search

sdv-dev / TGAN

Generative adversarial training for generating synthetic tabular data.
MIT License
278 stars 91 forks source link

census data postprocessing issue #23

Closed Venseven closed 5 years ago

Venseven commented 5 years ago

I have trained the census data ....after training and on restoring checkpoints it kind of stucks in generating samples...have attached the issue issue

ManuelAlvarezC commented 5 years ago

Hi @Venseven,

Would you mind sharing with us which version of TGAN you used and the commands you run?

Thanks

Venseven commented 5 years ago 
import tgan
num_samples=100
model_path="models/covertype_1.pkl"
from tgan.data import load_demo_data
data,c_m=load_demo_data("census")
from tgan.model import TGANModel
tgan_m=TGANModel(c_m,gpu=0)
tgan_m.fit(data)
sample=tgan_m.sample(num_samples)
print(sample)`

tgan --version-0.1.0

ManuelAlvarezC commented 5 years ago

I think that the issue here is that you are trying to sample too few records ( less than the batch size, that defaults to 200). If you try with 1000 samples it should work perfectly.