R : "weights = varIdent" helps to deal with unbalanced classes in a linear model?

Question

I want to fit a linear mixed model (because I have repeated measures over time).

LM.fit <- lme(VR ~ Group * Time + Age + sexe + study,
             random = ~1|subject,
             method = "REML",
             na.action = na.omit,
             data = Data)

I check the assumptions (which is ok I think ?!)

BUT I have unbalanced classes in Group : Subgroup1 have 80 samples and Subgroup2 have 20 samples..

LM.fit <- lme(VR ~ Group * Time + Age + sexe + study,
             random = ~1|subject,
             weights = varIdent(form = ~1|Group),
             method = "REML",
             na.action = na.omit,
             data = Data)

weights = varIdent(form = ~1|Group) I heard that adding this line can help accounting for unequal variances across groups and unbalanced classes. Can someone help me to know why and how?

If you have other methods, i am listening