SmolVLM: Small yet Mighty 💫

Play with HuggingFaceTB/SmolVLM-Instruct in this demo. To get started, upload an image and text or try one of the examples. This checkpoint works best with single turn conversations, so clear the conversation after a single turn.
Examples
Query Input Decoding strategy Sampling temperature Maximum number of new tokens to generate Repetition penalty Top P
Decoding strategy

Higher values is equivalent to sampling more low-probability tokens.
0 5
8 1024
0.01 5
0.01 0.99