])
            np.random.shuffle(splits)
            warp = np.concatenate(splits)
            # ret[i] = pat[0, warp]
            ret[i] = pat[:, warp]
        else:
            ret[i] = pat
    return torch.from_numpy(ret)