From 5f542a1ddfea912a44e451bab046cee55330e086 Mon Sep 17 00:00:00 2001
From: fuchao <1501312275@qq.com>
Date: Fri, 23 May 2025 10:38:11 +0800
Subject: [PATCH] =?UTF-8?q?=E4=BC=98=E5=8C=96=E6=8F=92=E5=85=A5=E5=8F=8D?=
 =?UTF-8?q?=E5=90=91TensorDump=E7=AE=97=E5=AD=90=E9=80=BB=E8=BE=91?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

---
 .../mindspore/dump/cell_dump_process.py       | 25 +++++++++----------
 1 file changed, 12 insertions(+), 13 deletions(-)

diff --git a/debug/accuracy_tools/msprobe/mindspore/dump/cell_dump_process.py b/debug/accuracy_tools/msprobe/mindspore/dump/cell_dump_process.py
index 1f17dc491..f4e0bf58c 100644
--- a/debug/accuracy_tools/msprobe/mindspore/dump/cell_dump_process.py
+++ b/debug/accuracy_tools/msprobe/mindspore/dump/cell_dump_process.py
@@ -58,6 +58,12 @@ if (ms.__version__ >= "2.5.0"):
     td_in = ops.TensorDump("in")
 else:
     td_in = ops.TensorDump()
+dump_gradient_op_existed = False
+if hasattr(ops, 'DumpGradient'):
+    gd = ops.DumpGradient()
+    dump_gradient_op_existed = True
+else:
+    logger.warning('The operator "DumpGradient" does not exist. Cell dump can not work in graph mode.')
 td.add_prim_attr(KEY_SIDE_EFFECT, False)
 td_in.add_prim_attr(KEY_SIDE_EFFECT, False)
 td.add_prim_attr(KEY_TD_FLAG, True)
@@ -147,7 +153,8 @@ def cell_construct_wrapper(func, self):
         for index, item in enumerate(args):
             if self.data_mode == "backward" or self.data_mode == "all":
                 if ops.is_tensor(item):
-                    item = self.output_clips[index](item)
+                    item = gd(gen_file_path(self.dump_path, self.cell_prefix, KEY_BACKWARD, KEY_OUTPUT, index),
+                              item, "out")
             if self.data_mode == "forward" or self.data_mode == "all":
                 if ops.is_tensor(item):
                     if need_tensordump_in(self, 'input_dump_mode'):
@@ -170,7 +177,8 @@ def cell_construct_wrapper(func, self):
             for index, item in enumerate(out):
                 if self.data_mode == "backward" or self.data_mode == "all":
                     if ops.is_tensor(item):
-                        item = self.input_clips[index](item)
+                        item = gd(gen_file_path(self.dump_path, self.cell_prefix, KEY_BACKWARD, KEY_INPUT, index),
+                                  item, "in")
                 if self.data_mode == "forward" or self.data_mode == "all":
                     if ops.is_tensor(item):
                         if need_tensordump_in(self, 'output_dump_mode'):
@@ -191,7 +199,8 @@ def cell_construct_wrapper(func, self):
             return out_list
         else:
             if self.data_mode == "backward" or self.data_mode == "all":
-                out = self.input_clips[0](out)
+                out = gd(gen_file_path(self.dump_path, self.cell_prefix, KEY_BACKWARD, KEY_INPUT, 0),
+                             out, "in")
             if self.data_mode == "forward" or self.data_mode == "all":
                 if ops.is_tensor(out):
                     if need_tensordump_in(self, 'output_dump_mode'):
@@ -831,16 +840,6 @@ def start(config: CellDumpConfig):
         logger.info(f"Cell {name}: construct function is wrapped!")
         cell.dump_path = dump_path
         cell.data_mode = data_mode
-        cell.input_clips = []
-        cell.output_clips = []
-        # It is assumed that each cell has a maximum of 50 outputs and 50 inputs.
-        for i in range(50):
-            cell.input_clips.append(
-                ops.InsertGradientOf(partial_func(clip_gradient, cell.dump_path, cell.cell_prefix, i, KEY_INPUT))
-            )
-            cell.output_clips.append(
-                ops.InsertGradientOf(partial_func(clip_gradient, cell.dump_path, cell.cell_prefix, i, KEY_OUTPUT))
-            )
 
     logger.info(f"==========The cell_dump_process_start phase is Finished!==========")
     if dump_task == CoreConst.TENSOR:
-- 
Gitee