W0607 15:11:35.219653 1625662 rnn_kernel.cu.cc:234] If the memory space of the Input WeightList is not continuous, less efficient calculation will be called. Please call flatten_parameters() to make the input memory continuous.
--------------------------------------
C++ Traceback (most recent call last):
--------------------------------------
No stack trace in paddle, may be caused by external reasons.
----------------------
Error Message Summary:
----------------------
FatalError: `Segmentation fault` is detected by the operating system.
[TimeInfo: *** Aborted at 1717744295 (unix time) try "date -d @1717744295" if you are using GNU date ***]
[SignalInfo: *** SIGSEGV (@0x0) received by PID 1625662 (TID 0x7f1aa1cb7740) from PID 0 ***]
print(
pd_model_state_dict['lstm.weight_ih_l0'] is pd_model_state_dict['lstm.0.cell.weight_ih'],
pd_model_state_dict['lstm.weight_hh_l0'] is pd_model_state_dict['lstm.0.cell.weight_hh'],
pd_model_state_dict['lstm.bias_ih_l0'] is pd_model_state_dict['lstm.0.cell.bias_ih'],
pd_model_state_dict['lstm.bias_hh_l0'] is pd_model_state_dict['lstm.0.cell.bias_hh']
)
非常感谢Paddle社区为开源做出的贡献!
环境
Python3.9 Paddle版本 2.6.0 cuda_12.2.r12.2/compiler.32965470_0
最小复现代码
以下是我写的 torch LSTM => Paddle LSTM 转换的函数脚本
这是测试代码,可以直接运行
如果不手动关闭 cudnn 运行,则会报错
报错内容:
我有俩个问题:
lstm_paddle.lstm.could_use_cudnn = False
这个flag,还是会报一样的段错误其他补充信息 Additional Supplementary Information
No response