yjtpesesu2

3 ms·

1. 1.▲
   by yjtpesesu2 7mo ago

   so, magic?
2. 2.▲
   by yjtpesesu2 7mo ago

   How does this differ from anything llama.cpp offers, regarding offloading layers? The repo consistently refers to "DDR4". Is there a reason DDR5 won't work with this?
3. 3.▲
   by yjtpesesu2 7mo ago

   Oh, that's just the infra for the infra. Then use something like graphllm from matteo, and of course llama.cpp from greg, tailor you model selection to your hardware.
4. 4.▲
   Hacker Infrastructure

   1 points by yjtpesesu2 7mo ago|1 comments
5. 5.▲
   by yjtpesesu2 9mo ago

   Death approaches. Slow burn until. When Death arrives, what you are doing now will be obviously irrelevant.
6. 6.▲
   by yjtpesesu2 9mo ago

   Any clue why you might be a favored/favoured high value user?