search

zju-muslab / AdvReverb

Code of paper "AdvReverb: AdvReverb: Rethinking the Stealthiness of Audio Adversarial Examples to Human Perception"
14 stars 1 forks source link

ASR weights #2

Open FreddieMG opened 3 weeks ago

FreddieMG commented 3 weeks ago

Can you please specify the exact Pretrained Wav2Vec2 weights used and where to place them to make the script work? Thanks

mangophant commented 1 week ago

You can find the pretrained weight at: https://huggingface.co/speechbrain/asr-wav2vec2-librispeech. Download the model files to the directory: ckpt/asr/asr-wav2vec2-librispeech You can also change the directory in the yaml file: config/asr/wav2vec2.yaml