changed to input ids

ref_seg_ger.py

@@ -18,7 +18,7 @@ from glob import glob
 import os
 import numpy as np
 from PIL import Image
-from tokenizers.pre_tokenizers import Whitespace
+from transformers import AutoTokenizer
 import datasets
 from itertools import chain
 import pandas as pd
@@ -62,7 +62,7 @@ _LABELS = [
 _FEATURES = datasets.Features(
     {
         #"id": datasets.Value("string"),
-        "input_ids": datasets.Sequence(datasets.Value("string")),
+        "input_ids": datasets.Sequence(datasets.Value("int64")),
         #"bbox": datasets.Sequence(datasets.Sequence(datasets.Value("int64"))),
         # "RGBs": datasets.Sequence(datasets.Sequence(datasets.Value("int64"))),
         # "fonts": datasets.Sequence(datasets.Value("string")),
@@ -136,7 +136,7 @@ class RefSeg(datasets.GeneratorBasedBuilder):
     # ]
 
     # DEFAULT_CONFIG_NAME = "small"  # It's not mandatory to have a default configuration. Just use one if it make sense.
-    TOKENIZER = Whitespace()
+    TOKENIZER = AutoTokenizer.from_pretrained("xlm-roberta-base")
 
     def _info(self):
         # TODO: This method specifies the datasets.DatasetInfo object which contains informations and typings for the dataset
@@ -214,25 +214,30 @@ class RefSeg(datasets.GeneratorBasedBuilder):
             labels = []
             for i, row in df.iterrows():
 
-                tokens = self.TOKENIZER.pre_tokenize_str(row['token'])
+                tokenized_input = self.TOKENIZER(
+                    row['token'],
+                    add_special_tokens=False,
+                    return_offsets_mapping=False,
+                    return_attention_mask=False,
+                )
                 #print(tokenized_input)
-                if len(tokens) > 1:
+                if len(tokenized_input['input_ids']) > 1:
                     if row['tag'] == 'B':
-                        input_ids.append(tokens[0][0])
+                        input_ids.append(tokenized_input['input_ids'][0])
                         labels.append(row['tag'] + '-' + row['label'])
-                        for input_id in tokens[1:]:
-                            input_ids.append(input_id[0])
+                        for input_id in tokenized_input['input_ids'][1:]:
+                            input_ids.append(input_id)
                             labels.append('I-' + row['label'])
                     elif row['tag'] == 'I':
-                        for input_id in tokens:
-                            input_ids.append(input_id[0])
+                        for input_id in tokenized_input['input_ids']:
+                            input_ids.append(input_id)
                             labels.append('I-' + row['label'])
                     else:
-                        for input_id in tokens:
-                            input_ids.append(input_id[0])
+                        for input_id in tokenized_input['input_ids']:
+                            input_ids.append(input_id)
                             labels.append('O')
-                elif len(tokens) == 1:
-                    input_ids.append(tokens[0][0])
+                elif len(tokenized_input['input_ids']) == 1:
+                    input_ids.append(tokenized_input['input_ids'][0])
                     if row['tag'] == 'O':
                         labels.append(row['tag'])
                     else:
@@ -247,16 +252,16 @@ class RefSeg(datasets.GeneratorBasedBuilder):
                 # split_rgbs = rgbs[index:index + self.CHUNK_SIZE]
                 # split_fonts = fonts[index:index + self.CHUNK_SIZE]
                 split_labels = labels[index:max(len(input_ids), index + self.CHUNK_SIZE)]
-                #split_labels_post = [item for sublist in split_labels for item in sublist]
-                # if(len(split_ids) != len(split_labels)):
-                #     print(f)
-                #     print(len(input_ids), input_ids)
-                #     print(len(split_labels), split_labels)
-                # for s in split_labels:
-                #     if type(s) is not str:
-                #         print(f)
-                #         print(len(input_ids), input_ids)
-                #         print(len(split_labels), split_labels)
+                split_labels_post = [item for sublist in split_labels for item in sublist]
+                if(len(split_ids) != len(split_labels)):
+                    print(f)
+                    print(len(input_ids), input_ids)
+                    print(len(split_labels), split_labels)
+                for s in split_labels:
+                    if type(s) is not str:
+                        print(f)
+                        print(len(input_ids), input_ids)
+                        print(len(split_labels), split_labels)
                     #print(len(split_labels_post), split_labels_post)
                 #print(split_labels, len(split_labels))
                 #print(split_ids, len(split_ids))