Nov 20, 2022
- UPDATE: March 2024
- java version: 1.8.0_392
- installed with eclipse temurin (via sdkman)
- sbt version: 1.0.2
- should be reflected in
project/build.properties
- installed with SDKMAN!
- should be reflected in
- scala version: 2.11.12
- should be reflected in
build.sbt
andscripts/config.sh
- installed with SDKMAN!
- should be reflected in
- see modified
project/plugins.sbt
- see references below
- make executables
chmod +x setup compile sbt scripts/config.sh
- run
sbt
first! (not./sbt
but justsbt
)setup
, etc scripts
- see
jamr_post_setup.sh
which fixes outdated perl expression - references:
- requires python2
pip2 install virtualenv
thenpython2 -m virtualenv python2-env
python2-env
virtualenv within ISI dir by default; or set up new one separately- UPDATE:
pyenv virtualenv 2.7.17 isi
mgizapp
requires Boost c++ librariessudo apt-get install libboost-all-dev
- read instructions in
INSTALL
insidemgizapp
- running
scripts/jamr2isi.py
on JAMR output prepares input data in required format
requires python3 (3.7.3 or 3.7.6 on ubuntu 22.04)- UPDATE: 3.7.17 seems required, along with spacy==2.3.7
- pyenv virtualenv by default; or set up local venv separately and configure the script acordingly
- torch
1.13.1+cu117
- relies on Stanza and spaCy whose version may be different from what model uses
- set up Stanza with
STANZA_RESOURCES_DIR=./leamr/leamr_stanza_resources python -c "import stanza; stanza.download('en')"
- set up Stanza with
- also need to set up
neuralmonkey
separatelypip install cython==0.29 --upgrade
- don't forget
python -m spacy download en
- contains minor modifications
- doesn't drop
.txt
in the outputs filename - uses pre-tokenized outputs
- doesn't drop
May have to set pyenv global
to Python2 + Python3, i.e.
pyenv global 2.7.15
- since Python3 are usually pre-installed, this exposes python2 and python3.
To run aligners in sequence (JAMR -> ISI -> LEAMR)
./scripts/run_aligners.sh [input_file]