mirror of
https://github.com/michaelthomson0797/fleet-infra.git
synced 2026-09-19 14:28:29 +00:00
qwen to gemma
This commit is contained in:
+15
-13
@@ -21,19 +21,21 @@ spec:
|
||||
type: nvidia
|
||||
number: 1
|
||||
models:
|
||||
create:
|
||||
- name: qwen3.5:9b-custom
|
||||
template: |
|
||||
FROM qwen3.5:9b
|
||||
PARAMETER num_ctx 8192
|
||||
PARAMETER temperature 1.0
|
||||
PARAMETER top_p 0.95
|
||||
PARAMETER top_k 20
|
||||
PARAMETER min_p 0.0
|
||||
PARAMETER presence_penalty 1.5
|
||||
PARAMETER repetition_penalty 1.0
|
||||
run:
|
||||
- qwen3.5:9b-custom
|
||||
pull:
|
||||
- gemma4:12b
|
||||
# create:
|
||||
# - name: qwen3.5:9b-custom
|
||||
# template: |
|
||||
# FROM qwen3.5:9b
|
||||
# PARAMETER num_ctx 8192
|
||||
# PARAMETER temperature 1.0
|
||||
# PARAMETER top_p 0.95
|
||||
# PARAMETER top_k 20
|
||||
# PARAMETER min_p 0.0
|
||||
# PARAMETER presence_penalty 1.5
|
||||
# PARAMETER repetition_penalty 1.0
|
||||
# run:
|
||||
# - qwen3.5:9b-custom
|
||||
ingress:
|
||||
enabled: true
|
||||
annotations:
|
||||
|
||||
Reference in New Issue
Block a user