Set a seed of generator with an option for more randomness when training several models with different seeds (#10486)

* set seed with parameter

Signed-off-by: Yonghye Kwon <developer.0hye@gmail.com>

* make seed to be a large number

* set seed with a parameter

* set a seed of dataloader with opt for more randomness

Signed-off-by: Yonghye Kwon <developer.0hye@gmail.com>
Co-authored-by: Glenn Jocher <glenn.jocher@ultralytics.com>

train.py

@@ -198,7 +198,8 @@ def train(hyp, opt, device, callbacks):  # hyp is path/to/hyp.yaml or hyp dictio
                                               image_weights=opt.image_weights,
                                               quad=opt.quad,
                                               prefix=colorstr('train: '),
-                                              shuffle=True)
+                                              shuffle=True,
+                                              seed=opt.seed)
     labels = np.concatenate(dataset.labels, 0)
     mlc = int(labels[:, 0].max())  # max label class
     assert mlc < nc, f'Label class {mlc} exceeds nc={nc} in {data}. Possible class labels are 0-{nc - 1}'