I tried to switch in a new model (built with the latest Kaldi) and I hit this error
ERROR: FstImpl::ReadHeader: FST not of type vector: <unspecified>
ERROR (fcgi-nnet3-decoder[5.3.15~1-f14e]:ReadFstKaldi():kaldi-fst-io.cc:40) Could not read fst from HCLG.fst
[ Stack-Trace: ]
kaldi::MessageLogger::HandleMessage(kaldi::LogMessageEnvelope const&, char const*)
kaldi::MessageLogger::~MessageLogger()
fst::ReadFstKaldi(std::string)
apiai::Nnet3LatgenFasterDecoder::Initialize(kaldi::OptionsItf&)
apiai::FcgiDecodingApp::Run(int, char**)
main
I found this posting
https://groups.google.com/forum/#!msg/kaldi-help/0SuCkeHyUmU/qIW-INChAAAJ
in which Dan Povey suggested the fix
I believe ReadFstKaldi() should now be ReadFstKaldiGeneric(), to allow reading const_fst.
I made the change in Nnet3LatgenFasterDecoder.cc and this particular load error went away. But the recognition itself is trashed, which suggests to me that there is something else going on.
I tried to switch in a new model (built with the latest Kaldi) and I hit this error
I found this posting https://groups.google.com/forum/#!msg/kaldi-help/0SuCkeHyUmU/qIW-INChAAAJ in which Dan Povey suggested the fix
I believe ReadFstKaldi() should now be ReadFstKaldiGeneric(), to allow reading const_fst.
I made the change inNnet3LatgenFasterDecoder.cc
and this particular load error went away. But the recognition itself is trashed, which suggests to me that there is something else going on.