[legacy] clean up legacy code (#4743)

* [legacy] remove outdated codes of pipeline (#4692)

* [legacy] remove cli of benchmark and update optim (#4690)

* [legacy] remove cli of benchmark and update optim

* [doc] fix cli doc test

* [legacy] fix engine clip grad norm

* [legacy] remove outdated colo tensor (#4694)

* [legacy] remove outdated colo tensor

* [test] fix test import

* [legacy] move outdated zero to legacy (#4696)

* [legacy] clean up utils (#4700)

* [legacy] clean up utils

* [example] update examples

* [legacy] clean up amp

* [legacy] fix amp module

* [legacy] clean up gpc (#4742)

* [legacy] clean up context

* [legacy] clean core, constants and global vars

* [legacy] refactor initialize

* [example] fix examples ci

* [example] fix examples ci

* [legacy] fix tests

* [example] fix gpt example

* [example] fix examples ci

* [devops] fix ci installation

* [example] fix examples ci

colossalai/zero/gemini/memory_tracer/__init__.py

@@ -3,9 +3,8 @@ from .memory_stats import MemStats    # isort:skip
 from .memory_monitor import AsyncMemoryMonitor, SyncCudaMemoryMonitor    # isort:skip
 from .memstats_collector import MemStatsCollector    # isort:skip
 from .chunk_memstats_collector import ChunkMemStatsCollector    # isort:skip
-from .static_memstats_collector import StaticMemStatsCollector    # isort:skip
 
 __all__ = [
-    'AsyncMemoryMonitor', 'SyncCudaMemoryMonitor', 'MemStatsCollector', 'ChunkMemStatsCollector',
-    'StaticMemStatsCollector', 'MemStats', 'OrderedParamGenerator'
+    'AsyncMemoryMonitor', 'SyncCudaMemoryMonitor', 'MemStatsCollector', 'ChunkMemStatsCollector', 'MemStats',
+    'OrderedParamGenerator'
 ]

colossalai/zero/gemini/memory_tracer/chunk_memstats_collector.py

@@ -1,7 +1,6 @@
 from typing import Optional
 
 from colossalai.utils import get_current_device
-from colossalai.utils.memory import colo_device_memory_capacity
 from colossalai.zero.gemini.chunk import ChunkManager
 
 from .memory_stats import MemStats
@@ -33,4 +32,5 @@ class ChunkMemStatsCollector(MemStatsCollector):
 
     @property
     def cuda_margin_mem(self) -> float:
+        from colossalai.legacy.utils.memory import colo_device_memory_capacity
         return colo_device_memory_capacity(get_current_device()) - self._memstats.max_overall_cuda

colossalai/zero/gemini/memory_tracer/memory_monitor.py

@@ -5,7 +5,7 @@ from time import sleep, time
 
 import torch
 
-from colossalai.utils import colo_device_memory_used, get_current_device
+from colossalai.utils import get_current_device
 
 
 class MemoryMonitor:
@@ -110,6 +110,7 @@ class AsyncMemoryMonitor(MemoryMonitor):
         return max_usage
 
     def _measure_usage(self):
+        from colossalai.legacy.utils import colo_device_memory_used
         max_usage = 0
         while self.keep_measuring:
             max_usage = max(

colossalai/zero/gemini/memory_tracer/memstats_collector.py

@@ -70,7 +70,7 @@ class MemStatsCollector:
         Sampling model data statistics.
         """
         if self._start_flag and not self.use_outside_memstats:
-            from colossalai.zero.legacy.gemini import StatefulTensor
+            from colossalai.legacy.zero.gemini import StatefulTensor
 
             # The following code work for ZeroInitContext, which is deprecated in v0.1.12
             cuda_mem = StatefulTensor.GST_MGR.total_mem['cuda']

colossalai/zero/gemini/memory_tracer/runtime_mem_tracer.py

@@ -1,12 +1,12 @@
 import torch.nn
 
-from colossalai.tensor.param_op_hook import ColoParamOpHookManager
-from colossalai.utils import _cast_float
-from colossalai.zero.legacy.gemini.ophooks.runtime_mem_tracer_hook import (
+from colossalai.legacy.zero.gemini.ophooks.runtime_mem_tracer_hook import (
     GradMemStats,
     GradMemTracerHook,
     ParamMemTracerHook,
 )
+from colossalai.tensor.param_op_hook import ColoParamOpHookManager
+from colossalai.utils import _cast_float
 
 from .memory_stats import MemStats