Requirements¶

To use Pinferencia's frontend with your model, there are some requirements of your model's predict function.

Templates¶

Currently, there are mainly two major catogory of the template's inputs and outputs. More (audio and video) will be supported in the future.

Base Templates¶

Derived Templates¶

Input¶

Based on the schema of the request, frontend end may parse the input into a list or simply a single string.

If you define your schema request as a list, for example, List[str], or simply list:

1. The predict function must be able to accept a list of data as inputs.
2. For text input, the input will be a list of strings.
3. For image input, the input will be a list of strings representing the base64 encoded images.

Otherwise,

1. The predict function must be able to accept a single data as the input.
2. For text input, the input will be a single string.
3. For image input, the input will be a single string representing the base64 encoded image.

Output¶

If you define your schema response as a list, for example, List[str], or simply list:

1. The predict function must produce a list of data as outputs.
2. For text output, the output must be a list.
3. For image output, the output must be a list of strings representing the base64 encoded images.

Otherwise,

1. The predict function must produce a single data as the output.
2. For text output, the output should be a single string.
3. For image output, the output should be a single string representing the base64 encoded image.

[
    [
        {"a": 1, "b": 2},
        {"a": 3, "b": 4},
        {"a": 5, "b": 6}
    ]
]

[
    {"a": 1, "b": 2},
    {"a": 3, "b": 4},
    {"a": 5, "b": 6}
]

[
    "Text output."
]

"Text output."

[
    [
        {"a": 1, "b": 2},
        1,
        "a"
    ]
]

{
    "a": 1,
    "b": 2
}