Question about stack_modelling use

[nikhailarum]

Hi Sylvain

I have finally been able to get the cluster to read my GIS data and occurrence data correctly. However when running the stack_modelling function I get a number of errors relating to the cluster. This is the code I am using for the stack_modelling function:

SSDM <- stack_modelling(c("GLM", "GAM", "MARS", "GBM", "RF", "CTA", "MAXENT", "ANN", "SVM"), Occurrences, Env, Xcol = "decimallon", Ycol = "decimallat", Pcol = NULL, Spcol = "species", rep = 1, name = "Stack", save = TRUE, path = getwd(), PA = NULL, cv = "holdout", cv.param = c(0.75, 1), thresh = 1001, axes.metric = "Pearson", uncertainty = TRUE, tmp = TRUE, ensemble.metric = c("AUC", "Kappa", "sensitivity", "specificity"), ensemble.thresh = c(0.75, 0.75, 0.75, 0.75), weight = TRUE, method = "bSSDM", metric = "SES", range = NULL, endemism = NULL, verbose = TRUE, GUI = FALSE, cores = 125)

In return I get a number of errors which I am struggling to interpret:

Error in checkForRemoteErrors(val) : 19 nodes produced errors; first error: comparison of these types is not implemented Calls: stack_modelling ... clusterApply -> staticClusterApply -> checkForRemoteErrors Execution halted Error in unserialize(node$con) : error reading from connection Calls: <Anonymous> ... doTryCatch -> recvData -> recvData.SOCKnode -> unserialize In addition: Warning message: In eval(e, x, parent.frame()) : Incompatible methods ("Ops.data.frame", "Ops.factor") for "==" Execution halted

This happens immediately after console states that it is exporting environment to clusters. I understand that on this run I may have allocated more cores than I am allowed however the same error message crops up when I allocate just a fraction of those cores (19). I doubt the issue is with any dataset as my rasters and occurrence data are read correctly in the environment and return the correct values. I am not sure what this error is trying to tell me or indeed how to fix this issue. With your experience of working with the package on an HPC I was hoping you could point me in the right direction as to how I could solve this error. Is this due to the cluster not recognizing which variables to export? Do i need to insert another function to fix this issue?

Many Thanks NIkhail

[sylvainschmitt]

Following mail exchanges, it appears that the issue is not coming from the cluster. So could you give more detail about what happen on your local machine ? Or close the issue if you solve the problem.

[nikhailarum]

Hi Sylvain

On my local machine I get the exact same error message as on the cluster. The same error message pops up when use the gui interface to run the analyses. I haven't had as much time to play with this issue as I would like as I have been busy this week, and have had laptop issues.

Regards Nikhail

On Thu, 22 Mar 2018, 10:00 Sylvain SCHMITT notifications@github.com wrote:

Following mail exchanges, it appears that the issue is not coming from the cluster. So could you give more detail about what happen on your local machine ? Or close the issue if you solve the problem.

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/sylvainschmitt/SSDM/issues/17#issuecomment-375209977, or mute the thread https://github.com/notifications/unsubscribe-auth/Ah4n6rXlw-MjPiisnnYZaWeyXSX8iC1fks5tg1oDgaJpZM4SNPN8 .

[sylvainschmitt]

Any update on the issue in order to close it or help you ?

[nikhailarum]

Hi Sylvain

Sorry for the lack of feedback. I have taken my analyses in a different direction and I am not using a niche modelling approach for my thesis anymore. However, when I have some spare time I will continue to try different solutions. All I can say is that when I tried it on my personal laptop with just the two species I got the exact same error message as on the cluster. However it doesn't tell me anything more specific in the error message.

Thanks, Nikhail

On Fri, 30 Mar 2018, 08:44 Sylvain SCHMITT notifications@github.com wrote:

Any update on the issue in order to close it or help you ?

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/sylvainschmitt/SSDM/issues/17#issuecomment-377459118, or mute the thread https://github.com/notifications/unsubscribe-auth/Ah4n6ugnr9VNSc37J0BVAFG0mYQpQJcPks5tjdRRgaJpZM4SNPN8 .

[sylvainschmitt]

Ok so I am closing the issue now but don't hesitate to reopen it in case of need

[nikhailarum]

That's perfect thanks Sylvain

On Fri, 30 Mar 2018, 08:55 Sylvain SCHMITT notifications@github.com wrote:

Ok so I am closing the issue now but don't hesitate to reopen it in case of need

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/sylvainschmitt/SSDM/issues/17#issuecomment-377460670, or mute the thread https://github.com/notifications/unsubscribe-auth/Ah4n6knBLEES5iT2Zg8Oy0cNaFb_tBogks5tjdbhgaJpZM4SNPN8 .

[grenyer]

Hi Sylvain - I've run into the same issue. Could you summarise either here or by email (grenyer@gmail.com) anything I should know from your discussions with nikhailarum above? It might save some time in the debugging...

[sylvainschmitt]

Hello, what type of model are you building ? Are you specifically using PPR.MEM ? Because another user spotted a bug that I corrected on GitHhub but which is still not on CRAN. Try to install the package from CRAN with following command and see if the bug persist:

devtools::install_github("sylvainschmitt/SSDM")

Let me know how it goes.