load error message of WarpCTC is confusing
Created by: reyoung
F0103 16:49:50.844861 2057601024 hl_dso_loader.cc:104] Check failed: nullptr != *dso_handle Failed to find dynamic library: /usr/local/cuda/lib/libwarpctc.dylib (dlopen(/usr/local/cuda/lib/libwarpctc.dylib, 5): image not found)
Please specify its path correctly using one of the following ways:
Method 1. set cuda and cudnn lib path at runtime. http://www.paddlepaddle.org/doc/ui/cmd_argument/argument_outline.html
For instance, issue command: paddle train --use_gpu=1 --cuda_dir=/usr/local/cuda/lib64 --cudnn_dir=/usr/local/cudnn/lib ...
Method 2. set environment variable LD_LIBRARY_PATH on Linux or DYLD_LIBRARY_PATH on Mac OS.
For instance, issue command: export LD_LIBRARY_PATH=...
Note: After Mac OS 10.11, using the DYLD_LIBRARY_PATH is impossible unless System Integrity Protection (SIP) is disabled. However, method 1 always work well.
It seems WarpCTC is not related to CUDA, but why the error message shows --cuda_dir
, --cudnn_dir
.
Also
However, method 1 always work well.
work => works.