15 points | by alister 10 hours ago
2 comments
Sounds like an interesting way to benchmark models. Put two instances of a model in a loop with itself. Run for n cycles. Then look at the output.
You can clearly see the side effects of the models being trained on Reddit content.
Sounds like an interesting way to benchmark models. Put two instances of a model in a loop with itself. Run for n cycles. Then look at the output.
You can clearly see the side effects of the models being trained on Reddit content.