The source code of our submission to the DCASE2023 challenge task 6a is available at https://github.com/Labbeti/conette-audio-captioning.
The model is called CoNeTTE and is almost the same model than our best submission to the DCASE challenge (apart for few hyperparameters).
It is described in the corresponding paper available at https://arxiv.org/abs/2309.00454.