Compare commits
4 Commits
c06fac8212
...
af9b5b82f9
| Author | SHA1 | Date | |
|---|---|---|---|
| af9b5b82f9 | |||
| ce0b13ebb3 | |||
| 516e157d39 | |||
| 73d6d1f15a |
@@ -8,8 +8,8 @@ macros:
|
|||||||
gemma3_ctx_128k: "--ctx-size 131072"
|
gemma3_ctx_128k: "--ctx-size 131072"
|
||||||
qwen35_ctx_256k: "--ctx-size 262144"
|
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 q8_0 -ctv q8_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 q8_0 -ctv q8_0"
|
||||||
qwen35_35b_heretic_mmproj: "--mmproj-url https://huggingface.co/unsloth/Qwen3.5-35B-A3B-GGUF/resolve/main/mmproj-F16.gguf --mmproj /root/.cache/llama.cpp/unsloth_Qwen3.5-35B-A3B-GGUF_mmproj-F16.gguf"
|
qwen35_35b_heretic_mmproj: "--mmproj-url https://huggingface.co/unsloth/Qwen3.5-35B-A3B-GGUF/resolve/main/mmproj-F16.gguf --mmproj /root/.cache/llama.cpp/unsloth_Qwen3.5-35B-A3B-GGUF_mmproj-F16.gguf"
|
||||||
qwen35_4b_heretic_mmproj: "--mmproj-url https://huggingface.co/unsloth/Qwen3.5-4B-GGUF/resolve/main/mmproj-F16.gguf --mmproj /root/.cache/llama.cpp/unsloth_Qwen3.5-4B-GGUF_mmproj-F16.gguf"
|
qwen35_4b_heretic_mmproj: "--mmproj-url https://huggingface.co/unsloth/Qwen3.5-4B-GGUF/resolve/main/mmproj-F16.gguf --mmproj /root/.cache/llama.cpp/unsloth_Qwen3.5-4B-GGUF_mmproj-F16.gguf"
|
||||||
glm47_flash_args: "--temp 0.7 --top-p 1.0 --min-p 0.01 --repeat-penalty 1.0"
|
glm47_flash_args: "--temp 0.7 --top-p 1.0 --min-p 0.01 --repeat-penalty 1.0"
|
||||||
@@ -82,7 +82,7 @@ models:
|
|||||||
--top-p 0.95
|
--top-p 0.95
|
||||||
--top-k 40
|
--top-k 40
|
||||||
--repeat-penalty 1.0
|
--repeat-penalty 1.0
|
||||||
-ctk q4_0 -ctv q4_0
|
-ctk q8_0 -ctv q8_0
|
||||||
${common_args}
|
${common_args}
|
||||||
|
|
||||||
"Qwen3.5-35B-A3B-GGUF:Q4_K_M":
|
"Qwen3.5-35B-A3B-GGUF:Q4_K_M":
|
||||||
|
|||||||
@@ -18,7 +18,7 @@ spec:
|
|||||||
spec:
|
spec:
|
||||||
containers:
|
containers:
|
||||||
- name: llama-swap
|
- name: llama-swap
|
||||||
image: ghcr.io/mostlygeek/llama-swap:v198-vulkan-b8390
|
image: ghcr.io/mostlygeek/llama-swap:v198-vulkan-b8400
|
||||||
imagePullPolicy: IfNotPresent
|
imagePullPolicy: IfNotPresent
|
||||||
command:
|
command:
|
||||||
- /app/llama-swap
|
- /app/llama-swap
|
||||||
|
|||||||
@@ -15,7 +15,7 @@ spec:
|
|||||||
- name: renovate
|
- name: renovate
|
||||||
# Update this to the latest available and then enable Renovate on
|
# Update this to the latest available and then enable Renovate on
|
||||||
# the manifest
|
# the manifest
|
||||||
image: renovate/renovate:43.77.9-full
|
image: renovate/renovate:43.83.1-full
|
||||||
envFrom:
|
envFrom:
|
||||||
- secretRef:
|
- secretRef:
|
||||||
name: renovate-gitea-token
|
name: renovate-gitea-token
|
||||||
|
|||||||
Reference in New Issue
Block a user