@@ -60,20 +60,67 @@ The sidecar supports the following command-line arguments:
60
60
61
61
# # Configuration Fields
62
62
- ` vLLMLoRAConfig` [**required**] base key
63
- - ` host` [*optional*]Model server's host. defaults to localhost
63
+ - ` host` [*optional*] Model server's host. defaults to localhost
64
64
- ` port` [*optional*] Model server's port. defaults to 8000
65
- - ` name` [*optional*] Name of this config
66
- - ` ensureExist` [*optional*] List of models to ensure existence on specified model server.
67
- - ` models` [**required**] [list]
68
- - ` base-model` [*optional*] Base model for lora adapter
69
- - ` id` [**required**] unique id of lora adapter
70
- - ` source` [**required**] path (remote or local) to lora adapter
65
+ - ` name` [*optional*] Name of this config
66
+ - ` defaultBaseModel` [*optional*] Default base model to use for all adapters when not specified individually
67
+ - ` ensureExist` [*optional*] List of models to ensure existence on specified model server.
68
+ - ` models` [**required**] [list]
69
+ - ` id` [**required**] unique id of lora adapter
70
+ - ` source` [**required**] path (remote or local) to lora adapter
71
+ - ` base-model` [*optional*] Base model for lora adapter (overrides defaultBaseModel)
71
72
- ` ensureNotExist` [*optional*]
72
- - ` models` [**required**] [list]
73
- - ` id` [**required**] unique id of lora adapter
74
- - ` source` [**required**] path (remote or local) to lora adapter
75
- - ` base-model` [*optional*] Base model for lora adapter
73
+ - ` models` [**required**] [list]
74
+ - ` id` [**required**] unique id of lora adapter
75
+ - ` source` [**required**] path (remote or local) to lora adapter
76
+ - ` base-model` [*optional*] Base model for lora adapter (overrides defaultBaseModel)
76
77
78
+ # # Example Configuration
79
+
80
+ Here's an example of using the `defaultBaseModel` field to avoid repetition in your configuration :
81
+
82
+ ` ` ` yaml
83
+ apiVersion: v1
84
+ kind: ConfigMap
85
+ metadata:
86
+ name: vllm-llama2-7b-adapters
87
+ data:
88
+ configmap.yaml: |
89
+ vLLMLoRAConfig:
90
+ name: vllm-llama2-7b
91
+ port: 8000
92
+ defaultBaseModel: meta-llama/Llama-2-7b-hf
93
+ ensureExist:
94
+ models:
95
+ - id: tweet-summary-1
96
+ source: vineetsharma/qlora-adapter-Llama-2-7b-hf-TweetSumm
97
+ - id: tweet-summary-2
98
+ source: mahimairaja/tweet-summarization-llama-2-finetuned
99
+ ` ` `
100
+
101
+ In this example, both adapters will use `meta-llama/Llama-2-7b-hf` as their base model without needing to specify it for each adapter individually.
102
+
103
+ You can still override the default base model for specific adapters when needed :
104
+
105
+ ` ` ` yaml
106
+ apiVersion: v1
107
+ kind: ConfigMap
108
+ metadata:
109
+ name: vllm-mixed-adapters
110
+ data:
111
+ configmap.yaml: |
112
+ vLLMLoRAConfig:
113
+ name: vllm-mixed
114
+ port: 8000
115
+ defaultBaseModel: meta-llama/Llama-2-7b-hf
116
+ ensureExist:
117
+ models:
118
+ - id: tweet-summary-1
119
+ source: vineetsharma/qlora-adapter-Llama-2-7b-hf-TweetSumm
120
+ - id: code-assistant
121
+ source: huggingface/code-assistant-lora
122
+ base-model: meta-llama/Llama-2-13b-hf # Override for this specific adapter
123
+ ` ` `
77
124
# # Example Deployment
78
125
79
126
The [deployment.yaml](deployment.yaml) file shows an example of deploying the sidecar with custom parameters :
0 commit comments