thats been fixing a lot of issues lately

Oh, thank u!

UnpicklingError Traceback (most recent call last) in () 5 print('cuDNN Benchmark:', hparams.cudnn_benchmark) 6 train(output_directory, log_directory, checkpoint_path, ----> 7 warm_start, n_gpus, rank, group_name, hparams, log_directory2) 3 frames /usr/local/lib/python3.7/dist-packages/torch/serialization.py in _legacy_load(f, map_location, pickle_module, **pickle_load_args) 775 "functionality.") 776 --> 777 magic_number = pickle_module.load(f, **pickle_load_args) 778 if magic_number != MAGIC_NUMBER: 779 raise RuntimeError("Invalid magic number; corrupt file?") UnpicklingError: invalid load key, '<'. Can someone help with this error from the tacotron training model?

Well, I tried to do some Link2Model on Fowler, and he didn't sound good because of low quality.

trained my Dorkly Sonic dataset, is it supposed to look like this at the end of training?

used the name1.wav transcription for this, it almost sounds the same.

it sounds KINDA impressive despite having only 29 voicelines

hello i am new to this synthesizer

i dont know which same device

Do I have to wait until it goes all the way 5001 epoch or does it stop at 50?

nvm i found a way to fix it

just dont use gpu

i am testing around with max length decoder to remove this max decoder warning

i figured out just put to 10

Is there any way to make the AI pause for a minute before reading the next sentence?

Like commas and periods

Have you used those?

Yea and its a 50 50 shot

Most of the time it does not work tho

got dorkly tails to say "jesus christ", even though there's an "oh" at the beginning despite there being no "oh" in the beginning. kinda makes me laugh

wait is this the right channel?

definitely not all the way to 5001 i recommend waiting until it's, say, 0.10 loss or until it stops improving

Well it's too late!

I screwed up the model twice!

yeah the new notebook is kind of a mess

only saves every 50 epochs, for example

and resuming from save is mildly complex

I don't know how this stuff works.. I even used the first training model and even that couldn't keep going with the training. It stopped saving and ended without finishing ultimately.

hmm interesting

next time if something happens please send screenshots of any and all error messages