Different results coming from LBFGS in Optim.jl

I’m trying to reproduce the solution of the LBFGS across different machines, but even under the same environment (same Manifest.toml), I get different minimizers. Is that possible?

Hi, could you also be more precise on how different are the results? Are the differences non-trivial?

The result is close to the optimal found in other machines by 2 decimal places.