You can get sampling with the effect of a particular temperature by trying different MC temperatures until you find one that works. I’m unaware of any way to meaningfully convert the temperature values used by Monte Carlo into Kelvins. Note that the conversion will vary depending on what scorefunction terms you are using and what weights they have.
Another consideration is that most of the community is not using Monte Carlo for thermodynamic sampling – we’re using it just for optimization. The temperature conversion is entirely empirical in this case – use the temperature that optimizes best. (Start in the range of 0.5-2 and see what happens).
You could try contacting Colin Smith in the Kortemme lab if you’re interested in thermodynamically valid sampling, he’s been bolting new pieces into Rosetta to allow for that.