4 ms·
You must not be doing much real work with these models, otherwise you would know that with the latest Claude models you can't even adjust the temperature or top
by weird-eye-issue 2mo ago
You must not be doing much real work with these models, otherwise you would know that with the latest Claude models you can't even adjust the temperature or top p/k
The real way to actually get a good output has always been in the prompt, not these parameters
- BoredomIsFun 2mo ago> You must not be doing much real work with these models, otherwise you would know that with the latest Claude models you can't even adjust the temperature or top p/k No, I do not. Good for me I guess. > The real way to actually get a good output has always been in the prompt, not these parameters What an absurd claim.
- weird-eye-issue 2mo agoHave fun toying around with your local models and leave the real work to the rest of us buddy
- BoredomIsFun 2mo agoHave fun using web-interfaces for overpriced US models while I am getting my work done with Chinese models for a fraction of price on openrouter using API. Besides, when using API I am sure all American models, including Claude honour sampling settings.
- weird-eye-issue 2mo agoNo, they don't. That's the whole point. I've spent several hundred thousand dollars in API costs in the last few years to power my app, I think I know what I'm talking about. Nobody besides you is talking about web interfaces Meanwhile you clearly have zero clue what you are talking about because a simple Google search would show you otherwise but that must be too difficult for you
- BoredomIsFun 2mo ago> I've spent several hundred thousand dollars in API costs in the last few years to power my app Which is kinda sad, if you've used Anthropic products, because you could get comparable performance from cheaper models for vast majority of tasks. > No, they don't. That's the whole point. Bad for them; do not use Anthropic then. Besides "the whole point" of conversation you have interrupted is that "low temperature and tight sampling produces stiff boring cliche prose" - which is truism, as those settings control the entropy of the output. And it is utterly irrelevant frankly if one has access to the Claude sampler through API or not; as Anthropic has apparently locked the sampler at very conservative temperature (probably as low as 0.2), there is no way to squeeze god prose out of it, as the logits have been severely messed up wrt to the actual word distribution in standard English.
- weird-eye-issue 2mo agoYou are dumber than a rock
- BoredomIsFun 2mo agoYou are dumber thanme
- Der_Einzige 2mo agoWrong: https://gist.github.com/Hellisotherpeople/71ba712f9f899adcb08b94bce20d5397 https://gist.github.com/Hellisotherpeople/71ba712f9f899adcb0...
- weird-eye-issue 2mo agoNice slop
- BoredomIsFun 2mo agoYou are flaunting your inability/unwillingness to use sampler settings.
- weird-eye-issue 2mo agoWhat? They aren't even available for the new Claude models which we use extensively so I think that should tell you something. I've made over $6 million with my AI app in the last few years without worrying about temperature and other parameters. What have you done?
- BoredomIsFun 2mo ago> I've made over $6 million with my AI app in the last few years Oh, wow. > without worrying about temperature and other parameters. Tells a lot about you.