Lumpiasty/klaster
0
0
Fork 0
Code Issues Pull Requests 4 Packages Projects Releases Wiki Activity

add max ctx on llama.cpp

Browse Source
This commit is contained in:
Lumpiasty
2026-03-17 01:33:35 +01:00
parent 9f315b38e3
commit 7e7b3e3d71
1 changed files with 24 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files

25
apps/llama/configs/config.yaml
View File

@@ -4,7 +4,9 @@ logToStdout: "both" # proxy and upstream
macros: macros:
base_args: "--no-warmup --port ${PORT}" base_args: "--no-warmup --port ${PORT}"
common_args: "--fit-target 1536 --fit-ctx 65536 --no-warmup --port ${PORT}" common_args: "--fit-target 1024 --no-warmup --port ${PORT}"
gemma3_ctx_128k: "--ctx-size 131072"
qwen35_ctx_256k: "--ctx-size 262144"
gemma_sampling: "--prio 2 --temp 1.0 --repeat-penalty 1.0 --min-p 0.00 --top-k 64 --top-p 0.95" gemma_sampling: "--prio 2 --temp 1.0 --repeat-penalty 1.0 --min-p 0.00 --top-k 64 --top-p 0.95"
qwen35_sampling: "--temp 0.6 --top-p 0.95 --top-k 20 --min-p 0.00 -ctk q4_0 -ctv q4_0" qwen35_sampling: "--temp 0.6 --top-p 0.95 --top-k 20 --min-p 0.00 -ctk q4_0 -ctv q4_0"
qwen35_35b_args: "--temp 1.0 --min-p 0.00 --top-p 0.95 --top-k 20 -ctk q4_0 -ctv q4_0" qwen35_35b_args: "--temp 1.0 --min-p 0.00 --top-p 0.95 --top-k 20 -ctk q4_0 -ctv q4_0"
@@ -39,6 +41,7 @@ models:
cmd: | cmd: |
/app/llama-server /app/llama-server
-hf unsloth/gemma-3-12b-it-GGUF:Q4_K_M -hf unsloth/gemma-3-12b-it-GGUF:Q4_K_M
${gemma3_ctx_128k}
${gemma_sampling} ${gemma_sampling}
${common_args} ${common_args}
@@ -46,6 +49,7 @@ models:
cmd: | cmd: |
/app/llama-server /app/llama-server
-hf unsloth/gemma-3-12b-it-GGUF:Q4_K_M -hf unsloth/gemma-3-12b-it-GGUF:Q4_K_M
${gemma3_ctx_128k}
${gemma_sampling} ${gemma_sampling}
--no-mmproj --no-mmproj
${common_args} ${common_args}
@@ -54,6 +58,7 @@ models:
cmd: | cmd: |
/app/llama-server /app/llama-server
-hf unsloth/gemma-3-4b-it-GGUF:Q4_K_M -hf unsloth/gemma-3-4b-it-GGUF:Q4_K_M
${gemma3_ctx_128k}
${gemma_sampling} ${gemma_sampling}
${common_args} ${common_args}
@@ -61,6 +66,7 @@ models:
cmd: | cmd: |
/app/llama-server /app/llama-server
-hf unsloth/gemma-3-4b-it-GGUF:Q4_K_M -hf unsloth/gemma-3-4b-it-GGUF:Q4_K_M
${gemma3_ctx_128k}
${gemma_sampling} ${gemma_sampling}
--no-mmproj --no-mmproj
${common_args} ${common_args}
@@ -83,6 +89,7 @@ models:
cmd: | cmd: |
/app/llama-server /app/llama-server
-hf unsloth/Qwen3.5-35B-A3B-GGUF:Q4_K_M -hf unsloth/Qwen3.5-35B-A3B-GGUF:Q4_K_M
${qwen35_ctx_256k}
${qwen35_35b_args} ${qwen35_35b_args}
${common_args} ${common_args}
@@ -90,6 +97,7 @@ models:
cmd: | cmd: |
/app/llama-server /app/llama-server
-hf unsloth/Qwen3.5-35B-A3B-GGUF:Q4_K_M -hf unsloth/Qwen3.5-35B-A3B-GGUF:Q4_K_M
${qwen35_ctx_256k}
${qwen35_35b_args} ${qwen35_35b_args}
${common_args} ${common_args}
${thinking_off} ${thinking_off}
@@ -101,6 +109,7 @@ models:
/app/llama-server /app/llama-server
-hf mradermacher/Qwen3.5-35B-A3B-heretic-GGUF:Q4_K_M -hf mradermacher/Qwen3.5-35B-A3B-heretic-GGUF:Q4_K_M
${qwen35_35b_heretic_mmproj} ${qwen35_35b_heretic_mmproj}
${qwen35_ctx_256k}
${qwen35_35b_args} ${qwen35_35b_args}
${common_args} ${common_args}
@@ -109,6 +118,7 @@ models:
/app/llama-server /app/llama-server
-hf mradermacher/Qwen3.5-35B-A3B-heretic-GGUF:Q4_K_M -hf mradermacher/Qwen3.5-35B-A3B-heretic-GGUF:Q4_K_M
${qwen35_35b_heretic_mmproj} ${qwen35_35b_heretic_mmproj}
${qwen35_ctx_256k}
${qwen35_35b_args} ${qwen35_35b_args}
${common_args} ${common_args}
${thinking_off} ${thinking_off}
@@ -117,6 +127,7 @@ models:
cmd: | cmd: |
/app/llama-server /app/llama-server
-hf unsloth/Qwen3.5-0.8B-GGUF:Q4_K_XL -hf unsloth/Qwen3.5-0.8B-GGUF:Q4_K_XL
${qwen35_ctx_256k}
${qwen35_sampling} ${qwen35_sampling}
${base_args} ${base_args}
${thinking_on} ${thinking_on}
@@ -134,6 +145,7 @@ models:
cmd: | cmd: |
/app/llama-server /app/llama-server
-hf unsloth/Qwen3.5-2B-GGUF:Q4_K_M -hf unsloth/Qwen3.5-2B-GGUF:Q4_K_M
${qwen35_ctx_256k}
${qwen35_sampling} ${qwen35_sampling}
${common_args} ${common_args}
${thinking_on} ${thinking_on}
@@ -142,6 +154,7 @@ models:
cmd: | cmd: |
/app/llama-server /app/llama-server
-hf unsloth/Qwen3.5-2B-GGUF:Q4_K_M -hf unsloth/Qwen3.5-2B-GGUF:Q4_K_M
${qwen35_ctx_256k}
${qwen35_sampling} ${qwen35_sampling}
${common_args} ${common_args}
${thinking_off} ${thinking_off}
@@ -150,6 +163,7 @@ models:
cmd: | cmd: |
/app/llama-server /app/llama-server
-hf unsloth/Qwen3.5-4B-GGUF:Q4_K_M -hf unsloth/Qwen3.5-4B-GGUF:Q4_K_M
${qwen35_ctx_256k}
${qwen35_sampling} ${qwen35_sampling}
${common_args} ${common_args}
${thinking_on} ${thinking_on}
@@ -158,6 +172,7 @@ models:
cmd: | cmd: |
/app/llama-server /app/llama-server
-hf unsloth/Qwen3.5-4B-GGUF:Q4_K_M -hf unsloth/Qwen3.5-4B-GGUF:Q4_K_M
${qwen35_ctx_256k}
${qwen35_sampling} ${qwen35_sampling}
${common_args} ${common_args}
${thinking_off} ${thinking_off}
@@ -167,6 +182,7 @@ models:
/app/llama-server /app/llama-server
-hf mradermacher/Qwen3.5-4B-heretic-GGUF:Q4_K_M -hf mradermacher/Qwen3.5-4B-heretic-GGUF:Q4_K_M
${qwen35_4b_heretic_mmproj} ${qwen35_4b_heretic_mmproj}
${qwen35_ctx_256k}
${qwen35_sampling} ${qwen35_sampling}
${common_args} ${common_args}
${thinking_on} ${thinking_on}
@@ -176,6 +192,7 @@ models:
/app/llama-server /app/llama-server
-hf mradermacher/Qwen3.5-4B-heretic-GGUF:Q4_K_M -hf mradermacher/Qwen3.5-4B-heretic-GGUF:Q4_K_M
${qwen35_4b_heretic_mmproj} ${qwen35_4b_heretic_mmproj}
${qwen35_ctx_256k}
${qwen35_sampling} ${qwen35_sampling}
${common_args} ${common_args}
${thinking_off} ${thinking_off}
@@ -184,6 +201,7 @@ models:
cmd: | cmd: |
/app/llama-server /app/llama-server
-hf unsloth/Qwen3.5-9B-GGUF:Q4_K_M -hf unsloth/Qwen3.5-9B-GGUF:Q4_K_M
${qwen35_ctx_256k}
${qwen35_sampling} ${qwen35_sampling}
${common_args} ${common_args}
${thinking_on} ${thinking_on}
@@ -192,6 +210,7 @@ models:
cmd: | cmd: |
/app/llama-server /app/llama-server
-hf unsloth/Qwen3.5-9B-GGUF:Q4_K_M -hf unsloth/Qwen3.5-9B-GGUF:Q4_K_M
${qwen35_ctx_256k}
${qwen35_sampling} ${qwen35_sampling}
${common_args} ${common_args}
${thinking_off} ${thinking_off}
@@ -200,6 +219,7 @@ models:
cmd: | cmd: |
/app/llama-server /app/llama-server
-hf unsloth/Qwen3.5-9B-GGUF:Q3_K_M -hf unsloth/Qwen3.5-9B-GGUF:Q3_K_M
${qwen35_ctx_256k}
${qwen35_sampling} ${qwen35_sampling}
${common_args} ${common_args}
${thinking_on} ${thinking_on}
@@ -208,6 +228,7 @@ models:
cmd: | cmd: |
/app/llama-server /app/llama-server
-hf unsloth/Qwen3.5-9B-GGUF:Q3_K_M -hf unsloth/Qwen3.5-9B-GGUF:Q3_K_M
${qwen35_ctx_256k}
${qwen35_sampling} ${qwen35_sampling}
${common_args} ${common_args}
${thinking_off} ${thinking_off}
@@ -216,6 +237,7 @@ models:
cmd: | cmd: |
/app/llama-server /app/llama-server
-hf unsloth/Qwen3.5-27B-GGUF:Q3_K_M -hf unsloth/Qwen3.5-27B-GGUF:Q3_K_M
${qwen35_ctx_256k}
${qwen35_sampling} ${qwen35_sampling}
${common_args} ${common_args}
${thinking_on} ${thinking_on}
@@ -224,6 +246,7 @@ models:
cmd: | cmd: |
/app/llama-server /app/llama-server
-hf unsloth/Qwen3.5-27B-GGUF:Q3_K_M -hf unsloth/Qwen3.5-27B-GGUF:Q3_K_M
${qwen35_ctx_256k}
${qwen35_sampling} ${qwen35_sampling}
${common_args} ${common_args}
${thinking_off} ${thinking_off}