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}'

utils/dataloaders.py

@@ -115,7 +115,8 @@ def create_dataloader(path,
                       image_weights=False,
                       quad=False,
                       prefix='',
-                      shuffle=False):
+                      shuffle=False,
+                      seed=0):
     if rect and shuffle:
         LOGGER.warning('WARNING ⚠️ --rect is incompatible with DataLoader shuffle, setting shuffle=False')
         shuffle = False
@@ -140,7 +141,7 @@ def create_dataloader(path,
     sampler = None if rank == -1 else distributed.DistributedSampler(dataset, shuffle=shuffle)
     loader = DataLoader if image_weights else InfiniteDataLoader  # only DataLoader allows for attribute updates
     generator = torch.Generator()
-    generator.manual_seed(6148914691236517205 + RANK)
+    generator.manual_seed(6148914691236517205 + seed + RANK)
     return loader(dataset,
                   batch_size=batch_size,
                   shuffle=shuffle and sampler is None,

utils/segment/dataloaders.py

@@ -37,7 +37,8 @@ def create_dataloader(path,
                       prefix='',
                       shuffle=False,
                       mask_downsample_ratio=1,
-                      overlap_mask=False):
+                      overlap_mask=False,
+                      seed=0):
     if rect and shuffle:
         LOGGER.warning('WARNING ⚠️ --rect is incompatible with DataLoader shuffle, setting shuffle=False')
         shuffle = False
@@ -64,7 +65,7 @@ def create_dataloader(path,
     sampler = None if rank == -1 else distributed.DistributedSampler(dataset, shuffle=shuffle)
     loader = DataLoader if image_weights else InfiniteDataLoader  # only DataLoader allows for attribute updates
     generator = torch.Generator()
-    generator.manual_seed(6148914691236517205 + RANK)
+    generator.manual_seed(6148914691236517205 + seed + RANK)
     return loader(
         dataset,
         batch_size=batch_size,