Closed cgrandin closed 2 years ago
Does the number of parameters change because time-varying selectivity is removed?
Yes, there are 5 less going from -6 years to -7 years, and then 5 less again going to -8. But -8, -9, and -10 seem fine
I thought that maybe we had went far enough back in time that the whole block was gone. Bad guess on my part. I think this is going to take exploration on a local machine or logging into AWS to diagnose.
It's just this if you want to try it:
run_retrospectives('models/2022.01.23_age1Survey', retrospective_yrs = 7, n_cores = 16)
Change the n_cores though
If you want to try it on AWS go to:
35.182.247.50:8783
User: kjohnson
pass: a
Click Terminal
tab, and copy the model output files from S3:
cp hakestore/models/2022.01.23_age1Survey/* models/2022.01.23_age1Survey
Then click Console
tab and run the code:
run_retrospectives('models/2022.01.23_age1Survey', retrospective_yrs = 7, n_cores = 16)
Currently, I am running this locally and Chris is re-running on AWS and they both appear to be working.
@cgrandin i think we can close this right?
Yes this has been done and incorporated into the results
run_retrospectives('models/2022.01.23_age1Survey', retrospective_yrs = 7, n_cores = 16)
fails, with this error:
Error in checkForRemoteErrors(val) : 16 nodes produced errors; first error: models/2022.01.23_age1Survey/retrospectives/retro-07/mcmc/chain_1/adaptation.csv exists = FALSE models/2022.01.23_age1Survey/retrospectives/retro-07/mcmc/chain_1/unbounded.csv exists = FALSE NUTS failed to run in chain 1. Check inputs.
So the
adaptation.csv
file is not being created by ADMB as it is in every other retrospective run. I tried a few things like increasing the initial adnuts run to 500 iterations, see the link below. I also changed the seed but it made no difference. I think the algorithm starting at the MLE values is not correct somehow but I don't know what to look for.See https://github.com/pacific-hake/hake-assessment/blob/27aeccb688149e0007b93dafa3740b85590c61f9/R/run-adnuts.R#L91