ray/examples/config/ray.1b.8gb.public.json at main · razroo/ray · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
{
  "profile": "1b",
  "asyncQueue": {
    "enabled": true,
    "storageDir": "/var/lib/ray/async-queue"
  },
  "auth": {
    "enabled": true,
    "apiKeyEnv": "RAY_API_KEYS"
  },
  "server": {
    "requestBodyLimitBytes": 64000
  },
  "model": {
    "maxOutputTokens": 256,
    "operational": {
      "recommendedPromptFormat": "native-template",
      "supportsJsonMode": true,
      "tokensPerSecondTarget": 18,
      "memoryClassMiB": 8192,
      "preferredCtxSize": 4096,
      "chatTemplateKnown": true
    },
    "adapter": {
      "kind": "llama.cpp",
      "timeoutMs": 24000,
      "slotStateTtlMs": 250,
      "slotSnapshotTimeoutMs": 300,
      "promptScaffoldCacheEntries": 384,
      "launchProfile": {
        "preset": "single-vps-1b-8gb",
        "binaryPath": "/usr/local/bin/llama-server",
        "modelPath": "/var/lib/ray/models/qwen2.5-1.5b-instruct-q4_k_m.gguf",
        "host": "127.0.0.1",
        "port": 8081,
        "alias": "qwen2.5-1.5b-instruct-q4_k_m",
        "ctxSize": 4096,
        "parallel": 2,
        "threads": 4,
        "threadsHttp": 2,
        "batchSize": 256,
        "ubatchSize": 128,
        "cachePrompt": true,
        "cacheReuse": 256,
        "cacheRamMiB": 768,
        "continuousBatching": true,
        "enableMetrics": true,
        "exposeSlots": true,
        "warmup": true,
        "enableUnifiedKv": true,
        "cacheIdleSlots": true,
        "contextShift": true
      }
    }
  },
  "scheduler": {
    "concurrency": 2,
    "maxQueue": 96,
    "maxQueuedTokens": 48000,
    "maxInflightTokens": 6144,
    "requestTimeoutMs": 28000,
    "dedupeInflight": true,
    "batchWindowMs": 10,
    "affinityLookahead": 24,
    "shortJobMaxTokens": 96
  },
  "cache": {
    "maxEntries": 512,
    "ttlMs": 120000
  },
  "gracefulDegradation": {
    "queueDepthThreshold": 20,
    "maxPromptChars": 8000,
    "degradeToMaxTokens": 160
  },
  "rateLimit": {
    "maxRequests": 150,
    "maxKeys": 8192
  },
  "tags": {
    "hardware": "8gb-vps-class",
    "engine": "llama.cpp",
    "modelSize": "1b-class",
    "exposure": "public"
  }
}