Created by: IrishCoffee

各位好，我在编译运行deepasr aishell这个例子的时候，先是运行sh profile.sh，会报一个memory不够的问题。

.........Traceback (most recent call last): File "../../tools/profile.py", line 210, in profile(args) File "../../tools/profile.py", line 193, in profile return_numpy=False) File "/usr/local/lib/python2.7/dist-packages/paddle/fluid/executor.py", line 470, in run self.executor.run(program.desc, scope, 0, True, True) paddle.fluid.core.EnforceNotMet: ptr_ should not be null Insufficient GPU memory to allocation. at [/paddle/paddle/fluid/framework/tensor.h:180] PaddlePaddle Call Stacks: 0 0x7fe12816db36p paddle::platform::EnforceNotMet::EnforceNotMet(std::__exception_ptr::exception_ptr, char const*, int) + 486

如果我直接运行sh train.sh, 会报如下错误。

xjia@nvidia:~/work/PaddlePaddle/models/fluid/DeepASR/examples/aishell$ sh train.sh
-----------  Configuration Arguments -----------
batch_size: 64
checkpoints: checkpoints
class_num: 3040
device: GPU
frame_dim: 80
hidden_dim: 1024
infer_models:
init_model_path: None
learning_rate: 6.4e-05
mean_var: data/global_mean_var
minimum_batch_size: 1
parallel: True
pass_num: 100
print_per_batches: 100
proj_dim: 512
stacked_num: 5
train_feature_lst: data/train_feature.lst
train_label_lst: data/train_label.lst
val_feature_lst: data/val_feature.lst
val_label_lst: data/val_label.lst
------------------------------------------------
API get_places is deprecated since 0.15.0. Please use ParallelExecutor instead.
Traceback (most recent call last):
  File "../../train.py", line 292, in <module>
    train(args)
  File "../../train.py", line 156, in train
    parallel=args.parallel)
  File "/home/xjia/work/PaddlePaddle/models/fluid/DeepASR/model_utils/model.py", line 91, in stacked_lstmp_model
    pd = fluid.layers.ParallelDo(places)
AttributeError: 'module' object has no attribute 'ParallelDo'

我是在一个单机1080ti的GPU上运行的，已经确认paddle成功安装。