Using a sample of 70 924 stars from the second data release of the GALAH optical spectroscopic survey, we construct median sequences of [X/Mg] vs. [Mg/H] for 21 elements, separating the high-$\alpha$/``low-Ia'' and low-$\alpha$/``high-Ia'' stellar populations through cuts in [Mg/Fe]. Previous work with the near-IR APOGEE survey has shown that such sequences are nearly independent of location in the Galactic disk, implying that they are determined by stellar nucleosynthesis yields with little sensitivity to other chemical evolution aspects. The separation between the two [X/Mg] sequences indicates the relative importance of prompt and delayed enrichment mechanisms, while the sequences' slopes indicate metallicity dependence of the yields. GALAH and APOGEE measurements agree for some of their common elements, but differ in sequence separation or metallicity trends for others. GALAH offers access to nine new elements. We infer that about $75\%$ of solar C comes from core collapse supernovae and $25\%$ from delayed mechanisms. We find core collapse fractions of $60-80\%$ for the Fe-peak elements Sc, Ti, Cu, and Zn, with strong metallicity dependence of the core collapse Cu yield. For the neutron capture elements Y, Ba, and La, we infer large delayed contributions with non-monotonic metallicity dependence. The separation of the [Eu/Mg] sequences implies that at least $\sim30\%$ of Eu enrichment is delayed with respect to star formation. We compare our results to predictions of several supernova and AGB yield models; C, Na, K, Mn, and Ca all show discrepancies with models that could make them useful diagnostics of nucleosynthesis physics.