[*] n-gpu-layers = 999 fit = off cache-type-k = q8_0 cache-type-v = q8_0 repeat-penalty = 1.15 [Qwen3.8-6-Loop-9B-Heretic-Q4_K_S] model = ./models/Qwen3.8-9B-Distill-uncensored-heretic.Q4_K_S.gguf # Optional vision projector (kept in CPU RAM via mmproj-offload = false): # mmproj = ./models/Qwen3.8-9B-Distill-uncensored-heretic.mmproj-f16.gguf # mmproj-offload = false lora = ./qwen3.8-9b-heretic-loop6-lora.gguf override-kv = qwen35.num_loops=int:6,qwen35.recurrent_kv_cache_mode=str:per_loop ctx-size = 8192 parallel = 1 cache-ram = 1024 ctx-checkpoints = 2 [Qwen3.8-4-Loop-9B-Heretic-Q4_K_S] model = ./models/Qwen3.8-9B-Distill-uncensored-heretic.Q4_K_S.gguf lora = ./qwen3.8-9b-heretic-loop6-lora.gguf override-kv = qwen35.num_loops=int:4,qwen35.recurrent_kv_cache_mode=str:per_loop ctx-size = 8192 parallel = 1 cache-ram = 1024 ctx-checkpoints = 2 [Qwen3.8-2-Loop-9B-Heretic-Q4_K_S] model = ./models/Qwen3.8-9B-Distill-uncensored-heretic.Q4_K_S.gguf lora = ./qwen3.8-9b-heretic-loop6-lora.gguf override-kv = qwen35.num_loops=int:2,qwen35.recurrent_kv_cache_mode=str:per_loop ctx-size = 8192 parallel = 1 cache-ram = 1024 ctx-checkpoints = 2 [Qwen3.8-1-Loop-9B-Heretic-Q4_K_S] model = ./models/Qwen3.8-9B-Distill-uncensored-heretic.Q4_K_S.gguf lora = ./qwen3.8-9b-heretic-loop6-lora.gguf override-kv = qwen35.num_loops=int:1,qwen35.recurrent_kv_cache_mode=str:per_loop ctx-size = 8192 parallel = 1 cache-ram = 1024 ctx-checkpoints = 2