Inference API

Method

$api->generate

Parameters

• model_id: Model ID on the HF Hub
• inputs: Input text
• min_length: Minimum number of tokens to generate (default 0)
• max_length: Maximum number of tokens to generate (default 256)
• top_k: Top K (default null)
• top_p: Top P (default null)
• temperature: Temperature (default 1.0)
• repetition_penalty: Repetition penalty (default null)
• max_time: Maximum generation time (default 120 seconds, can range from 1s to 120s)
• do_sample: Sample (default null)
• use_cache: Cache responses (default null)
• wait_for_model: Wait for model to load before generating, instead of erroring if model isn't loaded (default True)

Example

Complete the text "Hello:"

include 'hub.php';
$api = new Hub('hf_***');
echo($api->generate('gpt2', 'Hello')['generated_text']);