InvalidArgumentError: indices[127,7] = 43 is not in [0, 43) in Keras R

Question

The issue is related to: InvalidArgumentError (see above for traceback): indices[1] = 10 is not in [0, 10) I need it for R and therefore another solution than given in the link above.

maxlen <- 40
chars <- c("'",  "-",  " ",  "!",  "\"", "(",  ")",  ",",  ".",  ":",  ";",  "?",  "[",  "]",  "_",  "=",  "0", "a",  "b",  "c",  "d",  "e", "f",  "g",  "h",  "i",  "j",  "k",  "l",  "m",  "n",  "o",  "p",  "q",  "r",  "s",  "t",  "u",  "v",  "w",  "x",  "y",  "z")



tokenizer <- text_tokenizer(char_level = T, filters = NULL)

tokenizer %>% fit_text_tokenizer(chars)
unlist(tokenizer$word_index)

Output is:

 '  -     !  "  (  )  ,  .  :  ;  ?  [  ]  _  =  0  a  b  c  d  e  f  g  h  i  j  k  l  m  n  o  p  q  r  s  t  u  v  w  x  y  z 
 1  2  3  4  5  6  7  8  9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 

How can I change the indexing so it starts from 0 not from 1 in text_tokenizer?

The error I get after running fit() is as follows:

InvalidArgumentError: indices[127,7] = 43 is not in [0, 43)
     [[Node: embedding_3/embedding_lookup = GatherV2[Taxis=DT_INT32, Tindices=DT_INT32, Tparams=DT_FLOAT, _class=["loc:@training_1/RMSprop/Assign_1"], _device="/job:localhost/replica:0/task:0/device:CPU:0"](embedding_3/embeddings/read, embedding_3/Cast, training_1/RMSprop/gradients/embedding_3/embedding_lookup_grad/concat/axis)]]

But I believe that changing the Indexing will solve my problem.

Answer 1

Index 0 is often reserved for padding so it is not a wise idea to start your actual character indices from 0 as well. Instead you should venture to the Embedding layer and add 1 to the input size as suggested by the documentation:

input_dim: int > 0. Size of the vocabulary, i.e. maximum integer index + 1.

In your case this would be 43 + 1 = 44.