Modular designs are slightly easier to build, much easier to extend after being built, are much easier to isolate sections and allow you to have a pivot at the hub-end of each arm so the direction and strength of the artificial gravity can be maintained even under thrust.
Contiguous torus or cylindrical shapes will have weird effects at boost time and brake time, because the direction of gravity will not be "down" from the point of view of the contents. This isn't a problem for space stations, or spacecraft coasting, or even spacecraft on relatively short-haul flights with brief engine burns, but given that starships are likely to have exceedingly long boost and brake phases (possibly years) you want to design your gravity decks to handle that situation. There's a related question on this site that might be relevant (warning: it contains a long rambling answer by me): How fast can a ship with rotating habitats be accelerated?
If your starship is travelling reasonably quickly (by which I mean "could get to another star in under 100 years) you'll need a decent amount of forward shielding to protect your ship from collisions with interstellar dust, and reduce the radiation hazard of incoming particles. Shielding is dead mass, so really you want as little of it as possible. You therefore concentrate your shielding at the front of your ship (or habitation section) and have a loooong structure behind it. A cylindrical shape with a shielding endcap gives you much more volume for a given shielding area than a torus will. I found a nice paper on a segmented cylindrical starship design recently, and I seem to have lost it. Sorry :-(
Finally, do consider that frozen or dessicated meat cargo won't care about artificial gravity though, and will be easier to pack and shield.
In the absense of the segmented design I had hoped to share, consider instead the ISV Venture Star, the starship from Avatar.

Spun gravity section highlighted in red (modified image taken from Galaxy's spun gravity question linked above) currently folded up for thrust mode. The gravity section is tiny, it remains useful under high thrust, almost all of the payload is in unspun storage sections. Most of the rest of the design of the ship is pretty high quality too, with the slight exception of the mid-course flipover (which is suicide) and using rockets to brake instead of a magsail. That aside, this is an excellent starting point for any starship design question.
Lets have a quick look at the "continuous thrust" approach to providing artificial gravity, though.
It'll take you about 11 years, 285 days (from the point of view of an external observer) to fly to somewhere 10 light-years away at a continuous 1g and stop at your destination. At the turnaround point, you'll be travelling at about .987c, a Lorentz factor of ~6.16, and you'll have a kinetic energy of about 4.6x1017 joules per kilo of ship. A 100000 tonne starship requires an average engine power of about 2.5x1017 watts over its entire flight time, which is about 1.14 on the Kardashev scale.
Every nucleon you hit at that speed has about 6GeV of energy (compare with the ~2GeV of energy released when a nucleon and antinucleon annihilate), and you'll need shielding equivalent to several tens of metres of water per square metre of ship cross-section to keep all organic and electronic things safe. A 10m radius cylinder would therefore need at least 6000 tonnes of frontal water or ice shielding (realistically you'll want more more, especially if you have to handle mass loss over the journey!), not including shielding mass for galactic cosmic rays and for whatever drive system might be used.
The longer the distance flown, the worse this will get because the peak speeds will be that much higher.