Grad students often aren’t quite sure how much of their work time should be spent reading, and may also have trouble figuring out what to read during that time. (In principle this problem also applies to professors, but it’s not much of an issue in practice since we have almost no time for discretionary reading.) I usually tell people to err on the side of doing too much, as opposed to reading too much. Averaging one day a week on reading is probably about right, but it may be a lot higher or lower than this during certain periods of grad school.
Of the time spent reading, about half should go towards solving immediate research problems. The purpose of this is getting unstuck, learning about alternate approaches, avoiding reinvention of known techniques, and learning how and what the current players in your field are thinking. Of the remaining 50% of reading time, half of it should go towards “deep background reading” — reading in the thesis area that goes a bit further afield and a bit further back in the chain of references than is absolutely necessary to get the work done. This reading will form the basis for the related work sections of papers and the dissertation. The last 25% of reading time is purely discretionary: classic papers, good books, new results from conferences, etc. This material is not expected to be directly relevant (though it may turn out to be) but keeps a person up to date on random interesting topics.