search

Kyubyong / tacotron

A TensorFlow Implementation of Tacotron: A Fully End-to-End Text-To-Speech Synthesis Model
Apache License 2.0
1.83k stars 436 forks source link

single GPU is better? #56

Open kyoguan opened 7 years ago

kyoguan commented 7 years ago
6-gpu

batch_size = 32*6 and num_gpus = 6(train_multi_gpus.py)


1-gpu

batch_size = 32 (train.py)

it seems the single gpu is mush faster and the training curve is smoother.

candlewill commented 7 years ago

@kyoguan The loss curve is pretty good. How many training data did you use?

kyoguan commented 7 years ago

just use the WEB data set, download from https://dl.dropboxusercontent.com/u/42868014/WEB.zip

jarheadfa commented 7 years ago

@kyoguan - Is this the epoch loss or the batch loss?