Open Franck-Dernoncourt opened 4 years ago
I think it should ok to use the alternative? I have not tried it, but can you tell me if that solves the problem?
I'm not sure, as to run source parse.sh
we don't need to install libhdf5-serial-dev
:
docker run --interactive --tty ubuntu:18.04 bash
apt update; apt install -y git nano wget htop python3 python3-pip unzip; git clone https://github.com/KhalilMrini/LAL-Parser; cd LAL-Parser/; alias pip=pip3; source requirements.sh;
alias python=python3; source parse.sh
and I haven't tried to run the training.
The package probably is unnecessary in inference. I am running the pre-trained model on CentOS. It works well without the package. Is it needed for training a model?
https://github.com/KhalilMrini/LAL-Parser/blob/master/requirements.sh contains:
However
libhdf5-serial-dev=1.8.16+docs-4ubuntu1.1
is not available on Ubuntu 18.04 (https://packages.ubuntu.com/search?keywords=libhdf5-dev&searchon=names" only available on 16.04): is it ok to instead usesudo apt-get install libhdf5-serial-dev=1.10.0-patch1+docs-4
or should we stick tolibhdf5-serial-dev=1.8.16+docs-4ubuntu1.1
(in which case I'll add the .deb to the repo to make it easier for Ubuntu 18.04 users)?