@geek-fun/serverlessinsight 0.3.0 → 0.3.1

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
@@ -3,17 +3,6 @@ provider:
3
3
  name: aliyun
4
4
  region: cn-hangzhou
5
5
 
6
- vars:
7
- testv: testVarValue
8
- handler: index.handler
9
-
10
- stages:
11
- default:
12
- node_env: default
13
- dev:
14
- node_env: development
15
- prod:
16
- region: cn-hangzhou
17
6
 
18
7
  service: insight-poc-gpu
19
8
 
@@ -24,17 +13,18 @@ functions:
24
13
  insight_poc_fn:
25
14
  name: insight-poc-gpu-fn
26
15
  container:
27
- image: registry.cn-chengdu.aliyuncs.com/geek-fun/meke-api:latest
28
- cmd: "npm start"
16
+ image: registry.cn-hangzhou.aliyuncs.com/serverless_devs/modelscope:ubuntu20.04-cuda11.3.0-py37-torch1.11.0-tf1.15.5-1.5.0
17
+ cmd: "ollama run modelscope.cn/DeepSeek-R1-Distill-Qwen-1.5B-GGUF"
29
18
  port: 9000
30
- memory: 512
31
- gpu: TESLA_8
32
- timeout: 10
19
+ memory: 4096
20
+ gpu: TESLA_16
21
+ timeout: 900
22
+ log: true
33
23
  network:
34
- vpc_id: vpc-2vc8v9btc8470laqui9bk
24
+ vpc_id: vpc-bp1jpzq0807mrww7i8uko
35
25
  subnet_ids:
36
- - vsw-2vc9zrs5mojkxd14yo3zw
37
- - vsw-2vceshdo0xjp9q9t0oyt0
26
+ - vsw-bp1tp32pxn8c3mm07e9s2
27
+ - vsw-bp1gs2ggjq85apbf5yd9d
38
28
  security_group:
39
29
  name: insight-poc-gpu-fn-sg
40
30
  ingress:
@@ -49,6 +39,17 @@ functions:
49
39
  - mount_path: /mnt/nas
50
40
  storage_class: STANDARD_CAPACITY
51
41
  environment:
52
- NODE_ENV: ${stages.node_env}
53
- TEST_VAR: ${vars.testv}
54
- TEST_VAR_EXTRA: abcds-${vars.testv}-andyou
42
+ OLLAMA_MODELS: /mnt/nas/models
43
+ LLAMA_GPU_OVERHEAD: 0
44
+ OLLAMA_HOST: http://0.0.0.0:11434
45
+ OLLAMA_INTEL_GPU: false
46
+ OLLAMA_KEEP_ALIVE: 2562047h47m16.854775807s
47
+ OLLAMA_LOAD_TIMEOUT: 5m0s
48
+ OLLAMA_MAX_LOADED_MODELS: 1
49
+ OLLAMA_MAX_QUEUE: 1
50
+ OLLAMA_MULTIUSER_CACHE: false
51
+ OLLAMA_NOHISTORY: false
52
+ OLLAMA_NOPRUNE: false
53
+ OLLAMA_NUM_PARALLEL: 1
54
+ OLLAMA_ORIGINS: [ * http://localhost https://localhost http://localhost:* https://localhost:* http://127.0.0.1 https://127.0.0.1 http://127.0.0.1:* https://127.0.0.1:* http://0.0.0.0 https://0.0.0.0 http://0.0.0.0:* https://0.0.0.0:* app://* file://* tauri://* vscode-webview://* ]
55
+ OLLAMA_SCHED_SPREAD: fals