No way of testing my currently training models. Sorry 4 ping @hecko but it seems like a minor problem you might know of. I yope.

frick

typo

refresh

Oh, cool!

Someone Train This Model

Is the leaking caffe2 wqrning normal now when training HiFi-gan?

Is it necessary to copy the Tacotron before creating a voice model?

Can someone help me?

js
FP16 Run: False
Dynamic Loss Scaling: True
Distributed Run: False
cuDNN Enabled: True
cuDNN Benchmark: False
Loading checkpoint '/content/drive/MyDrive/Tacotron 2/models/Tacotron 2 model'
Loaded checkpoint '/content/drive/MyDrive/Tacotron 2/models/Tacotron 2 model' from iteration 0
---------------------------------------------------------------------------
ZeroDivisionError                         Traceback (most recent call last)
<ipython-input-11-26e552a6b230> in <module>
      7 print('cuDNN Benchmark:', hparams.cudnn_benchmark)
      8 train(output_directory, log_directory, checkpoint_path,
----> 9       warm_start, n_gpus, rank, group_name, hparams, log_directory2)

<ipython-input-2-f6cde9217cf2> in train(output_directory, log_directory, checkpoint_path, warm_start, n_gpus, rank, group_name, hparams, log_directory2)
    291                 learning_rate = _learning_rate
    292             iteration += 1  # next iteration is iteration + 1
--> 293             epoch_offset = max(0, int(iteration / len(train_loader)))
    294     else:
    295       os.path.isfile("/content/tacotron2/pretrained_model")

ZeroDivisionError: division by zero

I've been trouble using the Pipeline Tacotron 2 training model. I'm trying to extract the Dataset from the audio files and transcript I need to use but it's been fairly difficult. Is there another training model I can use Or can I have some help learning how to extract datasets?

Please I just want to be able to finish David freaking Gasman saying beep boop

Wish I could homie 💀 My best bet is that it's an outdated training model

What do you mean "extract" the dataset, exactly?

It's asking me to have the folder of the Wav's and transcript within google drive, Then have the dataset of the Google drive file pasted into the colab. It offers an option to just run step 2 manually without any input; so I can post the Wav's and transcript in the ``FILE_HERE`` folder, but it doesn't show up when I do that

Uh if you want to upload a datset manually,

Hold on let me find the post

Hh wait

Okay 👍

Is this about loading the dataset from a link or manually?

If you give the nb the zip right it should process it normally?

Sorry I'm just a little confused

You refer to the "step 2 loading datasets" sections, right?

Yes!

I was actually just screenshoting on mobile for an example on what it says

It doesn't leave an ``UPLOAD_HERE`` option for me; It only seems like doing the dataset way is the only option to do it. Which is super stressful when In the past I was just simply able to to put the respective Wav's and transcript into the side folder for other Training models

excuse the terrible scribbles. Should be sinilar methods for legacy and pipeline notebooks, which is 1) Running setup cells to get basically everytjing that does run later cells, If you upload manually, the folder will be created in this step. You can access that from the file picker in the side bar, if it's literally anywhere else on your screen idk. Formatted zip should be like the example I posted earlier.

Same format for the Gdrive method an manual

I do remember running that first; I still didn't get the file