Guest User

Untitled

a guest
Jun 13th, 2025
40
0
266 days
Not a member of Pastebin yet? Sign Up, it unlocks many cool features!
Batch 0.42 KB | None | 0 0
  1. @echo off
  2. set CUDA_VISIBLE_DEVICES=0,1
  3. llama-server.exe ^
  4. -m "T:\models\DeepSeek-V3-0324-IQ1_S_R4-00001-of-00003.gguf" ^
  5. --n-gpu-layers 999 ^
  6. -ts 23,23 ^
  7. --threads 18 ^
  8. --threads-batch 18 ^
  9. --ctx-size 16384 ^
  10. --batch-size 4096 ^
  11. --ubatch-size 4096 ^
  12. --no-mmap ^
  13. -amb 512 ^
  14. -mla 3 ^
  15. -fa ^
  16. -fmoe ^
  17. -rtr ^
  18. -ot "blk\.(0|1|2|3|4|5|6|7|8)\..*exps=CUDA0" ^
  19. -ot "blk\.(9|10|11|12|13|14)\..*exps=CUDA1" ^
  20. -ot "exps=CPU"
Advertisement
Add Comment
Please, Sign In to add comment