!921 Revert limitation of end_learning_rate

Merge pull request !921 from Simson/push-to-opensource

mindspore/nn/optim/lamb.py

@@ -114,7 +114,7 @@ def _check_param_value(decay_steps, warmup_steps, start_learning_rate,
     _ = warmup_steps
     validator.check_float_positive('start_learning_rate', start_learning_rate, prim_name)
     validator.check_float_legal_value('start_learning_rate', start_learning_rate, prim_name)
-    validator.check_float_positive('end_learning_rate', end_learning_rate, prim_name)
+    validator.check_value_type("end_learning_rate", end_learning_rate, [float], prim_name)
     validator.check_float_legal_value('end_learning_rate', end_learning_rate, prim_name)
     validator.check_float_positive('power', power, prim_name)
     validator.check_float_legal_value('power', power, prim_name)