【PaddlePaddle系列】报错解决方法合集 （不定时更新）

1、PaddlePaddle使用CPU时正常运行，但是使用GPU时却报出一堆错误信息，节选如下：

paddle.fluid.core.EnforceNotMet: enforce allocating <= available failed,  >
 at [/paddle/paddle/fluid/platform/gpu_info.cc:]
PaddlePaddle Call Stacks:
       0x7f89b8241736p paddle::platform::EnforceNotMet::EnforceNotMet(std::__exception_ptr::exception_ptr, char const*, int) +
       0x7f89b91f0afep paddle::platform::GpuMaxChunkSize() +
       0x7f89b9120aadp paddle::memory::GetGPUBuddyAllocator(int) +
       0x7f89b9120cbcp void* paddle::memory::Alloc<paddle::platform::CUDAPlace>(paddle::platform::CUDAPlace, unsigned long) + 

解决方法：

export FLAGS_fraction_of_gpu_memory_to_use=

避免每次运行或者每个终端都添加一次，可将此按个人喜好添加到用户级~/.bashrc或系统级/etc/profile

2、PaddlePaddle使用Fluid版本，使用exe.run时候报错：

---------------------------------------------------------------------------
EnforceNotMet                             Traceback (most recent call last)
<ipython-input--ca8c92bb26a4> in <module>()
              loss = exe.run(fluid.default_main_program(),
                                   feed=feeder.feed(data),
--->                               fetch_list=[avg_cost])
              print("Pass {0},Loss {1}".format(pass_id,loss))

/home/dzqiu/anaconda2/lib/python2./site-packages/paddle/fluid/executor.pyc in run(self, program, feed, fetch_list, feed_var_name, fetch_var_name, scope, return_numpy, use_program_cache)

             self._feed_data(program, feed, feed_var_name, scope)
-->          self.executor.run(program.desc, scope, , True, True)
             outs = self._fetch_data(fetch_list, fetch_var_name, scope)
             if return_numpy:

EnforceNotMet: enforce y_dims.size() > y_num_col_dims failed,  <=
The input tensor Y's rank of MulOp should be larger than y_num_col_dims. at [/paddle/paddle/fluid/operators/mul_op.cc:52]
PaddlePaddle Call Stacks:
       0x7f3db10d7736p paddle::platform::EnforceNotMet::EnforceNotMet(std::__exception_ptr::exception_ptr, char const*, int) +
       0x7f3db16da696p paddle::operators::MulOp::InferShape(paddle::framework::InferShapeContext*) const +
       0x7f3db1f0ef7bp paddle::framework::OperatorWithKernel::RunImpl(paddle::framework::Scope const&, boost::variant<paddle::platform::CUDAPlace, paddle::platform::CPUPlace, paddle::platform::CUDAPinnedPlace, boost::detail::variant::void_, boost::detail::variant::void_, boost::detail::variant::void_, boost::detail::variant::void_, boost::detail::variant::void_, boost::detail::variant::void_, boost::detail::variant::void_, boost::detail::variant::void_, boost::detail::variant::void_, boost::detail::variant::void_, boost::detail::variant::void_, boost::detail::variant::void_, boost::detail::variant::void_, boost::detail::variant::void_, boost::detail::variant::void_, boost::detail::variant::void_, boost::detail::variant::void_> const&) const +
       0x7f3db1f0c6edp paddle::framework::OperatorBase::Run(paddle::framework::Scope const&, boost::variant<paddle::platform::CUDAPlace, paddle::platform::CPUPlace, paddle::platform::CUDAPinnedPlace, boost::detail::variant::void_, boost::detail::variant::void_, boost::detail::variant::void_, boost::detail::variant::void_, boost::detail::variant::void_, boost::detail::variant::void_, boost::detail::variant::void_, boost::detail::variant::void_, boost::detail::variant::void_, boost::detail::variant::void_, boost::detail::variant::void_, boost::detail::variant::void_, boost::detail::variant::void_, boost::detail::variant::void_, boost::detail::variant::void_, boost::detail::variant::void_, boost::detail::variant::void_> const&) +
       0x7f3db11734afp paddle::framework::Executor::RunPreparedContext(paddle::framework::ExecutorPrepareContext*, paddle::framework::Scope*, bool, bool, bool) +
       0x7f3db1174500p paddle::framework::Executor::Run(paddle::framework::ProgramDesc const&, paddle::framework::Scope*, int, bool, bool) + 

　解决方法：

　在训练前先执行：exe.run(fluid.default_startup_program())