I ran into some confusion answering this question.
The OP asked for the number of partitions of an integer into odd numbers greater than $1$, with the additional constraint that there are two different choices (or colors) for each integer (it might be that the OP asked for compositions, but even if, I would like to treat partitions now). Eg we have four possibilities to partition $8$, namely $5_{red}+3_{red}$, $5_{red} + 3_{blue}$, $5_{blue}+3_{red}$ and $5_{blue} + 3_{blue}$.
Looking into the standard reference (Analytic Combinatorics), I found that partitions may be defined via the multiset construction (page 29ff). There, one first takes a combinatorial class $\mathcal B$ and its generating function $B(z)$ and sets $\mathrm{MSET}(\mathcal B) = \prod_{n \ge 1} (1-z^n)^{-B_n}$.
We have $B(z) = \sum_{n \ge 1} 2 \cdot z^{ 2n+1} $ (there are two possibilities for each odd number greater $1$), and so the generating function we are looking for is
$$
P_1(z) = \prod_{n \ge 1} (1-z^{2n+1})^{-2}.
$$
Yet this function produces wrong coefficients (already the coefficient of $z^6$ is $3$, which is clearly wrong). A different function which I derived in an earlier attempt,
$$
P_2(z) = \prod_{n \ge 1} (1-2z^{2n+1})^{-1},
$$
produces the right coefficients. So I guess $P_2(z)$ must be the right generating function, yet I don't clearly understand why. Also I don't know what exactly $P_1(z)$ counts. Did I misunderstand the multiset construction?