b04901014 / FT-w2v2-ser

Official implementation for the paper Exploring Wav2vec 2.0 fine-tuning for improved speech emotion recognition

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

Pre-training

BakingBrains opened this issue · comments

python run_pretrain.py --datadir Audio_Dir \
                       --labelpath Label_Path
                       --labeling_method hard \
                       --saving_path Saving_Path \
                       --training_step 10000 \
                       --save_top_k 1 \
                       --wav2vecpath Wav2vecCKPT \
                       --precision 16

Here --labelpath, should be a session label or metalabel.json?

In my experiment I only use session label since I don't want the test set to be revealed during pre-training, metalabel.json is simply a by-product.
But if you don't care, you can feed metalabel.json into it, it will pre-train on all of the IEMOCAP data regardless of the splits and the categories.