[zero] dump memory stats for sharded model (#548)

colossalai/utils/memory_tracer/memstats_collector.py

@@ -45,10 +45,18 @@ class MemStatsCollector:
     def overall_cuda(self):
         return self._overall_cuda
 
+    @property
+    def model_data_cuda_GB(self):
+        return [elem / 1e9 for elem in self._model_data_cuda]
+
     @property
     def model_data_cuda(self):
         return self._model_data_cuda
 
+    @property
+    def non_model_data_cuda_GB(self):
+        return [elem / 1e9 for elem in self.non_model_data_cuda]
+
     @property
     def non_model_data_cuda(self):
         """Non model data stats

colossalai/zero/sharded_model/sharded_model_v2.py

@@ -11,6 +11,9 @@ from colossalai.engine.ophooks import register_ophooks_recursively
 from colossalai.engine.ophooks.zero_hook import ZeroHook
 from colossalai.engine.paramhooks import BaseParamHookMgr
 from colossalai.logging import get_dist_logger
+from colossalai.utils import get_current_device
+from colossalai.utils.memory_tracer.model_data_memtracer import GLOBAL_MODEL_DATA_TRACER
+from colossalai.utils.memory_utils.utils import colo_cuda_memory_capacity
 from colossalai.utils.memory_tracer.memstats_collector import MemStatsCollector
 from colossalai.utils.memory_tracer.model_data_memtracer import \
     GLOBAL_MODEL_DATA_TRACER
@@ -131,6 +134,29 @@ class ShardedModelV2(nn.Module):
     def cpu_offload(self):
         return self._cpu_offload
 
+    def dump_memory_stats(self, filename: Optional[str] = 'dump_mem_stats.log') -> None:
+        """
+        dummy memory tracer collected infomation to a file.
+        try:
+            # forward: model(inputs)
+            # backward: optimizer.backward()
+        except Exception as e:
+            model.dump_memory_stats()
+            exit(0)
+        """
+        if self._use_memory_tracer:
+            self.logger.error(f'dump memort tracer collected infomation to a {filename}', ranks=[0])
+            if gpc.get_global_rank() == 0:
+                with open(filename, 'w+') as f:
+                    f.write(f'cuda reserved {torch.cuda.memory_reserved(get_current_device())/1e9} GB\n')
+                    f.write(f'cuda max allocated {torch.cuda.max_memory_allocated(get_current_device())/1e9} GB\n')
+                    f.write('model data\n')
+                    f.write(str(self._memstats_collector.model_data_cuda_GB))
+                    f.write('\n')
+                    f.write('non model data\n')
+                    f.write(str(self._memstats_collector.non_model_data_cuda_GB))
+                    f.write('\n')
+
     def forward(self, *args: Any, **kwargs: Any) -> torch.Tensor:
         if self._iter_cnter == 0 and self._memstats_collector:
             # the opeartion will affect the flag in ZeroHook
@@ -147,6 +173,7 @@ class ShardedModelV2(nn.Module):
 
     def backward_by_grad(self, tensor, grad):
         torch.autograd.backward(tensors=tensor, grad_tensors=grad)
+
         self._post_backward_operations()
         for ophook in self._ophook_list:
             ophook.post_iter()
@@ -154,9 +181,6 @@ class ShardedModelV2(nn.Module):
     def _update_memstats(self):
         if self._iter_cnter == 0 and self._memstats_collector:
             self._memstats_collector.finish_collection()
-            self.logger.debug(f'model data cuda, {self._memstats_collector.model_data_cuda}')
-            self.logger.debug(f'non-model data cuda, {self._memstats_collector.non_model_data_cuda}')
-
         if self._memstats_collector:
             self._memstats_collector.reset_sampling_cnter()
             # cuda margin space = cuda mem capacity - max fwd/bwd cuda mem used.

colossalai/zero/sharded_optim/sharded_optim_v2.py

@@ -24,7 +24,6 @@ from colossalai.utils.memory_utils.utils import colo_model_data_tensor_move, col
 from colossalai.utils.memory_tracer.model_data_memtracer import GLOBAL_MODEL_DATA_TRACER
 
 
-
 class OptimState(Enum):
     SCALED = 1
     UNSCALED = 2
@@ -139,6 +138,10 @@ class ShardedOptimizerV2(ColossalaiOptimizer):
         if self._use_memory_tracer:
             GLOBAL_MODEL_DATA_TRACER.register_optimizer(self)
 
+        self._use_memory_tracer = self.model.use_memory_tracer
+        if self._use_memory_tracer:
+            GLOBAL_MODEL_DATA_TRACER.register_optimizer(self)
+
     def get_memory_usage(self) -> Tuple[int, int]:
         """
         Get the memory usage of the optimizer. Including master_params (param fp32),
@@ -190,13 +193,15 @@ class ShardedOptimizerV2(ColossalaiOptimizer):
                 # Now p.data is sharded
                 # So optimizer states are sharded naturally
 
-        self._logger.debug(f"Before step ShardedOptimizerV2 consumes {self.get_memory_usage()[0]/1e6} MB CUDA Memory!",
-                           ranks=[0])
+        self._logger.debug(
+            f"Before step ShardedOptimizerV2 consumes {self.get_memory_usage()[0]/1e6} MB CUDA Memory, {self.get_memory_usage()[1]/1e6} MB CUDA Memory!",
+            ranks=[0])
 
         ret = self.optim.step(*args, **kwargs)
 
-        self._logger.debug(f"After step ShardedOptimizerV2 consumes {self.get_memory_usage()[0]/1e6} MB CUDA Memory!",
-                           ranks=[0])
+        self._logger.debug(
+            f"After step ShardedOptimizerV2 consumes {self.get_memory_usage()[0]/1e6} MB CUDA Memory, {self.get_memory_usage()[1]/1e6} MB CUDA Memory!",
+            ranks=[0])
         # Copy master param data (fp32) to payload of col_attr (fp16)
         # TODO() improve efficiency by gathering tensors into a chunk and transfering
         # a chunk.