Custom Model Settings

1
import data_designer.config as dd
2
from data_designer.interface import DataDesigner
3
4
# Create custom models using default providers
5
custom_models = [
6
    # High-temperature for more variability
7
    dd.ModelConfig(
8
        alias="creative-writer",
9
        model="nvidia/nemotron-3-nano-30b-a3b",
10
        provider="nvidia",  # Uses default NVIDIA provider
11
        inference_parameters=dd.ChatCompletionInferenceParams(
12
            temperature=1.2,
13
            top_p=0.98,
14
            max_tokens=4096,
15
        ),
16
    ),
17
    # Low-temperature for less variability
18
    dd.ModelConfig(
19
        alias="fact-checker",
20
        model="nvidia/nemotron-3-nano-30b-a3b",
21
        provider="nvidia",  # Uses default NVIDIA provider
22
        inference_parameters=dd.ChatCompletionInferenceParams(
23
            temperature=0.1,
24
            top_p=0.9,
25
            max_tokens=2048,
26
        ),
27
    ),
28
]
29
30
# Create DataDesigner (uses default providers)
31
data_designer = DataDesigner()
32
33
# Pass custom models to config builder
34
config_builder = dd.DataDesignerConfigBuilder(model_configs=custom_models)
35
36
# Add a topic column using a categorical sampler
37
config_builder.add_column(
38
    dd.SamplerColumnConfig(
39
        name="topic",
40
        sampler_type=dd.SamplerType.CATEGORY,
41
        params=dd.CategorySamplerParams(
42
            values=["Artificial Intelligence", "Space Exploration", "Ancient History", "Climate Science"],
43
        ),
44
    )
45
)
46
47
# Use your custom models
48
config_builder.add_column(
49
    dd.LLMTextColumnConfig(
50
        name="creative_story",
51
        model_alias="creative-writer",
52
        prompt="Write a creative short story about {{topic}}.",
53
    )
54
)
55
56
config_builder.add_column(
57
    dd.LLMTextColumnConfig(
58
        name="facts",
59
        model_alias="fact-checker",
60
        prompt="List 3 facts about {{topic}}.",
61
    )
62
)
63
64
# Preview your dataset
65
preview_result = data_designer.preview(config_builder=config_builder)
66
preview_result.display_sample_record()

1
import data_designer.config as dd
2
from data_designer.interface import DataDesigner
3
4
# Step 1: Define custom providers
5
custom_providers = [
6
    dd.ModelProvider(
7
        name="my-custom-provider",
8
        endpoint="https://api.my-llm-service.com/v1",
9
        provider_type="openai",  # OpenAI-compatible API
10
        api_key="MY_SERVICE_API_KEY",  # Environment variable name
11
    ),
12
    dd.ModelProvider(
13
        name="my-self-hosted-provider",
14
        endpoint="https://my-org.internal.com/llm/v1",
15
        provider_type="openai",
16
        api_key="SELF_HOSTED_API_KEY",
17
    ),
18
]
19
20
# Step 2: Define custom models
21
custom_models = [
22
    dd.ModelConfig(
23
        alias="my-text-model",
24
        model="openai/some-model-id",
25
        provider="my-custom-provider",  # References provider by name
26
        inference_parameters=dd.ChatCompletionInferenceParams(
27
            temperature=0.85,
28
            top_p=0.95,
29
            max_tokens=2048,
30
        ),
31
    ),
32
    dd.ModelConfig(
33
        alias="my-self-hosted-text-model",
34
        model="openai/some-hosted-model-id",
35
        provider="my-self-hosted-provider",
36
        inference_parameters=dd.ChatCompletionInferenceParams(
37
            temperature=0.7,
38
            top_p=0.9,
39
            max_tokens=1024,
40
        ),
41
    ),
42
]
43
44
# Step 3: Create DataDesigner with custom providers
45
data_designer = DataDesigner(model_providers=custom_providers)
46
47
# Step 4: Create config builder with custom models
48
config_builder = dd.DataDesignerConfigBuilder(model_configs=custom_models)
49
50
# Step 5: Add a topic column using a categorical sampler
51
config_builder.add_column(
52
    dd.SamplerColumnConfig(
53
        name="topic",
54
        sampler_type=dd.SamplerType.CATEGORY,
55
        params=dd.CategorySamplerParams(
56
            values=["Technology", "Healthcare", "Finance", "Education"],
57
        ),
58
    )
59
)
60
61
# Step 6: Use your custom model by referencing its alias
62
config_builder.add_column(
63
    dd.LLMTextColumnConfig(
64
        name="short_news_article",
65
        model_alias="my-text-model",  # Reference custom alias
66
        prompt="Write a short news article about the '{{topic}}' topic in 10 sentences.",
67
    )
68
)
69
70
config_builder.add_column(
71
    dd.LLMTextColumnConfig(
72
        name="long_news_article",
73
        model_alias="my-self-hosted-text-model",  # Reference custom alias
74
        prompt="Write a detailed news article about the '{{topic}}' topic.",
75
    )
76
)
77
78
# Step 7: Preview your dataset
79
preview_result = data_designer.preview(config_builder=config_builder)
80
preview_result.display_sample_record()