This is Hume's Treatise 3.1.2, paragraph 9:
But nature may also be oppos'd to artifice, as well as to what is rare and unusual; and in this sense it may be disputed, whether the notions of virtue be natural or not. We readily forget, that the designs, and projects, and views of men are principles as necessary in their operation as heat and cold, moist and dry: But taking them to be free and entirely our own, 'tis usual for us to set them in opposition to the other principles of nature. Shou'd it, therefore, be demanded, whether the sense of virtue be natural or artificial, I am of the opinion, that 'tis impossible for me at present to give any precise answer to thsi question. Perhaps it will appear afterwards, that our sense of some virtues is artificial and that of others natural. The discussion of this question will be more proper when we enter upon an exact detail of each particular vice and virtue.
The Nortons' notes to this paragraph says that "Treatise 3.2 shows that some virtues are artficial, while Treatise 3.3 discusses the natural virtues" (p. 539).
This is not what Hume is talking about, however. He does not say "Perhaps it will appear afterwards, that some virtues are artificial and others natural," but "Perhaps it will appear afterwards, that our sense of some virtues is artifical and that of others natural." This is actually a very different issue. Hume does regard some virtues as artificial, and some as natural; (property-)justice, for instance, is not a natural virtue, but one that depends on human artifice and convention. But Hume's position on our sense of virtue is quite different; our sense of virtue is natural. He explicitly says this at the very end of Book 3 (3.3.6, paragraphs 3 and 4). This turns out, I think, to be essential to Hume's view. Several of his arguments about the relation between virtues that are virtues because of 'agreeableness' and virtues that are virtues because of 'utility' actually require that our sense of morals, or sense of virtue, be fundamentally natural, not the result of convention. Convention as it were diversifies it, and creates new virtues; but all these virtues have to trace back (through sympathy, generally) to the original sentiments of approbation and disapprobation that are natural to us. So in the passage above I read Hume as saying that, as far as he has discussed to that point, it could well be that our sense of some virtues is artificial and of others natural; discussion of this would require a closer examination of particular virtues. But, when he performs such an examination, he concludes that while (property-)justice and others like it are artificial virtues, our sense of them is natural. So he ends up rejecting the view he says is 'perhaps' the case here; the 'perhaps' needs to be emphasized and confined to context. It's possible, of course, that I'm missing some nuance in Hume's later use of 'natural', but I don't think so.
By the way, sometime soon I'll be doing a post on Hume's philosophical account of good manners, which is very interesting and not sufficiently examined.