Virtual envrironment install questions

Help
1

I’m trying to install the latest version (2.5.8) on my FreeBSD and I have some questions on it, it seems to have some other issues since I last installed it.

scripts/build/build-phonetisaurus.sh has

"--with-openfst-includes=-I${openfst_build}/build/include" \
"--with-openfst-libs=-L${openfst_build}/build/lib" && \

I don’t think you need the -I and -L there, at least freeBSD doesn’t build with them, I also get a warning that you need to build FST with --enable-static --enable-shared --enable-ngram-fsts , the script only builds it with --enable-far

there’s output_file="$(realpath “$4”)" in the build scripts, freeBSDs realpath doesn’t work on files that do not exist, which is the case of the output files. (I guess linux will have the same issue)

I see eigen3 is built for kenml, but I get a warning it it not found when building kenml itself, should it be build with it?

Both phonitisaurus and kenml have fixed freeBSD bugs in the provided versions you host, it would be nice to update them to a later version (kenml > 8 march (https://github.com/kpu/kenlm/pull/263) phonetisaurus > 13 may (https://github.com/AdolfVonKleist/Phonetisaurus/pull/54), Kaldi > 29 jan (https://github.com/kaldi-asr/kaldi/pull/3881) (didn’t try Kaldi yet, might be up to date)

When installing the makefile has this entry for Larynx, requiring nanoTTS, is this true, or maybe a copy paste from the nanoTTS install? Also is the provided torch-1.6.0-cp37-cp37m-linux_x86_64 special or can I just use the pip version 1.7 from the pytorch project? (If I can get that to work on bsd)

# -----------------------------------------------------------------------------
# Larynx: text to speech
# https://github.com/rhasspy/larynx
# -----------------------------------------------------------------------------

.PHONY: larynx
ifeq ($(ENABLE_LARYNX),yes)
larynx: $(NANOTTS_PRE_FILE)

install-larynx: $(TORCH_WHEEL)
        "$(srcdir)/scripts/install/install-nanotts.sh" "$(NANOTTS_PRE_FILE)" "$(DESTDIR)$(prefix)"

$(TORCH_WHEEL):
        $(DO_DOWNLOAD) "$(TORCH_WHEEL)" "https://github.com/synesthesiam/prebuilt-apps/releases/download/v1.0/torch-1.6.0-cp37-cp37m-linux_$(LARYNX_ARCH).whl"
else # ENABLE_LARYNX
larynx:

install-larynx:
endif # ENABLE_LARYNX

A freeBSD only issue is that the install command used in the install scripts has a different syntax in freeBSD (it doesn’t have the --target-directory , the target directory should be listed as last parameter instead) Mac OSX might do the same.

2

Hey,
I was wondering if there are any instructions on how to install Rhasspy on FreeBSD?
Thank you

3

Rhasspy isn’t officially tested or supported on FreeBSD, but the continuous integration tests run on both Ubuntu and macOS, so most of the project should work on other Unix-like platforms too. That said, you’ll probably bump into issues like the one above. Your best bet would be to just try to follow the instructions for the installation in a Python virtual environment, after installing the dependencies yourself.

If you encounter some problems, feel free to open a topic on the forum or open an issue. Maybe we can help, and document the process later for other FreeBSD users.

4

Thanks for the reply. I dont think I m at that level but I might give it a go and see how that goes

5

I sort of documented my first try in my previous post about it last year Install on FreeBSD. It is not easy and needs some adjusting of the install/compile scripts. Last time I got most if it to work with the previous posts, I didn’t get larynx working yet, so I didn’t have a working TTS, but speech recognizing was working with Kaldi. I’ll have to try again some of these days with the current version.