Created by: mapingshuo

Hi, 我在使用encode.py来计算一句话的embedding, 正如Readme介绍的那样：https://github.com/PaddlePaddle/ERNIE#faq1-how-to-get-sentencetokens-embedding-of-ernie

我的encode脚本为：

TASK_DATA_PATH=.
MODEL_PATH=/path/to/model
export FLAGS_sync_nccl_allreduce=1
export CUDA_VISIBLE_DEVICES=4

python -u ernie_encoder.py \
                   --use_cuda true \
                   --batch_size 3 \
                   --output_dir "./test" \
                   --init_pretraining_params ${MODEL_PATH}/trained_chinese/params \
                   --data_set ${TASK_DATA_PATH}/baidu_input/dev.tsv \
                   --vocab_path ${MODEL_PATH}/vocab.txt \
                   --max_seq_len 128 \
                   --ernie_config_path ${MODEL_PATH}/ernie_config.json

其中，baidu_input/dev.tsv 如下：

label\ttext_a\n
0\t你吃了么\n
1\t我吃过了\n
0\t谢谢你啊\n

模型报错：

Device count: 1
Total num examples: 3
WARNING:root:paddle.fluid.layers.py_reader() may be deprecated in the near future. Please use paddle.fluid.io.PyReader() instead.
Traceback (most recent call last):
  File "ernie_encoder.py", line 182, in <module>
    main(args)
  File "ernie_encoder.py", line 130, in main
    args, pyreader_name='reader', ernie_config=ernie_config)
  File "ernie_encoder.py", line 77, in create_model
    unpad_enc_out = fluid.layers.sequence_unpad(enc_out, length=seq_lens)
  File "/home/mapingshuo/paddle_release_home/python-distribute/lib/python2.7/site-packages/paddle/fluid/layers/nn.py", line 4842, in sequence_unpad
    outputs={'Out': out})
  File "/home/mapingshuo/paddle_release_home/python-distribute/lib/python2.7/site-packages/paddle/fluid/layer_helper.py", line 43, in append_op
    return self.main_program.current_block().append_op(*args, **kwargs)
  File "/home/mapingshuo/paddle_release_home/python-distribute/lib/python2.7/site-packages/paddle/fluid/framework.py", line 2116, in append_op
    attrs=kwargs.get("attrs", None))
  File "/home/mapingshuo/paddle_release_home/python-distribute/lib/python2.7/site-packages/paddle/fluid/framework.py", line 1499, in __init__
    self.desc.infer_shape(self.block.desc)
paddle.fluid.core_avx.EnforceNotMet:

--------------------------------------------
C++ Call Stacks (More useful to developers):
--------------------------------------------
0   std::string paddle::platform::GetTraceBackString<std::string const&>(std::string const&, char const*, int)
1   paddle::platform::EnforceNotMet::EnforceNotMet(std::string const&, char const*, int)
2   paddle::operators::SequenceUnpadOp::InferShape(paddle::framework::InferShapeContext*) const
3   paddle::framework::OpDesc::InferShape(paddle::framework::BlockDesc const&) const

------------------------------------------
Python Call Stacks (More useful to users):
------------------------------------------
  File "/home/mapingshuo/paddle_release_home/python-distribute/lib/python2.7/site-packages/paddle/fluid/framework.py", line 2116, in append_op
    attrs=kwargs.get("attrs", None))
  File "/home/mapingshuo/paddle_release_home/python-distribute/lib/python2.7/site-packages/paddle/fluid/layer_helper.py", line 43, in append_op
    return self.main_program.current_block().append_op(*args, **kwargs)
  File "/home/mapingshuo/paddle_release_home/python-distribute/lib/python2.7/site-packages/paddle/fluid/layers/nn.py", line 4842, in sequence_unpad
    outputs={'Out': out})
  File "ernie_encoder.py", line 77, in create_model
    unpad_enc_out = fluid.layers.sequence_unpad(enc_out, length=seq_lens)
  File "ernie_encoder.py", line 130, in main
    args, pyreader_name='reader', ernie_config=ernie_config)
  File "ernie_encoder.py", line 182, in <module>
    main(args)

----------------------
Error Message Summary:
----------------------
PaddleCheckError: Expected len_dims.size() == 1, but received len_dims.size():2 != 1:1.
The shape of Input(Length) should be [batch_size]. at [/home/mapingshuo/Paddle/paddle/fluid/operators/sequence_ops/sequence_unpad_op.cc:41]
  [operator < sequence_unpad > error]

请问是什么原因，谢谢