The presumption is that a pitched sound consists of partials that have frequencies that are integer multiples of the fundamental frequency, so that a note with fundamental frequency f (e.g. 100Hz) has partials at f, 2f, 3f (100, 200, 300 Hz) and so on - or in terms of ratios, 1:1, 1:2, 1:3 and so on. It's these ratios that the deviation is from in an inharmonic sound - for example, this spectrum of a bell:
(Picture by user Hyacinth from page https://en.wikipedia.org/wiki/Inharmonicity)
We can see here that the ratios are 1:1, 1:2.23, 1:3.73, and so on, so they could be said to deviate from the presumed 1:1, 1:2, (1:3, for which a near-equivalent is missing), 1:4... rations of a pitched note.
If a sound is inharmonic, it's only when considering it in the frequency domain like this that we can actually see a fundamental frequency. Viewed in the time domain, a sound that has non-integer-ratio partials will not actually be periodic; you could therefore consider that its actual fundamental frequency is less well-defined than for a sound that consists of only harmonic partials.