From 81e06752e32e1990731a8fa5e7a25d1474b80d4a Mon Sep 17 00:00:00 2001
From: Charles-hit <56987902+Charles-hit@users.noreply.github.com>
Date: Tue, 13 Sep 2022 20:12:08 +0800
Subject: [PATCH] fix distributed bug caused by fill_any_like (#45978)

---
 python/paddle/distributed/auto_parallel/completion.py         | 4 ++--
 .../distributed/auto_parallel/operators/dist_default.py       | 4 ++--
 python/paddle/distributed/auto_parallel/utils.py              | 2 +-
 3 files changed, 5 insertions(+), 5 deletions(-)

diff --git a/python/paddle/distributed/auto_parallel/completion.py b/python/paddle/distributed/auto_parallel/completion.py
index 1775a823c5..8712288204 100644
--- a/python/paddle/distributed/auto_parallel/completion.py
+++ b/python/paddle/distributed/auto_parallel/completion.py
@@ -1037,7 +1037,7 @@ class Completer:
                     grad_op_dist_attr.set_output_dims_mapping(
                         output_name, ref_fwd_dims_mapping)
 
-                elif grad_op.type == 'fill_zeros_like':
+                elif grad_op.type == 'fill_any_like':
                     ref_var_name = grad_op.input_arg_names[0]
                     ref_var = vars[ref_var_name]
                     ref_dist_attr = self._dist_context.get_tensor_dist_attr_for_program(
@@ -1274,7 +1274,7 @@ class Completer:
                     grad_op_dist_attr.impl_type = "default"
                     grad_op_dist_attr.impl_idx = 0
 
-                elif grad_op.type == 'fill_zeros_like':
+                elif grad_op.type == 'fill_any_like':
                     ref_var_name = grad_op.input_arg_names[0]
                     ref_var = vars[ref_var_name]
                     ref_dist_attr = self._dist_context.get_tensor_dist_attr_for_program(
diff --git a/python/paddle/distributed/auto_parallel/operators/dist_default.py b/python/paddle/distributed/auto_parallel/operators/dist_default.py
index 08c81c4a30..a5139e0018 100644
--- a/python/paddle/distributed/auto_parallel/operators/dist_default.py
+++ b/python/paddle/distributed/auto_parallel/operators/dist_default.py
@@ -348,7 +348,7 @@ class DistributedDefaultImpl0(DistributedOperatorImpl):
             else:
                 batch_dim_mappings.append(dims_mapping[1])
         for arg_name in op_desc.output_arg_names():
-            if op_desc.type() == "fill_zeros_like":
+            if op_desc.type() == 'fill_any_like':
                 input_tensor = dist_op.get_serial_input(
                     op_desc.input_arg_names()[0])
                 if input_tensor.is_parameter:
@@ -387,7 +387,7 @@ class DistributedDefaultImpl0(DistributedOperatorImpl):
                     dims_mapping[1] = compatible_dim_mapping
                     changed = True
         for arg_name in op_desc.output_arg_names():
-            if op_desc.type() == "fill_zeros_like":
+            if op_desc.type() == 'fill_any_like':
                 input_tensor = dist_op.get_serial_input(
                     op_desc.input_arg_names()[0])
                 if input_tensor.is_parameter:
diff --git a/python/paddle/distributed/auto_parallel/utils.py b/python/paddle/distributed/auto_parallel/utils.py
index d276df6ddb..cfe2161286 100644
--- a/python/paddle/distributed/auto_parallel/utils.py
+++ b/python/paddle/distributed/auto_parallel/utils.py
@@ -1040,7 +1040,7 @@ def set_grad_var_shape(program, dist_context):
 
             if op.type in [
                     "c_allreduce_sum", "c_identity", "scale", "cast",
-                    "fill_zeros_like"
+                    'fill_any_like'
             ]:
                 forward_var_name = op.input_arg_names[0]
             elif op.type == "matmul_v2_grad" or op.type == "matmul_grad" or op.type == "mul_grad":
-- 
GitLab