FAILED TO LOAD CHECKPOINT

[wangjinzhuo]

When running python ram.py, it always says "fail to load checkpoint"

[hope-yao]

I also experienced the similar issue and I've attached the error message bellow.

W tensorflow/core/framework/op_kernel.cc:968] Not found: Tensor name "rnn_decoder/loop_function_4/Variable_2/Adam" not found in checkpoint files save-3scales/save10000.ckpt [[Node: save/restore_slice_89 = RestoreSlice[dt=DT_FLOAT, preferred_shard=-1, _device="/job:localhost/replica:0/task:0/cpu:0"](_recv_save/Const_0, save/restore_slice_89/tensor_name, save/restore_slice_89/shape_and_slice)]] FAILED TO LOAD CHECKPOINT

In the begining the error message I got is:

     "This module is deprecated.  Use tf.nn.seq2seq instead."

Then I changed

      _from tensorflow.models.rnn import seq2seq
      from tensorflow.models.rnn import rnn_cell

to

      from tensorflow.python.ops import seq2seq
      from tensorflow.python.ops import rnn_cell

respectively, that is when I got the error "fail to load checkpoint"

I've tried r0.9 r0.10 and cutting edge version of tensorflow, but none of them work. I'm wondering should I try the earliest version since this code is written in Dec 2015. But early version don't have good support over latest cuda, which my theano is using.... Have you got a good solution to it?

Thanks.