Changxin Miao

Right, this will hurt HDD performance a lot. How about using an environment variable as a switch? for some cases, user can trade off performance for stability/reliability.

Not yet. I didn't intentionally construct test cases for this scenario. Just suppose it's a common practice in visitor pattern when reading the code.

Hi, what do you mean by `work for me`? This web page is always free.

Dears, any progress on the status code? Just FYI, my current fix is just to leave the status code as it is: https://github.com/polyrabbit/go-gin/commit/e7a519105e6f2aec8c1cfc001c43278f043ec007

@brunodoamaral Thanks for reporting. @thiswillbeyourgithub also mentioned the same [issue](https://github.com/polyrabbit/hacker-news-digest/issues/35) and suggested to use `logit_bias` to avoid those words. I gave a quick try but it didn't work as expected....

Oops, I see lots of weird ` 196 interspersely` now. Need to seek another model...

Haven't tried other values - both are set to 1 currently. I suppose we cannot get rid of those magic words completely in Gemma, need to find a better model....

> not stop before banning like 10 tokens I fine-tuned some code and switched to llama3 now, I'll use it for a while and see how it goes. Hope I...

> I'm considering using it as a fallback for dynamic web pages. Done, now we have summaries for substack etc.

@jiefenghuang Sorry I missed your last message. Which part in prefetch_test.go do you think needs to be changed?