Name and order of the relative frequencies and rates

[maremita]

Hello, In the CSV sampling file, the parameters of the substitution models are tagged by their index (substmodel.frequencies.[0, 1,..], substmodel.rates.[0, 1,..]) and not by their names (for example [A, G, C, T] and [AG, AC, AT, GC, GT, CT] for nucleotide). I couldn't find in the source code documentation of their name/order, and it's not easy to guess it from the code of q() that builds the substitution matrix of the GTR model.

It will be helpful to document their names somewhere to facilitate comparisons and post-analysis tasks.

Thank you Mathieu @4ment. Amine.

[maremita]

I ran some simulations with different frequencies and rates values and compared the results to PhyML. I think the order for torchtree is frequencies = [A, C, G, T] and rates = [AC, AG, AT, CG, CT, GT], which follows the standard order (sorted alphabetically and found in popular tools). However, we can find different orders in other tools and textbooks. For example the recent variational phylogenetics tool VBPI uses frequencies = [A, G, C, T] and rates = [AG, AC, AT, GC, GT, CT]. EvoVGM, a generative variational model, adopted the same order as VBPI.