clika_runtime.modelverse.pipelines

Task pipelines: pipeline(task, model) returns a callable that loads the right model kind for the task and answers in the shapes model-hub pipelines use::

import clika_runtime.modelverse as mv

generate = mv.pipeline("text-generation", model="Qwen/Qwen2.5-0.5B-Instruct", device="cuda:0")
print(generate("hello", max_new_tokens=32)[0]["generated_text"])

transcribe = mv.pipeline("automatic-speech-recognition", model="openai/whisper-small")
print(transcribe("clip.wav")["text"])

The eleven tasks: text-generation, automatic-speech-recognition, text-to-speech, feature-extraction, image-feature-extraction, zero-shot-image-classification, object-detection, depth-estimation, translation, image-to-text, text-to-image. Images are [H, W, 3] tensors; audio a file path, encoded bytes, or a features tensor.

Name	Kind
AutomaticSpeechRecognitionPipeline	class
DepthEstimationPipeline	class
FeatureExtractionPipeline	class
ImageFeatureExtractionPipeline	class
ImageToTextPipeline	class
ObjectDetectionPipeline	class
Pipeline	class
TextGenerationPipeline	class
TextToImagePipeline	class
TextToSpeechPipeline	class
TranslationPipeline	class
ZeroShotImageClassificationPipeline	class
functions	module functions