I think there is a bug in your beta schedule calculation, as you go with the default beta_startvalue of 0.02 and only provide the end value of 1e-4. However, you scale the provided end value (beta_max in the configuration) with 1/N_timesteps while the start value of 0.02 is kept as is.
I guess that the values are taken from VPSDE (20 and 0.1) and and already scaled down by the number of timesteps (20/1000, 0.1/1000), so what you wanted to do is more something like this:
I think there is a bug in your beta schedule calculation, as you go with the default
beta_start
value of0.02
and only provide the end value of1e-4
. However, you scale the provided end value (beta_max
in the configuration) with 1/N_timesteps while the start value of0.02
is kept as is.I guess that the values are taken from VPSDE (20 and 0.1) and and already scaled down by the number of timesteps (20/1000, 0.1/1000), so what you wanted to do is more something like this:
Is this correct or is there areason to only scaling the
linear_end
variable?