You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
When we quantize a model using a minifloat format and the fx backend, the residual handler tries to access QuantAct modules' int_scaling_impl, as seen here.
This throws an error as minifloat configurations will have a FloatQuant quantizer and therefore no int_scaling_impl but a float_scaling_impl.
This only occurs with the fx backend.
The text was updated successfully, but these errors were encountered:
When we quantize a model using a minifloat format and the
fx
backend, the residual handler tries to access QuantAct modules'int_scaling_impl
, as seen here.This throws an error as minifloat configurations will have a
FloatQuant
quantizer and therefore noint_scaling_impl
but afloat_scaling_impl
.This only occurs with the
fx
backend.The text was updated successfully, but these errors were encountered: