File size: 10,647 Bytes
cc9c7ee |
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 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 |
from src.utils.mapper import configmapper
from transformers import AutoTokenizer
import pandas as pd
from datasets import load_dataset, Dataset
from evaluation.fix_spans import _contiguous_ranges
@configmapper.map("datasets", "toxic_spans_multi_spans")
class ToxicSpansMultiSpansDataset:
def __init__(self, config):
self.config = config
self.tokenizer = AutoTokenizer.from_pretrained(
self.config.model_checkpoint_name
)
self.dataset = load_dataset("csv", data_files=dict(self.config.train_files))
self.test_dataset = load_dataset("csv", data_files=dict(self.config.eval_files))
temp_key_train = list(self.dataset.keys())[0]
self.intermediate_dataset = self.dataset.map(
self.create_train_features,
batched=True,
batch_size=1000000, ##Unusually Large Batch Size ## Needed For Correct ID mapping
remove_columns=self.dataset[temp_key_train].column_names,
)
temp_key_test = list(self.test_dataset.keys())[0]
self.intermediate_test_dataset = self.test_dataset.map(
self.create_test_features,
batched=True,
batch_size=1000000, ##Unusually Large Batch Size ## Needed For Correct ID mapping
remove_columns=self.test_dataset[temp_key_test].column_names,
)
self.tokenized_inputs = self.intermediate_dataset.map(
self.prepare_train_features,
batched=True,
remove_columns=self.intermediate_dataset[temp_key_train].column_names,
)
self.test_tokenized_inputs = self.intermediate_test_dataset.map(
self.prepare_test_features,
batched=True,
remove_columns=self.intermediate_test_dataset[temp_key_test].column_names,
)
def create_train_features(self, examples):
features = {
"context": [],
"id": [],
"question": [],
"title": [],
"start_positions": [],
"end_positions": [],
}
id = 0
# print(examples)
for row_number in range(len(examples["text"])):
context = examples["text"][row_number]
question = "offense"
title = context.split(" ")[0]
start_positions = []
end_positions = []
span = eval(examples["spans"][row_number])
contiguous_spans = _contiguous_ranges(span)
for lst in contiguous_spans:
lst = list(lst)
dict_to_write = {}
start_positions.append(lst[0])
end_positions.append(lst[1])
features["context"].append(context)
features["id"].append(str(id))
features["question"].append(question)
features["title"].append(title)
features["start_positions"].append(start_positions)
features["end_positions"].append(end_positions)
id += 1
return features
def create_test_features(self, examples):
features = {"context": [], "id": [], "question": [], "title": []}
id = 0
for row_number in range(len(examples["text"])):
context = examples["text"][row_number]
question = "offense"
title = context.split(" ")[0]
features["context"].append(context)
features["id"].append(str(id))
features["question"].append(question)
features["title"].append(title)
id += 1
return features
def prepare_train_features(self, examples):
"""Generate tokenized features from examples.
Args:
examples (dict): The examples to be tokenized.
Returns:
transformers.tokenization_utils_base.BatchEncoding:
The tokenized features/examples after processing.
"""
# Tokenize our examples with truncation and padding, but keep the
# overflows using a stride. This results in one example possible
# giving several features when a context is long, each of those
# features having a context that overlaps a bit the context
# of the previous feature.
pad_on_right = self.tokenizer.padding_side == "right"
print("### Batch Tokenizing Examples ###")
tokenized_examples = self.tokenizer(
examples["question" if pad_on_right else "context"],
examples["context" if pad_on_right else "question"],
**dict(self.config.tokenizer_params),
)
# Since one example might give us several features if it has
# a long context, we need a map from a feature to
# its corresponding example. This key gives us just that.
sample_mapping = tokenized_examples.pop("overflow_to_sample_mapping")
# The offset mappings will give us a map from token to
# character position in the original context. This will
# help us compute the start_positions and end_positions.
offset_mapping = tokenized_examples.pop("offset_mapping")
# Let's label those examples!
tokenized_examples["start_positions"] = []
tokenized_examples["end_positions"] = []
for i, offsets in enumerate(offset_mapping):
# We will label impossible answers with the index of the CLS token.
input_ids = tokenized_examples["input_ids"][i]
# Grab the sequence corresponding to that example
# (to know what is the context and what is the question).
sequence_ids = tokenized_examples.sequence_ids(i)
# One example can give several spans, this is the index of
# the example containing this span of text.
sample_index = sample_mapping[i]
start_positions = examples["start_positions"][sample_index]
end_positions = examples["end_positions"][sample_index]
start_positions_token_wise = [0 for x in range(len(input_ids))]
end_positions_token_wise = [0 for x in range(len(input_ids))]
# If no answers are given, set the cls_index as answer.
if len(start_positions) != 0:
for position in range(len(start_positions)):
start_char = start_positions[position]
end_char = end_positions[position] + 1
# Start token index of the current span in the text.
token_start_index = 0
while sequence_ids[token_start_index] != (1 if pad_on_right else 0):
token_start_index += 1
# End token index of the current span in the text.
token_end_index = len(input_ids) - 1
while sequence_ids[token_end_index] != (1 if pad_on_right else 0):
token_end_index -= 1
# Detect if the answer is out of the span (in which case we continue).
if not (
offsets[token_start_index][0] <= start_char
and offsets[token_end_index][1] >= end_char
):
continue
else:
# Otherwise move the token_start_index and token_end_index to the two ends of the answer.
# Note: we could go after the last offset if the answer is the last word (edge case).
while (
token_start_index < len(offsets)
and offsets[token_start_index][0] <= start_char
):
token_start_index += 1
start_positions_token_wise[token_start_index - 1] = 1
while offsets[token_end_index][1] >= end_char:
token_end_index -= 1
end_positions_token_wise[token_end_index + 1] = 1
tokenized_examples["start_positions"].append(start_positions_token_wise)
tokenized_examples["end_positions"].append(start_positions_token_wise)
return tokenized_examples
def prepare_test_features(self, examples):
"""Generate tokenized validation features from examples.
Args:
examples (dict): The validation examples to be tokenized.
Returns:
transformers.tokenization_utils_base.BatchEncoding:
The tokenized features/examples for validation set after processing.
"""
# Tokenize our examples with truncation and maybe
# padding, but keep the overflows using a stride.
# This results in one example possible giving several features
# when a context is long, each of those features having a
# context that overlaps a bit the context of the previous feature.
print("### Tokenizing Validation Examples")
pad_on_right = self.tokenizer.padding_side == "right"
tokenized_examples = self.tokenizer(
examples["question" if pad_on_right else "context"],
examples["context" if pad_on_right else "question"],
**dict(self.config.tokenizer_params),
)
# Since one example might give us several features if it has a long context,
# we need a map from a feature to its corresponding example. This key gives us just that.
sample_mapping = tokenized_examples.pop("overflow_to_sample_mapping")
# We keep the example_id that gave us this feature and we will store the offset mappings.
tokenized_examples["example_id"] = []
for i in range(len(tokenized_examples["input_ids"])):
# Grab the sequence corresponding to that example
# (to know what is the context and what is the question).
sequence_ids = tokenized_examples.sequence_ids(i)
context_index = 1 if pad_on_right else 0
# One example can give several spans,
# this is the index of the example containing this span of text.
sample_index = sample_mapping[i]
tokenized_examples["example_id"].append(str(examples["id"][sample_index]))
# Set to None the offset_mapping that are not part
# of the context so it's easy to determine if a token
# position is part of the context or not.
tokenized_examples["offset_mapping"][i] = [
(o if sequence_ids[k] == context_index else None)
for k, o in enumerate(tokenized_examples["offset_mapping"][i])
]
return tokenized_examples
|