run recompute's real backward with amp disabled (#40042)

python/paddle/distributed/fleet/utils/recompute.py

@@ -182,9 +182,10 @@ class RecomputeFunction(PyLayer):
                     "none of output has requires_grad=True, this recompute() is not necessary"
                 )
 
-            # actually backward            
-            paddle.autograd.backward(forward_outputs_with_grad,
-                                     backward_inputs_with_grad)
+            # actually backward
+            with paddle.amp.auto_cast(enable=False):
+                paddle.autograd.backward(forward_outputs_with_grad,
+                                         backward_inputs_with_grad)
 
             grads = list(inp._grad_ivar() for inp in detached_inputs
                          if isinstance(inp, core.VarBase))