Hacks territory there, the internal pull-ups are not low resistance like the SCSI bus terminators so technically the bus isn't "in spec" if you go that route, but I'd vouch I personally prefer that hack if it works because it also negates the need for transceivers with high-current sinking capability. And, therefore, you cut most of the cost of RaSCSI.
And actually, that's a good point to mention, looks like this is using "direct connection" to the SCSI bus and I wouldn't think the STM32 is rated for 48mA on each I/O pin. And it isn't, I checked the datasheet and it says 25mA as the "absolute maximum."
Note that 48mA is the maximum rating in the SCSI spec assuming a bus has two passive terminators on each end, which supply 23mA each. Of course our typical Macintosh SCSI buses fare well with only a single passive terminator.