运行Ernie,报错Windows not support stack backtrace yet.和paddle.fluid.core.EnforceNotMet: Invoke operator fill_constant error.
Created by: Walleclipse
有关我的配置 1)PaddlePaddle版本:1.3.2 2)GPU:Tesla M40, CUDA: 8.0, cuDNN Version: 7.0 3)系统环境:Windows Server 2016, Python 3.6 我的模型: ERNIE 当我按照 Readme中的提取embedding的方法,运行 ernir_encoder.py时到 exe.run(startup_prog) 会出现错误,具体错误信息如下:
File "C:\Users\Python\Python36\site-packages\paddle\fluid\executor.py", line 591, in _run
exe.run(program.desc, scope, 0, True, True, fetch_var_name)
paddle.fluid.core.EnforceNotMet: Invoke operator fill_constant error.
Python Callstacks:
File "C:\Users\Python\Python36\site-packages\paddle\fluid\framework.py", line 1382, in _prepend_op
attrs=kwargs.get("attrs", None))
File "C:\Users\Python\Python36\site-packages\paddle\fluid\initializer.py", line 167, in __call__
stop_gradient=True)
File "C:\Users\Python\Python36\site-packages\paddle\fluid\framework.py", line 1300, in create_parameter
initializer(param, self)
File "C:\Users\Python\Python36\site-packages\paddle\fluid\layer_helper.py", line 339, in create_parameter
**attr._to_kwargs(with_initializer=True))
File "C:\Users\Python\Python36\site-packages\paddle\fluid\layer_helper.py", line 424, in append_bias_op
attr=bias_attr, shape=size, dtype=input_var.dtype, is_bias=True)
File "C:\Users\Python\Python36\site-packages\paddle\fluid\layers\nn.py", line 297, in fc
pre_activation = helper.append_bias_op(pre_bias, dim_start=num_flatten_dims)
File "D:\pycharm\course\IR_chat\LARK-develop\LARK-develop\ERNIE\model\ernie.py", line 155, in get_pooled_output
bias_attr="pooled_fc.b_0")
File "D:/pycharm/course/IR_chat/LARK-develop/LARK-develop/ERNIE/ernie_encoder.py", line 79, in create_model
cls_feats = ernie.get_pooled_output()
File "D:/pycharm/course/IR_chat/LARK-develop/LARK-develop/ERNIE/ernie_encoder.py", line 131, in main
args, pyreader_name='reader', ernie_config=ernie_config)
File "D:/pycharm/course/IR_chat/LARK-develop/LARK-develop/ERNIE/ernie_encoder.py", line 185, in <module>
main(args)
C++ Callstacks:
Enforce failed. Expected allocating <= available, but received allocating:11440875110 > available:11143610112.
Insufficient GPU memory to allocation. at [D:\1.3.2\paddle\paddle\fluid\platform\gpu_info.cc:216]
PaddlePaddle Call Stacks:
Windows not support stack backtrace yet.
W0411 20:25:15.478150 40984 device_context.cc:263] Please NOTE: device: 0, CUDA Capability: 52, Driver API Version: 9.0, Runtime API Version: 8.0
W0411 20:25:15.478150 40984 device_context.cc:271] device: 0, cuDNN Version: 7.0.
W0411 20:25:15.478150 40984 device_context.cc:295] WARNING: device: 0. The installed Paddle is compiled with CUDNN 7.1, but CUDNN version in your machine is 7.0, which may cause serious incompatible bug. Please recompile or reinstall Paddle with compatible CUDNN version.
并且 我按照Readme中的说明运行其他程序时也会报错。比如 run_classifier.py ,会报下面的错误
File "run_classifier.py", line 178, in main
main_program=train_program)
File "C:\Users\AppData\Roaming\Python\Python36\site-packages\paddle\fluid\parallel_executor.py", line 191, in __init__
local_scopes, exec_strategy, build_strategy)
paddle.fluid.core.EnforceNotMet: Not compiled with CUDA at [D:\1.3.2\paddle\paddle\fluid\framework\parallel_executor.cc:250]
PaddlePaddle Call Stacks:
Windows not support stack backtrace yet.
这个问题该怎么解决呢?谢谢