I can not install it

[pjssilva]

I tried installing EMP.jl by cloning the repository and activating it. It fails when installing ReSHOP (that is also written by you) because the file libreshop.tar.xz is 0 bytes long. Then, tar fails to cope with it. Are EMP.jl and ReSHOP.jl still active? Do you have any plans to update them? If not, any suggestions of another Julia package with similar capabilities?

[xhub]

Thanks for your interest in this package. I've tagged a release of ReSHOP which removed the old install method and only used a jll.

EMP.jl has not been updated to a few breaking changes in the Julia ecosystem.
Unfortunately, it's time-consuming to provide fixes with no real gains.
I hope to come back to updating this package soon.
Note that the package is not registered, reflecting on the beta-status

What capabilities are you interested in?

[pjssilva]

I can now install it, but as you said it is still broken due to changes in JuMP/MOI.

For me, the most important capability is to be able to derive the KKT conditions of multiple NLPs and construct an EPEC that is suitable to a solver like PATH to search for an equilibrium. Unfortunately, I could only find EMP.jl to do it "easily". Thanks a lot for your effort and I hope that JuMP/MOI become more stable, decreasing the burden on package matainance.

[xhub]

EPECs (Nash equilibrium problem with individual agents being bilevel/MPEC) are challenging. They cannot be solved by PATH directly. One strategy would be to use NLPEC to reformulate each bilevel/MPEC problem into an NLP. This yields a (parametric) MOPEC and PATH can be used to solved it. By solving a sequence of MOPEC, where the parameters introduced by NLPEC are driven towards appropriate values, one can find a "solution" to the EPEC.

A cursory look at https://www.gams.com/38/docs/S_JAMS.html#EMP_EMPINFO_FILE_DETAILS seems to indicate that this is not currently possible in GAMS. Do you have a GAMS model of the EPEC to be solved?

If each individual agent is an NLP, then this constitute a NEP or MOPEC. Then, this would be much easier to solve as one just needs to stack all KKT systems together and solve the resulting MCP/VI. In that case this package could be considered (once fixed)