[chore] move fair_dev into fairscale (#1078)

Co-authored-by: Min Xu <min.xu.public@gmail.com>

tests/optim/test_oss_adascale.py

@@ -22,8 +22,8 @@ from torch.nn import Linear, Sequential
 from torch.nn.parallel import DistributedDataParallel as DDP
 from torch.optim import SGD
 
-from fair_dev.testing.golden_testing_data import adascale_test_data
-from fair_dev.testing.testing import skip_if_single_gpu
+from fairscale.fair_dev.testing.golden_testing_data import adascale_test_data
+from fairscale.fair_dev.testing.testing import skip_if_single_gpu
 from fairscale.optim import OSS, AdaScale, AdaScaleWrapper
 
 

tests/optim/test_single_node_adascale.py

@@ -19,9 +19,9 @@ from torch.nn import Linear, Sequential
 from torch.optim import SGD
 from torch.optim.lr_scheduler import LambdaLR
 
-from fair_dev.testing.golden_testing_data import adascale_test_data
-from fair_dev.testing.testing import make_cudnn_deterministic, skip_if_no_cuda
-from fair_dev.testing.testing_memory import find_tensor_by_shape
+from fairscale.fair_dev.testing.golden_testing_data import adascale_test_data
+from fairscale.fair_dev.testing.testing import make_cudnn_deterministic, skip_if_no_cuda
+from fairscale.fair_dev.testing.testing_memory import find_tensor_by_shape
 from fairscale.optim import AdaScale
 
 

tests/utils/test_reduce_scatter_bucketer.py

@@ -12,7 +12,7 @@ from unittest import mock
 from parameterized import parameterized
 import torch
 
-from fair_dev.testing.testing import dist_init, spawn_for_all_world_sizes
+from fairscale.fair_dev.testing.testing import dist_init, spawn_for_all_world_sizes
 from fairscale.internal import torch_version
 from fairscale.internal.reduce_scatter_bucketer import ReduceScatterBucketer