paddle v2 dssm demo 训练报错

[gonghaie]

• 把dssm拉到测试机上之后，在dssm目录下运行以下命令正常输出model：
  python train.py --model_type 1 --model_arch 0 --model_output_prefix "model_dir/" --class_num 2
• 但是，将命令改为以下参数之后，训练报错：
  python train.py --model_type 1 --model_arch 2 --model_output_prefix "model_dir/" --class_num 2
  报错日志如下：

Traceback (most recent call last):
  File "train.py", line 267, in <module>
    use_gpu=args.use_gpu)
  File "train.py", line 186, in train
    share_embed=share_embed)()
  File "/home/yanchunwei/third_party/haie/dssm/network_conf.py", line 73, in __call__
    return self.model_type_creater()
  File "/home/yanchunwei/third_party/haie/dssm/network_conf.py", line 207, in _build_rank_model
    x = self.model_arch_creater(input, prefix=prefixs[id])
  File "/home/yanchunwei/third_party/haie/dssm/network_conf.py", line 57, in _model_arch_creater
    sent_vec = _model_arch.get(str(model_arch))(emb, prefix)
  File "/home/yanchunwei/third_party/haie/dssm/network_conf.py", line 106, in create_rnn
    input=emb)
  File "/home/yanchunwei/third_party/paddle_env/python27/lib/python2.7/site-packages/paddle/v2/config_base.py", line 52, in wrapped
    out = f(*args, **xargs)
  File "/home/yanchunwei/third_party/paddle_env/python27/lib/python2.7/site-packages/paddle/trainer_config_helpers/default_decorators.py", line 53, in __wrapper__
    return func(*args, **kwargs)
  File "/home/yanchunwei/third_party/paddle_env/python27/lib/python2.7/site-packages/paddle/trainer_config_helpers/default_decorators.py", line 53, in __wrapper__
    return func(*args, **kwargs)
  File "/home/yanchunwei/third_party/paddle_env/python27/lib/python2.7/site-packages/paddle/trainer_config_helpers/default_decorators.py", line 53, in __wrapper__
    return func(*args, **kwargs)
  File "/home/yanchunwei/third_party/paddle_env/python27/lib/python2.7/site-packages/paddle/trainer_config_helpers/default_decorators.py", line 53, in __wrapper__
    return func(*args, **kwargs)
  File "/home/yanchunwei/third_party/paddle_env/python27/lib/python2.7/site-packages/paddle/trainer_config_helpers/default_decorators.py", line 53, in __wrapper__
    return func(*args, **kwargs)
  File "/home/yanchunwei/third_party/paddle_env/python27/lib/python2.7/site-packages/paddle/trainer_config_helpers/layers.py", line 350, in wrapper
    return method(*args, **kwargs)
  File "/home/yanchunwei/third_party/paddle_env/python27/lib/python2.7/site-packages/paddle/trainer_config_helpers/layers.py", line 1345, in grumemory
    assert input.size is not None and input.size % 3 == 0
AssertionError

• 初步怀疑此分支没有进行单元测试。

[luotao1]

@NHZlX 和我复现 python train.py --model_type 1 --model_arch 2 --model_output_prefix "model_dir/" --class_num 2后出现的问题：

Traceback (most recent call last):
  File "train.py", line 266, in <module>
    use_gpu=args.use_gpu)
  File "train.py", line 186, in train
    share_embed=share_embed)()
  File "/home/luotao02/models/dssm/network_conf.py", line 73, in __call__
    return self.model_type_creater()
  File "/home/luotao02/models/dssm/network_conf.py", line 207, in _build_rank_model
    x = self.model_arch_creater(input, prefix=prefixs[id])
  File "/home/luotao02/models/dssm/network_conf.py", line 57, in _model_arch_creater
    sent_vec = _model_arch.get(str(model_arch))(emb, prefix)
  File "/home/luotao02/models/dssm/network_conf.py", line 105, in create_rnn
    gru = paddle.layer.gru_memory(
AttributeError: 'module' object has no attribute 'gru_memory'

[Superjomn]

请参考 https://github.com/PaddlePaddle/models/pull/212/files 的修改
@gonghaie