In June we had successful training runs against Tailstorm(ll) discount on an alpha schedule list. We're not able to reproduce this against the new Tailstorm. To rule out any problems we want to reproduce the June results with the new simulator. This PR re-adds the old protocol in the new simulator.
As mentioned at the top of tailstorm_june.ml, the diff to tailstomll.ml is very small. Even if we keep the attack space tailstorm_june_ssz.ml for future reference, we probably still want to get rid of the protocol tailstorm_june.
The june version of the protocol does not support observation normalization to the unit interval. This PR makes observation normalization optional.
In June we had successful training runs against Tailstorm(ll) discount on an alpha schedule list. We're not able to reproduce this against the new Tailstorm. To rule out any problems we want to reproduce the June results with the new simulator. This PR re-adds the old protocol in the new simulator.
As mentioned at the top of tailstorm_june.ml, the diff to tailstomll.ml is very small. Even if we keep the attack space tailstorm_june_ssz.ml for future reference, we probably still want to get rid of the protocol tailstorm_june.
The june version of the protocol does not support observation normalization to the unit interval. This PR makes observation normalization optional.