Congratulations on this very interesting and nice paper.

I was just looking through your released AnimalSpeak CSV file at HuggingFace. The file contains only 894284 entries, but in the paper, you mention 1.1M audio-caption pairs. Could you maybe clarify where the discrepancy comes from?

Will you release a script that downloads all the snippets and processes everything so that it can be used to recreate the results from your paper? That would be a great help. Thank you very much.

Hello,
Big thanks for your work!!

I would like to compute the audio embeddings to try to cluster several bioacoustic events.
Therefore, I would like the pretrained model to be trained on AnimalSpeak (BioLingual model).
In the huggingface description/documentation looks like BioLingual can only be used with the pipeline, and that to get the audio embeddings it is necessary to load the CLAP pretrained weights.
I was wondering if this code below would be correct?

`
import torch
import torchaudio

from transformers import ClapModel, ClapProcessor

waveform, fs = torchaudio.load('wav_example.wav')
device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
model = ClapModel.from_pretrained("davidrrobinson/BioLingual").to(device)
processor = ClapProcessor.from_pretrained("davidrrobinson/BioLingual", sampling_rate=fs)
inputs = processor(audios=waveform[0, :], return_tensors="pt", sampling_rate=fs).to(device)
audio_embed = model.get_audio_features(**inputs)
`

Thank you for your advice!

Best,

Clea