报错
tensorflow.python.framework.errors_impl.AlreadyExistsError:
Resource __per_step_8/BiLSTM/bidirectional_rnn/fw/fw/while/fw/coupled_input_forget_gate_lstm_cell/ArithmeticOptimizer/AddOpsRewrite_Leaf_1_add_2/tmp_var/N10tensorflow19TemporaryVariableOp6TmpVarE
[[]]
[[]]
问题描述：My problem was also caused by the model size, disregard the “calling tf.nn.dynamic_rnn in the same scope twice” in my previous comment.
问题解决
config = tf.compat.v1.ConfigProto()
off = rewriter_config_pb2.RewriterConfig.OFF
config.graph_options.rewrite_options.memory_optimization = off
