MCPcopy Create free account
hub / github.com/OpenSparseLLMs/Linear-MoE / tokenize

Method tokenize

linear_moe/data/starcoder.py:101–135  ·  view source on GitHub ↗

This function tokenizes the source strings given the tokenizer and returns a dictionary containing the tokenized inputs and labels. Args: strings (List[str]): The list of input strings. tokenizer (PreTrainedTokenizer): The tokenizer to use.

(self, strings, tokenizer)

Source from the content-addressed store, hash-verified

99 return dict(input_ids=input_ids, labels=labels)
100
101 def tokenize(self, strings, tokenizer):
102 """
103 This function tokenizes the source strings given the tokenizer and returns a dictionary containing the
104 tokenized inputs and labels.
105 Args:
106 strings (List[str]): The list of input strings.
107 tokenizer (PreTrainedTokenizer): The tokenizer to use.
108 max_input_length (Optional[int]): The maximum length of the input sequences.
109 max_target_length (Optional[int]): The maximum length of the target sequences.
110 Returns:
111 Dict[str, Any]: A dictionary containing input_ids, labels, input_ids_lens and labels_lens.
112 """
113
114 tokenized_list = [
115 tokenizer(
116 text,
117 return_tensors='np',
118 padding='max_length',
119 max_length=self.max_padding_length,
120 truncation=True,
121 ) for text in strings
122 ]
123 input_ids = labels = [
124 tokenized.input_ids[0] for tokenized in tokenized_list
125 ]
126 input_ids_lens = labels_lens = [
127 (tokenized.input_ids != tokenizer.pad_token_id).sum().item()
128 for tokenized in tokenized_list
129 ]
130 return dict(
131 input_ids=input_ids,
132 labels=labels,
133 input_ids_lens=input_ids_lens,
134 labels_lens=labels_lens,
135 )
136
137 def gpt_convert_example_to_feature(self, sample):
138 input_ids, labels = sample

Callers 1

preprocessMethod · 0.95

Calls

no outgoing calls

Tested by

no test coverage detected