3.6 Robyn_Run Issue

[F1nalFortune]

3.6 Robyn_Run Issue

Issue

Issue arises when running robyn_run function. Here's how I'm calling the robyn_run function:

OutputModels <- robyn_run(
  InputCollect = InputCollect # feed in all model specification
  , cores = 6 # default
  ,iterations = 6000 
  ,trials = 5 
  , outputs = FALSE
)

and here's the error:

Initial model is built on rolling window of 200 week: ...
Using weibull_pdf adstocking with xx hyperparameters (xx to iterate + 0 fixed) on 1 core (Windows fallback)
>>> Starting 5 trials with 6000 iterations each using TwoPointsDE nevergrad algorithm...
  Running trial 1 of 5
Error in seq.default(log(lambda_max), log(lambda_max * lambda_min_ratio),  : 
  'from' must be a finite number

The error is fixed when I comment out the prophet_vars with robyn_inputs, however, I would like to keep the prophet_vars in the model.

Environment & Robyn version

R version 4.1.2 (2021-11-01)
Robyn version Robyn_3.6.0

[laresbernardo]

Can you please share a reproducible example? There must be an issue with lambda_max calculation which is calculated on Robyn:::lambda_seq(). Do you have NA values in dt_mod? Can you share that data.frame as CSV, preferably anonymized?

[gufengzhou]

FYI We got the same issue in the FB group too, see here. It's due to NA/Inf/-Inf values. Would you please see the post and check too?

[gufengzhou]

FYI we've just fixed this error. Let me know if adding back prophet works now d748aa7

[F1nalFortune]

@laresbernardo @gufengzhou Issue resolved! Thanks for the support.

Best,
Brian

[ekenheim]

Having the same error message currently - fresh install of R (all packages updated)

[gufengzhou]

We just had a new release V3.8.0. Please try it out!