Temperature 1.0 used to be bad for sampling. 0.7 was the better choice, and the difference in results were noticeable. You may want to experiment with this.

▲

danielhanchen a day ago | parent [-]

You might be right, but Google's recommendation was temp 1 etc primarily because all their benchmarks were used with these numbers, so it's better reproducibility for downstream tasks

▲

sillysaurusx a day ago | parent [-]

Fair, though putting a note in the readme about temperature 0.7 couldn't hurt.

I wonder why they do benchmarks with 1 instead of 0.7... that's strange. 0.7 or 0.8 at most gives noticeably better samples.

▲

davedx a day ago | parent [-]

Reproducibility. They're benchmarks.

	▲	sillysaurusx a day ago \| parent [-]
		Reproducibility is a matter of using the same input seeds, which jax can do. 0.7 vs 1.0 would make no difference for that. Without seeds, 0.7 would be less random than 1.0, so it'd be (slightly) more reproducible.