search

beauchamplab / rave

For all RAVE info, including installation instructions, visit
https://beauchamplab.github.io/rave
23 stars 5 forks source link

RAVE install error #48

Closed satra closed 3 years ago

satra commented 3 years ago

following the instructions on the wiki, i'm running into an install error:

ERROR: dependency ‘rave’ is not available for package ‘ravebuiltins’
* removing ‘/opt/conda/lib/R/library/ravebuiltins’
Warning messages:
1: In i.p(...) :
  installation of package ‘fftwtools’ had non-zero exit status
2: In i.p(...) :
  installation of package ‘/tmp/RtmpOOhBpX/filec34d80ec92/rave_1.0.2.tar.gz’ had non-zero exit status
3: In i.p(...) :
  installation of package ‘/tmp/RtmpOOhBpX/filec346b11e6b/ravebuiltins_0.1.1.tar.gz’ had non-zero exit status

system: "Ubuntu 20.04.2 LTS" R version 4.0.3

also if there is a new Docker image/Dockerfile that shows how to install the tool automatically would be much appreciated.

jmagnotti commented 3 years ago

Thanks for posting this error. It may be that the system doesn't have the fftw library installed. Can you try to install it through the package manager?

 sudo apt-get install libfftw3-dev libfftw3-doc

If this works we'll be sure to update the Linux install instructions and work on updating the docker image as well.

satra commented 3 years ago

thanks @jmagnotti - i installed conda's r-fftwtools package and then the error i get is:

Cannot create directory at '/data/rave_data/cache_dir' 
Error : Cannot create directory at '/data/rave_data/cache_dir'

any idea why it's trying to write to /data and how i can have it not use that location. i'm trying to do this without any superuser privileges.

jmagnotti commented 3 years ago

Thanks for reporting this further issue. Sounds like the installer isn't detecting the home directory correctly... We'll have to dig into this a bit and get back to you. Is this a fairly vanilla Ubuntu install?

satra commented 3 years ago

Sounds like the installer isn't detecting the home directory correctly..

do you save any config info somewhere. if so i was playing with various environments including running the rave docker image. so it could have created some files somewhere that i should perhaps remove.

Is this a fairly vanilla Ubuntu install?

yes we are just building of the latest iteration of the jupyter datascience container and i'm doing things inside a conda environment that has various R packages installed.

dipterix commented 3 years ago

Hi, @satra . Thanks for reporting this installation issue. We develop RAVE on osx, and the instructions on Linux is a little bit out-dated. I'm sorry for the inconvenience. We just tested installing RAVE on Ubuntu 20.04. The updates will be posted pretty soon. @jmagnotti could also set up a zoom call to troubleshoot the issue.

Normally rave sets it data repository under your home address. I remember the only place that I manually set the data repo is in the docker images. Therefore I guess you were trying to use the Dockerfile to install on your native linux system or migrate to your own docker image. This approach is not recommended.

(Below are the instructions to reset RAVE settings. Skip if you are building docker images)

You can either use R command to reset RAVE settings, or go to the settings file to edit.

R command solution

Open R (in your terminal, type "R" and return) or RStudio console, type and return

raveio::raveio_resetopt()

This should reset your rave configurations. Then you can use rave::rave_options() to launch the browser to edit.

Direct editing

You can find under the settings yaml file under ~/.config/R/raveio. Open with text editor, then you should be able to find something like this:

data_dir: /home/pi/rave_data/data_dir
delay_input: 20.0
test_mode: no
py2_path: ''
py_virtualenv: ''
session_string: AfvMGMQ0sv0mQ9n
..temp: []
module_root_dir: ~/rave_modules/
verbose_level: DEBUG
file_structure: native
module_lookup_file: ~/rave_modules/modules.csv
image_height: 768
disable_startup_speed_check: no
py3_path: ''
suma_nodes_per_electrodes: 42
rave_ver: 1.0.2
bids_data_dir: /home/pi/rave_data/bids_dir/
max_mem: 7.6301994
drive_speed:
- 250.0
- 120.0
server_time_zone: America/Chicago
fast_cache: yes
raw_data_dir: /home/pi/rave_data/raw_dir
image_width: 1280
max_worker: 3

Correct data_dir, raw_data_dir, and probably bids_data_dir.

You can also set maximum number of workers by changing max_worker. If you are setting up RAVE for shared use, please set it to 1 to save the memory

After the fix

Once the configuration file is fixed, you probably need to finalize installation in R (since the original installation failed). Go to https://openwetware.org/wiki/RAVE:Install#Install_RAVE section 1, start from step 3.

You should then be able to launch RAVE via R code:

rave::start_rave()

If you want to download some demo subjects, run the following R code. It will download from this page 

# You can change YAB to KC, YAD, YAI, YAF, YAH...
options('timeout' = 1e10); rave::download_sample_data("YAB")
satra commented 3 years ago

@dipterix - thanks so much for this. indeed it was me playing around with the docker image that caused the issue. i now have rave working on dandi hub ! (@yarikoptic @bendichter) the ideal thing would be to have a docker image for rave since the install process takes a bit of time.

i was also able to get the current released docker image working as well (the image also says v1.0.2 - so does it mean its the same version as what i installed with the commands).

in the hub terminal you can do

singularity pull docker://beauchamplab/rave

# this is not necessary, but i wanted to fully isolate rave
mkdir -p home
mkdir -p home/data
mkdir -p home/jovyan

singularity exec --env NCPUS=4 -e --no-home -B $(pwd)/home:/home -B $(pwd)/home/data:/data rave_latest.sif start_rave

and then point the browser to: https://hub.dandiarchive.org/user/YOURUSERID/proxy/6767/ (the last / is important for rave to find the relevant pieces)

so both a standalone install and a docker image works on dandihub.

image