← Back to context Comment by throwa356262 1 day ago Better performance than TQ and better quality than FP16?Am I reading this right?? 6 comments throwa356262 Reply qeternity 1 day ago It's not better quality: 59.3% vs 59.4% fp16 on AIME 25 sheepscreek 18 hours ago 0.1% is within margin of error. Depending on the performance boost, it might be worthwhile taking a minuscule quality hit. electroglyph 21 hours ago any divergence (even if the benchmark is better) from full precision is error 7e 14 hours ago Just pretend that it is the next step update when training. You didn’t train your model to step=inf, I hope? thefox96 1 day ago Faster than Fp16, not better quality i guess pbich 1 day ago [dead]
qeternity 1 day ago It's not better quality: 59.3% vs 59.4% fp16 on AIME 25 sheepscreek 18 hours ago 0.1% is within margin of error. Depending on the performance boost, it might be worthwhile taking a minuscule quality hit.
sheepscreek 18 hours ago 0.1% is within margin of error. Depending on the performance boost, it might be worthwhile taking a minuscule quality hit.
electroglyph 21 hours ago any divergence (even if the benchmark is better) from full precision is error 7e 14 hours ago Just pretend that it is the next step update when training. You didn’t train your model to step=inf, I hope?
7e 14 hours ago Just pretend that it is the next step update when training. You didn’t train your model to step=inf, I hope?
It's not better quality: 59.3% vs 59.4% fp16 on AIME 25
0.1% is within margin of error. Depending on the performance boost, it might be worthwhile taking a minuscule quality hit.
any divergence (even if the benchmark is better) from full precision is error
Just pretend that it is the next step update when training. You didn’t train your model to step=inf, I hope?
Faster than Fp16, not better quality i guess
[dead]