Codette/mainpreventchema.json at main · Raiffs-bits/Codette · 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
{
  "name": "prevent_model_collapse",
  "description": "Preemptive safeguard against model collapse, ensuring consistent learning and retention of data integrity across model generations.",
  "strict": true,
  "parameters": {
    "type": "object",
    "required": [
      "initial_data",
      "training_steps",
      "model_capacities",
      "sampling_method"
    ],
    "properties": {
      "initial_data": {
        "type": "array",
        "description": "The initial clean data used for training the first model (model 0).",
        "items": {
          "type": "object",
          "properties": {
            "text": {
              "type": "string",
              "description": "Text data sample that will be used for training."
            },
            "label": {
              "type": "string",
              "description": "Label associated with the data sample."
            }
          },
          "additionalProperties": false,
          "required": [
            "text",
            "label"
          ]
        }
      },
      "training_steps": {
        "type": "number",
        "description": "Number of iterations for training the models to evaluate convergence and performance."
      },
      "model_capacities": {
        "type": "object",
        "required": [
          "max_samples",
          "memory_limit"
        ],
        "properties": {
          "max_samples": {
            "type": "number",
            "description": "Maximum number of samples to retain for each model's training dataset."
          },
          "memory_limit": {
            "type": "number",
            "description": "Memory limit for training each individual model."
          }
        },
        "additionalProperties": false
      },
      "sampling_method": {
        "type": "string",
        "description": "Method used for data sampling during each training phase.",
        "enum": [
          "Monte_Carlo",
          "stratified",
          "random"
        ]
      }
    },
    "additionalProperties": false
  }
}