search

syang1993 / gst-tacotron

A tensorflow implementation of the "Style Tokens: Unsupervised Style Modeling, Control and Transfer in End-to-End Speech Synthesis"
368 stars 110 forks source link

training time #8

Closed Young-Sun closed 6 years ago

Young-Sun commented 6 years ago

@syang1993 It's nice work! :-) The latest demo samples are impressing. How long does it take to GST-Tacotron model using segmented Blizzard2013 datasets? And would you let me know GPU specification?

syang1993 commented 6 years ago

@Young-Sun Thanks. I trained this model with my friend's Tesla P40 GPU. It takes about 4 days to get this model using bc2013 data.

butterl commented 6 years ago

@syang1993 Is it normal to run model training with 3.8s/step ? the hparam is same ,using THCH30 dataset and 16K with max frame limit set to 1500

Young-Sun commented 6 years ago

@syang1993 Thanks for your reply. It would be great help to estimate training time :-)

fazlekarim commented 6 years ago

@Young-Sun I took me about 2 seconds per step. So if you want 200k steps, it'll take around 400k seconds.