Fix `sharding_pass` and "nop" op to improve GC strategy (#56283)

* Improve GC for pipeline parallel

* Delete print

* fix bug of nop_op and sharding

---------
Co-authored-by: Nchenruibiao <chenruibiao@baidu.com>

paddle/fluid/operators/nop_op.cc

@@ -45,14 +45,9 @@ establish the dependency between input and output tensors.
   }
 };
 
-DECLARE_NO_NEED_BUFFER_VARS_INFERER(NopNoNeedBufferVarsInferer, "X", "Out");
-
 }  // namespace operators
 }  // namespace paddle
 
 namespace ops = paddle::operators;
 
-REGISTER_OP_WITHOUT_GRADIENT(nop,
-                             ops::NopOp,
-                             ops::NopOpMaker,
-                             ops::NopNoNeedBufferVarsInferer);
+REGISTER_OP_WITHOUT_GRADIENT(nop, ops::NopOp, ops::NopOpMaker);

python/paddle/distributed/passes/auto_parallel_sharding.py

@@ -295,12 +295,12 @@ class ShardingPass(PassBase):
         self._insert_optimizer_broadcasts(main_block, startup_block)
 
     def _shard_amp_related_op_and_vars(self, main_block):
-        if self.stage < 2:
+        if self.stage < 1:
             return
 
         for idx, op in reversed(list(enumerate(main_block.ops))):
             # shard amp related param_grad cast
-            if _is_param_grad_fp32_cast_op(main_block, op):
+            if _is_param_grad_fp32_cast_op(main_block, op) and self.stage > 1:
                 output_name = op.output_arg_names[0]
                 param_name = output_name[: output_name.find("@")]
                 if not self._is_parameter_in_local_shard(param_name):