HarshitSoni1903
commented
5 years ago Nevermind, I wrote a rough workaround for it. Turns out, that it occurred because when model is imported, all the parameters are encoded with it except for the LabelEncoder. It worked for you because you used notebook and all the variables were reused. but when it comes to standalone execution error occurs. Hence, to reuse the variable "lb"(instance of LableEncoder) we pickle dump it in a file and load it again in the same variable name. Hence the inverse_transform method of the lb object can be reused without fit or fit_transform, but one time execution is mandatory to initialize the variable and fit it. the code for dumping is as follows: import pickle file=open('lbsave.txt','wb') pickle.dump(lb,file) file.close()
loading the file is through: import pickle file=open('lbsave.txt','rb') lb=pickle.load(file)
MiteshPuthran
mohammedsuhail85
MK-43
anas2908
The model was imported perfectly. But the LabelEncoder did not work as:
added in cell 496
from sklearn.preprocessing import LabelEncoder lb = LabelEncoder() livepredictions = (lb.inverse_transform((liveabc))) livepredictions
throws the error:
If you could help with this, it will be of great help.
PS: I started by importing all files in cell 1,2,3. Then added 'opt = keras.optimizers.rmsprop(lr=0.00001, decay=1e-6)' to cell 137, since opt was not defined then executed all the blocks in Demo section.