Closed gfouilhe closed 1 year ago
Thanks for pointing this error out, we have a fix coming up in the develop
branch and will be fixed in the next release. For the moment, defining the InverseHessianVectorProduct
object before instantiating the InfluenceCalculator
should do the trick.
Module
Common
Contact Details
No response
Current Behavior
When using, as expected, the from_string interface to IHVPCalculator, e.g. :
influence_calculator_cgd = FirstOrderInfluenceCalculator(influence_model, batched_ds, "cgd")
or
influence_calculator_lissa = FirstOrderInfluenceCalculator(influence_model, batched_ds, "lissa")
we get the following error :
TypeError: ConjugateGradientDescentIHVP.__init__() missing 1 required positional argument: 'train_dataset'
Note that this behavior isn't observed when using
"exact"
string.Expected Behavior
Having
train_dataset
argument ofConjugateGradientDescentIHVP.__init__()
andLissaIHVP.__init__()
automatically set tobatched_ds
(that we already have to put inFirstOrderInfluenceCalculator
Version
v0.1.0
Environment
Relevant log output
No response
To Reproduce