ksangeeta2429 / edgel3 Goto Github PK

I observed that the window size is 1 second by default, and that the last embedding corresponds to a window whose center is 0.5 second (half a window size) less than the total audio length. Is there a way to change the window size so that I can continue to get embeddings for the last 0.5 seconds?

Alternatively, I can adding half a second's length of zero padding at the end (like _center_audio does at the beginning), so that I can get embeddings corresponding to a window centered at the very end of the audio.

Just like OpenL3, the EdgeL3 package is not compatible with the newer version of Kapre (1.7). And while the package can still be installed successfully via pip install edgel3 it is unusable, (i.e. calls to edgel3.get_embedding or edgel3.models.load_embedding_model error out).

The Kapre version should be pinned to kapre==1.4, just like OpenL3 (see commit here

You can check out the below Colab Notebooks for examples of the package failing with kapre==1.7 and working with kapre==1.4

• Kapre 1.4 (working)
• Kapre 1.7 (broken)

Hi ksangeeta2429,

Thank you for making the code and models available. I am experimenting with edgel3 and noticed that the model parameter files are the same size (~35M) as the openl3 models. Shouldn't they be much smaller?

Thank you,
Matthias