Closed hainan-xv closed 7 years ago
which branch?
shortcut
It cannot compile.
fixed. Try pull'ing the new version?
still failed....
did you make clean and re-./configure ? it might have to do with the CUDA update I feel
yes.
On Sat, Jan 21, 2017 at 12:07 PM, hainan-xv notifications@github.com wrote:
did you make clean and re-./configure ? it might have to do with the CUDA update I feel
— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/hainan-xv/kaldi/issues/2#issuecomment-274274158, or mute the thread https://github.com/notifications/unsubscribe-auth/ADWAklLv-u6FBuxjX0GKz00g3r_vVp8lks5rUjtRgaJpZM4Lpx1P .
-- Yiming Wang Department of Computer Science The Johns Hopkins University 3400 N. Charles St. Baltimore, MD 21218
Send me an email with the error messages then I will have a look
In file src/rnnlm/rnnlm-component.cc, functions
void LmLinearComponent::Propagate(const SparseMatrix<BaseFloat> &sp, CuMatrixBase<BaseFloat> *out) const
void LmLinearComponent::UpdateSimple(const SparseMatrix<BaseFloat> &in_value, const CuMatrixBase<BaseFloat> &out_deriv)
have very inefficient implementations. @freewym Can you implement them at the CuMatrix level and write cuda kernel function for them? Thanks.