Change the default values for generation in GUI

[pingpongching]

the defaults that we have right now aren't ideal, i tried many models(nous-vicuna, snoozy, groovy....) with the values given here: https://www.reddit.com/r/LocalLLaMA/wiki/index/#wiki_prompting , i used the "Precise" preset, and all of the models gave much better results.
Amount of times the models got stuck in a loop writing the same thing over and over significantly decreased using the new values, that allowed to make them write more longer responses, and follow the conversation for longer, without going schitzo.
And i believe the newcomers will appreciate not having to figure out why the models are constantly getting stuck in loops whenever the conversation gets a little long.

Also changed the batch size from 9 to 128, 9 is too conservative imo.

[AndriyMulyar]

We would need some heavy testing to make these the defaults. Ideally, we should be sending per model defaults by specifying them in the models.json file.

[manyoso]

The current settings were mostly on the advice/direction of some people in discord and some scant testing I did early on. We currently don't have a good way to objectively test default values across our model zoo. I'm all for letting people experimenting and letting the hive mind decide what our defaults should be at this point. @Andriy can you do some kind of poll in discord maybe?