GPU vs CPU performance

[sohanasarah]

I am training neural-prophet with a large dataset. I am trying to use accelerator: gpu feature but I noticed that training performance is better on cpu than gpu. I tried to increase batch size and decrease epoch still I get the same result. Am I missing something? Below is the parameter setup that I used:

params = dict(
weekly_seasonality = False,
daily_seasonality = False,
seasonality_mode = 'multiplicative',
epochs=500,
learning_rate=0.05,
accelerator = "gpu"
)

[Kevin-Chen0]

^ @karl-richter

[lilGao16]

me too,really wonder if it's normal or there is other parameters i can adjust