Fix whole word masking labels (#326)

huggingface · Oct 4, 2022 · 4cd2c0e · 4cd2c0e
1 parent 2a61a64
commit 4cd2c0e
Show file tree

Hide file tree

Showing 5 changed files with 10 additions and 0 deletions.
diff --git a/chapters/en/chapter7/3.mdx b/chapters/en/chapter7/3.mdx
@@ -523,6 +523,7 @@ def whole_word_masking_data_collator(features):
             for idx in mapping[word_id]:
                 new_labels[idx] = labels[idx]
                 input_ids[idx] = tokenizer.mask_token_id
+        feature["labels"] = new_labels
 
     return default_data_collator(features)
 ```
@@ -563,6 +564,7 @@ def whole_word_masking_data_collator(features):
             for idx in mapping[word_id]:
                 new_labels[idx] = labels[idx]
                 input_ids[idx] = tokenizer.mask_token_id
+        feature["labels"] = new_labels
 
     return tf_default_data_collator(features)
 ```

diff --git a/chapters/fr/chapter7/3.mdx b/chapters/fr/chapter7/3.mdx
@@ -524,6 +524,7 @@ def whole_word_masking_data_collator(features):
             for idx in mapping[word_id]:
                 new_labels[idx] = labels[idx]
                 input_ids[idx] = tokenizer.mask_token_id
+        feature["labels"] = new_labels
 
     return default_data_collator(features)
 ```
@@ -564,6 +565,7 @@ def whole_word_masking_data_collator(features):
             for idx in mapping[word_id]:
                 new_labels[idx] = labels[idx]
                 input_ids[idx] = tokenizer.mask_token_id
+        feature["labels"] = new_labels
 
     return tf_default_data_collator(features)
 ```

diff --git a/chapters/ja/chapter7/3.mdx b/chapters/ja/chapter7/3.mdx
@@ -535,6 +535,7 @@ def whole_word_masking_data_collator(features):
             for idx in mapping[word_id]:
                 new_labels[idx] = labels[idx]
                 input_ids[idx] = tokenizer.mask_token_id
+        feature["labels"] = new_labels
 
     return default_data_collator(features)
 ```
@@ -575,6 +576,7 @@ def whole_word_masking_data_collator(features):
             for idx in mapping[word_id]:
                 new_labels[idx] = labels[idx]
                 input_ids[idx] = tokenizer.mask_token_id
+        feature["labels"] = new_labels
 
     return tf_default_data_collator(features)
 ```

diff --git a/chapters/vi/chapter7/3.mdx b/chapters/vi/chapter7/3.mdx
@@ -523,6 +523,7 @@ def whole_word_masking_data_collator(features):
             for idx in mapping[word_id]:
                 new_labels[idx] = labels[idx]
                 input_ids[idx] = tokenizer.mask_token_id
+        feature["labels"] = new_labels
 
     return default_data_collator(features)
 ```
@@ -563,6 +564,7 @@ def whole_word_masking_data_collator(features):
             for idx in mapping[word_id]:
                 new_labels[idx] = labels[idx]
                 input_ids[idx] = tokenizer.mask_token_id
+        feature["labels"] = new_labels
 
     return tf_default_data_collator(features)
 ```

diff --git a/chapters/zh-CN/chapter7/3.mdx b/chapters/zh-CN/chapter7/3.mdx
@@ -524,6 +524,7 @@ def whole_word_masking_data_collator(features):
             for idx in mapping[word_id]:
                 new_labels[idx] = labels[idx]
                 input_ids[idx] = tokenizer.mask_token_id
+        feature["labels"] = new_labels
 
     return default_data_collator(features)
 ```
@@ -564,6 +565,7 @@ def whole_word_masking_data_collator(features):
             for idx in mapping[word_id]:
                 new_labels[idx] = labels[idx]
                 input_ids[idx] = tokenizer.mask_token_id
+        feature["labels"] = new_labels
 
     return tf_default_data_collator(features)
 ```