• 13 Posts
  • 557 Comments
Joined 7 months ago
cake
Cake day: March 22nd, 2024

help-circle






  • BG3 would beg to differ, though that’s quite an exception.

    And honestly… A lot of games are great if you just wait to play them. I played Cyberpunk 2077 way after release, with mods to fix bugs, make combat more punchy and some other utilities, and it was an absolute blast. Stellar writing, the immersion… nothing like what I read about release, nothing like I’ve played before.

    That’s my most dramatic example, but it’s happened to me in other AAAs too.






  • Wouldn’t Turkey or someone sour this?

    But if it’s actually possible, that’s fascinating… if Ukraine can’t push back quickly, wouldn’t it “force” an end to the war? Russia would have a red line it absolutely can’t cross, no hope of advancement, and likely just claim everything on the other side. Surely they wouldn’t continue a grinding stalemate where Ukraine has a “safe zone” to operate out of.

    If Ukraine does retain its ability to push back hard by the time this happens, and doesn’t go for a truce, then that’s especially peculiar. Walling off a part of their territory as actually untouchable seems like a massive strategic advantage for Ukraine.




  • Well, it’s not over.

    This is coming next week. Path is unclear, and its not as big as Helene, but anything near a 930mb in Tampa Bay and plowing over Orlando at 950mb, especially at this angle, is a catastrophe.

    Katrina was 920mb at landfall, and these intensity forecasts have been undershooting hurricanes recently.

    And there’s another low pressure system at the edge of the GFS that I don’t like, taking a similar path to Helene:

    This is what the upcoming hurricane looked like a few days ago.


  • Pretty much everything has an API :P

    ollama is OK because its easy and automated, but you can get higher performance, better vram efficiency, and better samplers from either kobold.cpp or tabbyAPI, with the catch being that more manual configuration is required. But this is good, as it “forces” you to pick and test an optimal config for your system.

    I’d recommend kobold.cpp for very short context (like 6K or less) or if you need to partially offload the model to CPU because your GPU is relatively low VRAM. Use a good IQ quantization (like IQ4_M, for instance).

    Otherwise use TabbyAPI with an exl2 quantization, as it’s generally faster (but GPU only) and much better at long context through its great k/v cache quantization.

    They all have OpenAI APIs, though kobold.cpp also has its own web ui.