Update set_default_tensor_type to set_default_dtype (#5734)

pytorch · Nov 29, 2023 · d3172e7 · d3172e7
1 parent d7927ce
commit d3172e7
Show file tree

Hide file tree

Showing 11 changed files with 11 additions and 11 deletions.
diff --git a/test/bench.py b/test/bench.py
@@ -128,7 +128,7 @@ def run_benchmarks(args):
   args, benchs = parser.parse_known_args()
   args.benchs = benchs
 
-  torch.set_default_tensor_type('torch.FloatTensor')
+  torch.set_default_dtype(torch.float32)
   torch_xla._XLAC._xla_set_use_full_mat_mul_precision(
       use_full_mat_mul_precision=True)
   run_benchmarks(args)
diff --git a/test/spmd/test_train_spmd_imagenet.py b/test/spmd/test_train_spmd_imagenet.py
@@ -372,7 +372,7 @@ def test_loop_fn(loader, epoch):
   if FLAGS.profile:
     server = xp.start_server(FLAGS.profiler_port)
 
-  torch.set_default_tensor_type('torch.FloatTensor')
+  torch.set_default_dtype(torch.float32)
   accuracy = train_imagenet()
   if accuracy < FLAGS.target_accuracy:
     print('Accuracy {} is below target {}'.format(accuracy,

diff --git a/test/test_operations_hlo.py b/test/test_operations_hlo.py
@@ -69,7 +69,7 @@ def test_dropout_by_u8_mask(self):
 
 
 if __name__ == '__main__':
-  torch.set_default_tensor_type('torch.FloatTensor')
+  torch.set_default_dtype(torch.float32)
   torch.manual_seed(42)
   torch_xla._XLAC._xla_set_use_full_mat_mul_precision(
       use_full_mat_mul_precision=True)

diff --git a/test/test_profile_mp_mnist.py b/test/test_profile_mp_mnist.py
@@ -198,7 +198,7 @@ def test_loop_fn(loader):
 
 
 def _mp_fn(index, flags):
-  torch.set_default_tensor_type('torch.FloatTensor')
+  torch.set_default_dtype(torch.float32)
   accuracy = train_mnist(flags, dynamic_graph=True, fetch_often=True)
   if flags.tidy and os.path.isdir(flags.datadir):
     shutil.rmtree(flags.datadir)

diff --git a/test/test_train_mp_imagenet.py b/test/test_train_mp_imagenet.py
@@ -358,7 +358,7 @@ def test_loop_fn(loader, epoch):
 def _mp_fn(index, flags):
   global FLAGS
   FLAGS = flags
-  torch.set_default_tensor_type('torch.FloatTensor')
+  torch.set_default_dtype(torch.float32)
   accuracy = train_imagenet()
   if accuracy < FLAGS.target_accuracy:
     print('Accuracy {} is below target {}'.format(accuracy,

diff --git a/test/test_train_mp_imagenet_amp.py b/test/test_train_mp_imagenet_amp.py
@@ -298,7 +298,7 @@ def test_loop_fn(loader, epoch):
 def _mp_fn(index, flags):
   global FLAGS
   FLAGS = flags
-  torch.set_default_tensor_type('torch.FloatTensor')
+  torch.set_default_dtype(torch.float32)
   accuracy = train_imagenet()
   if accuracy < FLAGS.target_accuracy:
     print('Accuracy {} is below target {}'.format(accuracy,

diff --git a/test/test_train_mp_imagenet_fsdp.py b/test/test_train_mp_imagenet_fsdp.py
@@ -385,7 +385,7 @@ def test_loop_fn(loader, epoch):
 def _mp_fn(index, flags):
   global FLAGS
   FLAGS = flags
-  torch.set_default_tensor_type('torch.FloatTensor')
+  torch.set_default_dtype(torch.float32)
   accuracy = train_imagenet()
   if accuracy < FLAGS.target_accuracy:
     print('Accuracy {} is below target {}'.format(accuracy,

diff --git a/test/test_train_mp_mnist.py b/test/test_train_mp_mnist.py
@@ -203,7 +203,7 @@ def test_loop_fn(loader):
 
 
 def _mp_fn(index, flags):
-  torch.set_default_tensor_type('torch.FloatTensor')
+  torch.set_default_dtype(torch.float32)
   accuracy = train_mnist(flags)
   if flags.tidy and os.path.isdir(flags.datadir):
     shutil.rmtree(flags.datadir)

diff --git a/test/test_train_mp_mnist_amp.py b/test/test_train_mp_mnist_amp.py
@@ -211,7 +211,7 @@ def test_loop_fn(loader):
 
 
 def _mp_fn(index, flags):
-  torch.set_default_tensor_type('torch.FloatTensor')
+  torch.set_default_dtype(torch.float32)
   accuracy = train_mnist(flags)
   if flags.tidy and os.path.isdir(flags.datadir):
     shutil.rmtree(flags.datadir)

diff --git a/test/test_train_mp_mnist_fsdp_with_ckpt.py b/test/test_train_mp_mnist_fsdp_with_ckpt.py
@@ -313,7 +313,7 @@ def test_loop_fn(model, loader):
 
 
 def _mp_fn(index, flags):
-  torch.set_default_tensor_type('torch.FloatTensor')
+  torch.set_default_dtype(torch.float32)
   accuracy = train_mnist(flags)
   if flags.tidy and os.path.isdir(flags.datadir):
     shutil.rmtree(flags.datadir)

diff --git a/test/test_train_mp_mnist_zero1.py b/test/test_train_mp_mnist_zero1.py
@@ -184,7 +184,7 @@ def test_loop_fn(loader):
 
 
 def _mp_fn(index, flags):
-  torch.set_default_tensor_type('torch.FloatTensor')
+  torch.set_default_dtype(torch.float32)
   accuracy = train_mnist(flags)
   if flags.tidy and os.path.isdir(flags.datadir):
     shutil.rmtree(flags.datadir)