[polish] use GLOBAL_MODEL_DATA_TRACER (#417)

colossalai/engine/ophooks/zero_hook.py

@@ -5,7 +5,7 @@ from colossalai.zero.shard_utils import BaseShardStrategy
 
 from ._base_ophook import BaseOpHook
 from colossalai.utils.memory_tracer.memstats_collector import MemStatsCollector
-from colossalai.utils.memory_tracer.model_data_memtracer import ModelDataTracer
+from colossalai.utils.memory_tracer.model_data_memtracer import GLOBAL_MODEL_DATA_TRACER
 from typing import Optional
 
 
@@ -25,7 +25,6 @@ class ZeroHook(BaseOpHook):
 
     def pre_fwd_exec(self, module: torch.nn.Module, *args):
         tensor_list = []
-        global_model_data_tracer = ModelDataTracer()
         for param in module.parameters():
             assert hasattr(param, 'col_attr')
             tensor_list.append(param.col_attr.data)
@@ -33,7 +32,7 @@ class ZeroHook(BaseOpHook):
         for param in module.parameters():
             if param.col_attr.data.device != self.computing_device:
                 param.col_attr.data.to(self.computing_device)
-                global_model_data_tracer.add_tensor(param.col_attr.data.payload)
+                GLOBAL_MODEL_DATA_TRACER.add_tensor(param.col_attr.data.payload)
             param.data = param.col_attr.data.payload
 
         if self._memstarts_collector:
@@ -50,7 +49,6 @@ class ZeroHook(BaseOpHook):
 
     def pre_bwd_exec(self, module: torch.nn.Module, input, output):
         tensor_list = []
-        global_model_data_tracer = ModelDataTracer()
         for param in module.parameters():
             assert hasattr(param, 'col_attr')
             tensor_list.append(param.col_attr.data)
@@ -58,7 +56,7 @@ class ZeroHook(BaseOpHook):
         for param in module.parameters():
             if param.col_attr.data.device != self.computing_device:
                 param.col_attr.data.to(self.computing_device)
-                global_model_data_tracer.add_tensor(param.col_attr.data.payload)
+                GLOBAL_MODEL_DATA_TRACER.add_tensor(param.col_attr.data.payload)
             param.data = param.col_attr.data.payload
             # Store local accumulated grad shard
             if param.grad is not None:

colossalai/utils/memory_tracer/allocator.py

@@ -1,5 +1,5 @@
 import torch
-from colossalai.utils.memory_tracer.model_data_memtracer import ModelDataTracer
+from colossalai.utils.memory_tracer.model_data_memtracer import GLOBAL_MODEL_DATA_TRACER
 
 
 def col_move_to_cpu(t: torch.Tensor):
@@ -7,7 +7,7 @@ def col_move_to_cpu(t: torch.Tensor):
     if t.device.type == 'cpu':
         return
 
-    ModelDataTracer().delete_tensor(t)
+    GLOBAL_MODEL_DATA_TRACER.delete_tensor(t)
     t.data = t.data.cpu()
 
 

colossalai/utils/memory_tracer/memstats_collector.py

@@ -1,4 +1,4 @@
-from colossalai.utils.memory_tracer.model_data_memtracer import ModelDataTracer
+from colossalai.utils.memory_tracer.model_data_memtracer import GLOBAL_MODEL_DATA_TRACER
 from .async_memtracer import get_cuda_memory_used
 from colossalai.utils import get_current_device
 
@@ -54,7 +54,7 @@ class MemStatsCollector:
         if self._start_flag:
             sampling_cnt = self._sampling_cnter.sampling_cnt
             assert sampling_cnt == len(self._overall_cuda)
-            self._model_data_cuda.append(ModelDataTracer().cuda_usage)
+            self._model_data_cuda.append(GLOBAL_MODEL_DATA_TRACER.cuda_usage)
             self._overall_cuda.append(get_cuda_memory_used(torch.device(f'cuda:{get_current_device()}')))
         self._sampling_cnter.advance()
 

colossalai/utils/memory_tracer/model_data_memtracer.py

@@ -5,10 +5,9 @@ import torch
 
 class ModelDataTracer(metaclass=SingletonMeta):
     """
-    A singleton to trace model data usage during runtime.
+    A tracer singleton to trace model data usage during runtime.
-    We have to trigger our API (trace_tensor, detach_tensor) when do model-data memory operation,
+    The tracer is designed to trace the memory layout change during model-data tensors allocation, releasing, and moving.
-    including allocation, releasing and moving.
+    To achieve this goal, the developers have to call `ModelDataTracer` in the corresponding code explicitly.
-
     NOTE() now the class only trace cuda memory usage
     """
 
@@ -32,3 +31,6 @@ class ModelDataTracer(metaclass=SingletonMeta):
     @property
     def cuda_usage(self):
         return self._cuda_usage
+
+
+GLOBAL_MODEL_DATA_TRACER = ModelDataTracer()

colossalai/utils/memory_tracer/test_memstats_collector.py

@@ -1,5 +1,5 @@
 from colossalai.utils.memory_tracer.memstats_collector import MemStatsCollector
-from colossalai.utils.memory_tracer.model_data_memtracer import ModelDataTracer
+from colossalai.utils.memory_tracer.model_data_memtracer import GLOBAL_MODEL_DATA_TRACER
 import torch
 
 
@@ -14,7 +14,7 @@ def test_mem_collector():
     collector.sample_memstats()
 
     m_a = torch.randn(10).cuda()
-    ModelDataTracer().add_tensor(m_a)
+    GLOBAL_MODEL_DATA_TRACER.add_tensor(m_a)
     b = torch.randn(10).cuda()
 
     # sampling at time 1
@@ -26,7 +26,7 @@ def test_mem_collector():
     collector.sample_memstats()
 
     collector.finish_collection()
-    collector.reset()
+    collector.reset_sampling_cnter()
 
     # do nothing after collection, just advance sampling cnter
     collector.sample_memstats()

colossalai/zero/init_ctx/init_context.py

@@ -3,7 +3,7 @@ import functools
 import torch
 from colossalai.zero.shard_utils import BaseShardStrategy
 from colossalai.zero.sharded_param import ShardedParamV2
-from colossalai.utils.memory_tracer.model_data_memtracer import ModelDataTracer
+from colossalai.utils.memory_tracer.model_data_memtracer import GLOBAL_MODEL_DATA_TRACER
 
 # Inserts _post_init_method at the end of init method
 
@@ -153,7 +153,7 @@ class ZeroInitContext(InsertPostInitMethodToModuleSubClasses):
 
             if self.shard_param:
                 self.shard_strategy.shard(tensor_list=[param.col_attr._data_sharded_tensor])
-                ModelDataTracer().add_tensor(param.col_attr._data_sharded_tensor.payload)
+                GLOBAL_MODEL_DATA_TRACER.add_tensor(param.col_attr._data_sharded_tensor.payload)
             if param.col_attr.grad and self.shard_grad:
                 self.shard_strategy.shard(tensor_list=[param.col_attr._grad_sharded_tensor])
-                ModelDataTracer().add_tensor(param.col_attr._grad_sharded_tensor.payload)
+                GLOBAL_MODEL_DATA_TRACER.add_tensor(param.col_attr._grad_sharded_tensor.payload)

tests/test_amp/test_naive_fp16.py

@@ -26,15 +26,15 @@ def run_naive_amp():
     test_models = ['repeated_computed_layers', 'nested_model']
     for test_name in test_models:
         get_component_func = non_distributed_component_funcs.get_callable(test_name)
-        model_builder, train_dataloader, _, optim_builder, _ = get_component_func()
+        model_builder, train_dataloader, _, optim_class, _ = get_component_func()
 
         # create model
         amp_model = model_builder(checkpoint=True).cuda()
         torch_model = copy.deepcopy(amp_model)
 
         # create optimizer
-        amp_optimizer = optim_builder(amp_model)
+        amp_optimizer = optim_class(amp_model.parameters(), lr=1e-3)
-        torch_optimizer = optim_builder(torch_model)
+        torch_optimizer = optim_class(torch_model.parameters(), lr=1e-3)
 
         # inject naive amp
         amp_config = dict(initial_scale=1)

tests/test_zero_data_parallel/test_init_context.py

@@ -14,7 +14,7 @@ from colossalai.zero.shard_utils import (BucketTensorShardStrategy, TensorShardS
 from tests.components_to_test.registry import non_distributed_component_funcs
 
 from common import CONFIG
-from colossalai.utils.memory_tracer.model_data_memtracer import ModelDataTracer
+from colossalai.utils.memory_tracer.model_data_memtracer import GLOBAL_MODEL_DATA_TRACER
 
 
 def run_dist(rank, world_size, port, init_device, shard_strategy):
@@ -37,10 +37,10 @@ def run_dist(rank, world_size, port, init_device, shard_strategy):
             assert param.col_attr.data.payload.device.type == init_device.type, \
                 f'{param.col_attr.data.payload.device.type} vs. {init_device.type}'
 
-    print(f'cuda usgae {ModelDataTracer().cuda_usage}')
+    print(f'cuda usgae {GLOBAL_MODEL_DATA_TRACER.cuda_usage}')
     print(f'numel {model_numel_tensor}')
     if init_device.type == 'cuda':
-        assert (ModelDataTracer().cuda_usage > 0)
+        assert (GLOBAL_MODEL_DATA_TRACER.cuda_usage > 0)
 
 
 @pytest.mark.dist