Kaizhi Qian

GPU

@yuanyuan0209 You could simply convert them to the proper datatype. For inference, I used the "Fast WaveNet" algorithm, which is a little tricky to implement. However, you could just follow...

@ucasiggcas 文档里面有说

I have a few models, but I can't remember which one of them works with the current branch. I will share it once I figure that out.

Does it make sense that the waveform generated under training mode sounds worse than that generated under inference mode? I mean, under training mode, since the model has access to...

@avanitanna Just compute the mean and std of the mfcc feature.

``` dctmx = scipy.fftpack.dct(np.eye(80), type=2, axis=1, norm='ortho') # compute mfcc stats using all spectrograms mfcc_all = sp_all.dot(dctmx) mfcc_mean, mfcc_std = np.mean(mfcc_all,axis=0), np.std(mfcc_all,axis=0) # normalize each mfcc cc_tmp = sp_tmp.dot(dctmx) cc_norm...

@avanitanna sp_all is the concatenation of all mel spectrogram, sp_tmp is the spectrogram you want to normalize

You need to use the fairseq version set up by this repo to run inference. The same fairseq framework you used to train the contentvec. That is why the legacy...

The run.sh will override the config.