-
Running
bash ./nn_se/_1_perprocess.sh
to prepare data. -
Change "root_dir" parameter in nn_se/FLAGS.py to the root of the project. For example "root_dir = /home/user/ulaw-SGAN-for-SE".
-
Ensure "PARAM = dse_ulawV2_G_FTmagmse_Ndloss_ssnr_001" is set in last line of nn_se/FLAGS.py.
-
Running
cp nn_se dse_ulawV2_G_FTmagmse_Ndloss_ssnr_001 -r
to create the Experiment config code dir.
Running python -m dse_ulawV2_G_FTmagmse_Ndloss_ssnr_001._2_train
to start training of config "dse_ulawV2_G_FTmagmse_Ndloss_ssnr_001".
Running python -m dse_ulawV2_G_FTmagmse_Ndloss_ssnr_001._3_enhance_testsets
to get the metrics of Experiment "dse_ulawV2_G_FTmagmse_Ndloss_ssnr_001". The last ckpt is selected as the default ckpt to load. Alse, you can use --ckpt
to specify the path of ckpt.
See "nn_se/_1_preprocess.sh", "nn_se/_2_train.py" and "nn_se/_3_enhance_testsets.py".