shorten context for qwen3-vl-2b and lower kv cache quant
This commit is contained in:
@@ -535,8 +535,9 @@ models:
|
|||||||
cmd: |
|
cmd: |
|
||||||
/app/llama-server
|
/app/llama-server
|
||||||
-hf unsloth/Qwen3-VL-2B-Instruct-GGUF:Q4_K_M
|
-hf unsloth/Qwen3-VL-2B-Instruct-GGUF:Q4_K_M
|
||||||
--ctx-size 16384
|
--ctx-size 6144
|
||||||
--predict 4096
|
--cache-type-k q8_0
|
||||||
|
--cache-type-v q8_0
|
||||||
--temp 0.7
|
--temp 0.7
|
||||||
--top-p 0.8
|
--top-p 0.8
|
||||||
--top-k 20
|
--top-k 20
|
||||||
|
|||||||
Reference in New Issue
Block a user