sTinyStories A spoken version of TinyStories Synthesized with LJ voice using FastSpeech2. The dataset was synthesized to boost the training of Speech Language Models as detailed in the paper " Slamming : Training a Speech Language Model on One GPU in a Day". It was first suggested by Cuervo et. al 2024. We refer you to the SlamKit codebase to see how you can train a SpeechLM with this dataset. Usage sTinyStories is big, you can download parital sets using the streaming=True variable. You can read more about handling large datasets using the streaming variable here. Another possibility to download a partial set using hf hub download : Data Fields The data has several fields: audio : The synthesized audio. text : The text of the prompt audio. aligned text : The text of the prompt audio, aligned to the time in the audio: [("word", "start time", "end time") ... ] split sentence : boolean some of the stories were generated while splitting the sentences to provide better audio quality and seperation between setnences. Due to compute budget, we generated only some of the texts with this technique. Citation If you use this version of the dataset please cite our work:
We use cookies for essential functionality and analytics. You can accept or reject analytics cookies.Cookie policy