[Gemini] rename ParamTracerWrapper -> RuntimeMemTracer (#2073)

colossalai/gemini/memory_tracer/runtime_mem_tracer.py

@@ -1,13 +1,14 @@
 import torch.nn
 
-from colossalai.tensor.param_op_hook import ParamOpHookManager
-from colossalai.gemini.ophooks.param_trace_hook import ParamTracerHook, GradHook
 from colossalai.gemini.memory_tracer.model_data_memtracer import GLOBAL_CUDA_MEM_INFO
+from colossalai.gemini.ophooks.param_trace_hook import GradHook, ParamTracerHook
 from colossalai.nn.parallel.data_parallel import _cast_float
+from colossalai.tensor.param_op_hook import ParamOpHookManager
 
-__all__ = ['ParamTracerWrapper']
+__all__ = ['RuntimeMemTracer']
 
-class ParamTracerWrapper():
+
+class RuntimeMemTracer():
 
     def __init__(self, module: torch.nn.Module, dtype: torch.dtype = torch.half):
         super().__init__()
@@ -25,12 +26,18 @@ class ParamTracerWrapper():
     def __call__(self, *args, **kwargs):
         return self.forward(*args, **kwargs)
 
-    def _save_param_data_on_cpu(self):
+    def _backup_params(self):
+        """
+        The function is called before forward. Backup model params on cpu.
+        """
         for p in self.module.parameters():
             self.cpu_param_data_dict[p] = torch.empty(p.data.shape, dtype=self.dtype, device="cpu")
             self.cpu_param_data_dict[p].copy_(p.data)
 
-    def _restore_param_data(self):
+    def _restore_params(self):
+        """
+        This function is called after backward. Restore model params.
+        """
         for p in self.module.parameters():
             p.data = torch.empty(p.data.shape, dtype=self.dtype, device="cpu", requires_grad=p.data.requires_grad)
             p.data.copy_(self.cpu_param_data_dict[p])
@@ -38,7 +45,7 @@ class ParamTracerWrapper():
 
     def _pre_forward(self):
         self._clear_cuda_mem_info()
-        self._save_param_data_on_cpu()
+        self._backup_params()
         self.grad_hook.register_grad_hook()
         self.param_op_hook.mem_monitor.start()
 
@@ -60,7 +67,7 @@ class ParamTracerWrapper():
         last_model_data = GLOBAL_CUDA_MEM_INFO.model_data_list[-1]
         GLOBAL_CUDA_MEM_INFO.non_model_data_list.append(cuda_volume - last_model_data)
         self.grad_hook.remove_grad_hook()
-        self._restore_param_data()
+        self._restore_params()
 
     def _clear_cuda_mem_info(self):
         GLOBAL_CUDA_MEM_INFO.model_data_list.clear()

tests/test_gemini/test_runtime_mem_tracer.py

@@ -1,11 +1,15 @@
+from copy import deepcopy
+
 import numpy as np
 import torch
 
-from colossalai.gemini.memory_tracer.param_tracer_wrapper import ParamTracerWrapper
 from colossalai.gemini.memory_tracer.model_data_memtracer import GLOBAL_CUDA_MEM_INFO
+from colossalai.gemini.memory_tracer.runtime_mem_tracer import RuntimeMemTracer
 from colossalai.utils.model.colo_init_context import ColoInitContext
+from tests.components_to_test import run_fwd_bwd
 from tests.components_to_test.registry import non_distributed_component_funcs
 
+
 def run_fwd_bwd(model, data, label, criterion, enable_autocast=False, dtype=torch.half):
     with torch.cuda.amp.autocast(enabled=enable_autocast):
         if criterion:
@@ -16,9 +20,9 @@ def run_fwd_bwd(model, data, label, criterion, enable_autocast=False, dtype=torc
         loss = loss.to(dtype)
     model.backward(loss)
 
+
 def run_param_wrapper_testing():
     test_models = ['simple_net', 'repeated_computed_layers', 'nested_model']
-
     for model_name in test_models:
         get_components_func = non_distributed_component_funcs.get_callable(model_name)
         model_builder, train_dataloader, _, _, criterion = get_components_func()
@@ -26,7 +30,8 @@ def run_param_wrapper_testing():
         with ColoInitContext(device=torch.device('cpu')):
             model = model_builder(checkpoint=False)
 
-        model = ParamTracerWrapper(model)
+        model_bk = deepcopy(model)
+        runtime_mem_tracer = RuntimeMemTracer(model)
 
         for i, (data, label) in enumerate(train_dataloader):
             if i > 1:
@@ -34,7 +39,10 @@ def run_param_wrapper_testing():
             data = data.cuda()
             label = label.cuda()
 
-            run_fwd_bwd(model, data, label, criterion, False)
+            run_fwd_bwd(runtime_mem_tracer, data, label, criterion, False)
+
+        for p1, p2 in zip(model_bk.parameters(), model.parameters()):
+            torch.allclose(p1.to(torch.half), p2)
 
         cuda_non_model_data_list = np.array(GLOBAL_CUDA_MEM_INFO.non_model_data_list) / 1024**2
         print("cuda_non_model_data_list", len(cuda_non_model_data_list))
@@ -43,6 +51,5 @@ def run_param_wrapper_testing():
         del model
 
 
-
 if __name__ == '__main__':
     run_param_wrapper_testing()