Unsupported characters in Function `_wrapped_model` contains input name(s)

TensorFlow General Discussion
1

I’ve been using Tensorflow 2.4 and recently upgraded to 2.5 and got this warning, I know it’s only a warning but it’s rather odd as the rename just put it all on lowercase? Hoping someone has insight on the warning.

WARNING - Function `_wrapped_model` contains input name(s) BP, BPA, CF, CP, CRS, Class, D, DIST, DLR, FU, JA, JT, LSC, LSD, LSL, LSP, LST, LSW, NR, P, Race, SP, SU, TA, TIM, TU, UP, WT with 
unsupported characters which will be renamed to bp, bpa, cf, cp, crs, class, d, dist, dlr, fu, ja, jt, lsc, lsd, lsl, lsp, lst, lsw, nr, p, race, sp, su, ta, tim, tu, up, wt in the SavedModel.
2 Likes
2

It was introduced with

https://github.com/tensorflow/tensorflow/commit/9f14e9eb15ba26fbf1e30750494cdd4f0b42ab5f

Just to inform the user when an input name Is normalized.

2 Likes
3

Don’t use uppercase letters in names. There’s automatic conversion until there isn’t. Here’s a nice little nonsense model that will compile and print the summary but which cannot be saved. Because one name is ‘a’ and another name is ‘A’.

input1 = keras.Input(shape=2, name='a')
input2 = keras.Input(shape=2, name='A')
outputs = tf.keras.layers.Average()([input1, input2])

dummy = keras.Model(inputs=[input1, input2], outputs=outputs)
dummy.summary()

dummy.compile(loss='categorical_crossentropy', optimizer='adam', metrics=['accuracy'])

dummy.save("dummy.savedmodel")

So listen to the warning ; ]