search

casperkaae / parmesan

Variational and semi-supervised neural network toppings for Lasagne
Other
208 stars 31 forks source link

add seed option to sample layers #31

Closed casperkaae closed 8 years ago

casperkaae commented 8 years ago

add seed option to SampleLayer and SimpleSampleLayer init

wuaalb commented 8 years ago

Additionally, could we add a function to re-seed existing instances?

def seed(self, seed=None):
   self._srng.seed(seed)

(I don't know how to make changes to your PR on github)

I may

  1. Instantiate model
  2. Train
  3. Save weights
  4. Re-seed
  5. Use for something (e.g. compute log likelihood, ..)
  6. Re-seed
  7. Use for something else

Or

  1. Instantiate model
  2. Load weights
  3. Re-seed
  4. Use for something (e.g. compute log likelihood, ..)

In those cases I don't want to create new instances of the model just to fix the RNG seed..

casperkaae commented 8 years ago

ok, better now?

wuaalb commented 8 years ago

Great, thanks.

casperkaae commented 8 years ago

merged #31