jych / nips2015_vrnn Goto Github PK

I'm trying to train your Blizzard 2013 models and it looks like there are some preprocessing steps to go from the Blizzard dataset dumps to the HD5 and .npy files read by the models.

Could you let me know what needs to be done to reproduce? Many thanks!

Hi,

I started the training the iamondb script and wanted to know if there is a test script where I could test the output of the model?

I do not see a test script in the package. It would be a great help if you could point me to a script where I could test the code.

Regards,
Vivek

Trying to download blizzard 2013 datasets. Link down?
www.synsig.org/index.php/Blizzard_Challenge_2013

Im trying to run the IAMOnDB example.

I get an error when i run fetch_iamondb(...)

Cannot find concatenated train.txt and valid.txtfiles! See the README in data

I downloaded and unpacked

1. ascii-all.tar.gz
2. lineImages-all.tar.gz
3. lineStrokes-all.tar.gz
4. original-xml-all.tar.gz
5. original-xml-part.tar.gz

Can you point me in the right direction?

I also had some trouble figuring out from which folder i should execute the files such that CLE is properly imported?

Entering main loop
Traceback (most recent call last):
File "vrnn_gauss.py", line 371, in
main(params)
File "vrnn_gauss.py", line 349, in main
mainloop.run()
File "cle/cle/train/init.py", line 83, in run
while self.run_epoch():
File "cle/cle/train/init.py", line 89, in run_epoch
for batch in self.data:
File "cle/cle/data/init.py", line 90, in iter
yield self.data.slices(idx, idx + self.batch_size)
File "../../../nips2015_vrnn/datasets/iamondb.py", line 97, in slices
mask = self.create_mask(batches[0].swapaxes(0, 1))
ValueError: bad axis2 argument to swapaxes

Hi, I run the model of iamondb ,and got the eroor message , how can I fix it ? thx !

Hi, read-me file says, "The original wave files have been read by numpy and saved into '.npz' format. There is a function that reads the numpy formatted files and generate a hdf5 format file." I can't find the script which converts to .npz formatted files, or to hdf5.

I also wanted to know as to what do you mean by "full_blizzard" and "blizzard_tbptt" while reading the blizzard data in the file "blizzard.py".

We can get it from Kratarth's repo - unless you forgot to include the files and have them locally

Error: no module named 'cle.cle'
But I have installed the module 'cle'.How to solve it?