deepglugs
deepglugs
set pin_memory to false (and possibly workers to 1) in your preset.json
I have to add this to the training loop to get it to fit on my 24G card (dataset dependant): ```python if txt.size(1) 550: continue ``` This essentially skips data...
A bit of an update... I'm training on the LJ dataset and I don't get noise. So something about my dataset is troublesome for flowtron. My data has a lot...
Another update. Looks like 32bit wav data was my issue. Now I get jibberish output with the model never attending to the text. Attention weights look poor after 1.6m steps...
Yes another update: Still trying to figure out the differences between my dataset and ljs. There are two remaining possibilities that come to mind: utterance length and total dataset size....
LJS with 2500 samples I have attention starting at 85k. here's 185k 
That seems to have done the trick! The directions for training from scratch seem to apply to pre-trained models as well. I'm seeing a lot of stuttering in the audio...
see #119. For inference, you don't need nearly as much VRAM, but the longer the utterance, the more VRAM you'll need in general.
> What's the GPU memory requirements to run this model? Doesn't look like I can run it on a 8GB 2070 Super even if I reduce batch size to 1....
try changing 'state_dict' to 'model' ```python state_dict = torch.load(model_path, map_location='cpu')['model'].state_dict() ```