i'm using random embedding nets to do weak self supervision & it works ok. i "make" my random embedding nets by just initing & then not training. however, since the dft init is in prep for training, & i'm not training, i wonder is there another init approach i should use?
-
-
you mean one hot encoding?
-
i'm assuming tf.keras.initializers.Orthogonal
End of conversation
New conversation -
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.