microsoft / LightGBM

A fast, distributed, high performance gradient boosting (GBT, GBDT, GBRT, GBM or MART) framework based on decision tree algorithms, used for ranking, classification and many other machine learning tasks.
https://lightgbm.readthedocs.io/en/latest/
MIT License
16.68k stars 3.83k forks source link

got an error on installing lightgbm on linux #2057

Closed franciszmy closed 5 years ago

franciszmy commented 5 years ago

I have used lightgbm by using rstudio on centOS before, but for some reasons, yesterday I deleted my lightgbm package on rstudio server and when I tried to reinstall it later, it posted an error which I can't solve.

The error is

** testing if installed package can be loaded Error in UseMethod("conditionCall") : no applicable method for 'conditionCall' applied to an object of class "character" Calls: ... tryCatchList -> tryCatchOne -> -> conditionCall Execution halted ERROR: loading failed

this error seems a base function in R, but I don't know why something is wrong with this base function.

Anyone could help me? Thanks so much in advance!!!

jameslamb commented 5 years ago

Hi @franciszmy , we're here to help!

Can you please share the following?

  1. The steps you used to remove lightgbm
  2. The steps you used to reinstall it
  3. The output for running sessionInfo()
franciszmy commented 5 years ago

thanks james, here are some information about what you mentioned above.

1.I just removed package from rstudio GUI. No other extra moves 2.I have lightgbm 2.0.1.0 package. I reinstalled it by using install.package in R

  1. sessionInfo() R version 3.4.4 (2018-03-15) Platform: x86_64-pc-linux-gnu (64-bit) Running under: CentOS release 6.10 (Final)

Matrix products: default BLAS: /opt/rinstall/R-3.4.4/lib/libRblas.so LAPACK: /opt/rinstall/R-3.4.4/lib/libRlapack.so

locale: [1] LC_CTYPE=en_US.UTF-8 LC_NUMERIC=C LC_TIME=en_US.UTF-8 LC_COLLATE=en_US.UTF-8
[5] LC_MONETARY=en_US.UTF-8 LC_MESSAGES=en_US.UTF-8 LC_PAPER=en_US.UTF-8 LC_NAME=C
[9] LC_ADDRESS=C LC_TELEPHONE=C LC_MEASUREMENT=en_US.UTF-8 LC_IDENTIFICATION=C

attached base packages: [1] stats graphics grDevices utils datasets methods base

other attached packages: [1] R6_2.4.0 xgboost_0.6-4

loaded via a namespace (and not attached): [1] compiler_3.4.2 magrittr_1.5 Matrix_1.2-11 tools_3.4.2 stringi_1.1.6
[6] grid_3.4.2 data.table_1.10.4-3 lattice_0.20-35

Besides, the reason why I delete my lightgbm package is because I cannot use lightgbm package that I used to use successfully, the reason why I cannot use lightgbm is because I find later that the R is updated to 3.5 , so I reinstall R with 3.4 version.

below is the details of installation information:

Installing package into ‘/home/xxxx/R/x86_64-pc-linux-gnu-library/3.4’ (as ‘lib’ is unspecified)

thank you so much!

jameslamb commented 5 years ago

Hmmm I'm still trying to figure this one out.

That error message is a developer error somewhere in the code used to load packages. It indicates that some R code is relying on S3 dispatching but got a type of data with no registered method.

You're saying that you downgraded from R 3.5.x to R 3.4.x? Why did you have to do that? My best guess right now is that something was left behind on your system and that R or cmake thinks something is cached that actually needs to be recompiled.

R 3.5 was a significant departure from 3.4 (https://cran.r-project.org/doc/manuals/r-devel/NEWS.html) so it's possible for a package that worked on 3.5 to not be installable on 3.4.

The R side of our CI needs some work, so I cannot say with certainty what versions of R we support. I'll try to reproduce your environment and see if I can get the same error!

While I try to do that, could you manually remove /home/xxxx/R/x86_64-pc-linux-gnu-library/3.4/lightgbm and try re-installing?

One other thing...is there still a /home/xxxx/R/x86_64-pc-linux-gnu-library/3.5 left behind? When I've used RStudio Server with a shared file system in the past, sometimes I've needed to remove other R/ directories left behind when changing versions.

StrikerRUS commented 5 years ago

ping @franciszmy

StrikerRUS commented 5 years ago

@franciszmy Have you managed to try @jameslamb's advices about remaining folders?

Laurae2 commented 5 years ago

@jameslamb @franciszmy Just had the same issue last week when switching to R 3.6 (except it's from R 3.5.3) on another package. RStudio Server is not fun to use when switching R versions.

In that case, it was better to nuke (delete) the hidden .rstudio folder to fully reset RStudio Server. Note that doing it will delete also all customizations done in RStudio Server (which have to be re-done manually).

You can find it in your home directory: ls ./.rstudio from your home directory.

jameslamb commented 5 years ago

😬 I feel that we have a set of strange R 3.6.0 issues coming our way 😬

@franciszmy could you try mine and @Laurae2 's suggestions and let us know if the issue gets resolved?

StrikerRUS commented 5 years ago

Closing due to inactivity.