Releases: RF5/simple-speech-commands
Releases · RF5/simple-speech-commands
Add pretrained resnext sc09 model
This includes the pretrained model adapted from https://github.com/HazyResearch/state-spaces/tree/main/sashimi/sc09_classifier . Please see details in original repo for model training specification. Its usage is the same as existing models, however gradient flow from waveform to output is not supported due to reliance on librosa mel-spectrogram transform.
v0.6 SC09 Model Release
Add convgru_classifier_sc09
checkpoint file. Include checkpoint both from end (3k steps), or from lowest validation loss (ckpt_lowest_validation).
Initial model release
Add convgru classifier checkpoint.