fixed zero level 3 dtype bug (#76)

colossalai/amp/apex_amp/__init__.py

@@ -1,7 +1,6 @@
 from .apex_amp import ApexAMPOptimizer
 import torch.nn as nn
 from torch.optim import Optimizer
-import apex.amp as apex_amp
 
 
 def convert_to_apex_amp(model: nn.Module,
@@ -19,6 +18,7 @@ def convert_to_apex_amp(model: nn.Module,
     :return: (model, optimizer)
     :rtype: Tuple
     """
+    import apex.amp as apex_amp
     model, optimizer = apex_amp.initialize(model, optimizer, **amp_config)
     optimizer = ApexAMPOptimizer(optimizer)
     return model, optimizer

colossalai/zero/__init__.py

@@ -30,11 +30,7 @@ def convert_to_zero(model: nn.Module,
     :rtype: Tuple
     """
     assert level == 2 or level == 3, 'Only ZERO Optimizer Level 2 and 3 are provided'
-    if level == 2:
-        if is_no_pp_or_last_stage():
-            model = NaiveAMPModel(model, output_to_fp32=True)
-        else:
-            model = NaiveAMPModel(model, output_to_fp32=False)
+    model = NaiveAMPModel(model, output_to_fp32=False)
 
     if level == 2:
         optimizer = ZeroRedundancyOptimizer_Level_2(init_optimizer=optimizer, **zero_config)

colossalai/zero/zero_redundancy_optimizer_level_3.py

@@ -695,13 +695,23 @@ class ZeroRedundancyOptimizer_Level_3(Optimizer):
                 },
                 "aio": aio_config
             }
-            remote_device = offload_param_config['device']
+
+            if offload_param_config is not None:
+                remote_device = offload_param_config['device']
+            else:
+                remote_device = None
+
+            if offload_optimizer_config is not None:
+                pin_memory = offload_optimizer_config.get(OFFLOAD_OPTIMIZER_PIN_MEMORY, False)
+            else:
+                pin_memory = False
+
             group = None
             if gpc.is_initialized(ParallelMode.DATA):
                 group = gpc.get_group(ParallelMode.DATA)
             Init(module=module, data_parallel_group=group, dtype=self.dtype,
                  remote_device=remote_device, config_dict_or_path=ds_config,
-                 pin_memory=offload_optimizer_config[OFFLOAD_OPTIMIZER_PIN_MEMORY])
+                 pin_memory=pin_memory)
 
         for m in module.modules():
             _init_external_params(m)

tests/test_zero_data_parallel/test_zero_level_3.py

@@ -89,10 +89,10 @@ def run_dist(rank, world_size):
     model.train()
     for idx, (data, label) in enumerate(train_dataloader):
         engine.zero_grad()
-        data = data.cuda().half()
+        data = data.cuda()
         label = label.cuda()
 
-        output = engine(data).float()
+        output = engine(data)
         loss = engine.criterion(output, label)
 
         engine.backward(loss)
@@ -104,7 +104,6 @@ def run_dist(rank, world_size):
 
 
 @pytest.mark.dist
-@pytest.mark.skip("Level 3 has unknown bug so skip this test for now")
 def test_zero_level_3():
     world_size = 4
     run_func = partial(run_dist, world_size=world_size)

tests/test_zero_tensor_parallel/test_vit_2d_level_3.py

@@ -108,7 +108,6 @@ def run_2d_parallel_vision_transformer_level_3(rank, world_size):
 
 
 @pytest.mark.dist
-@pytest.mark.skip("Level 3 has unknown bug so skip this test for now")
 def test_3d_vit_zero_level_3():
     world_size = 8
     run_func = partial(run_2d_parallel_vision_transformer_level_3, world_size=world_size)