Spaces:
Sleeping
Sleeping
File size: 2,360 Bytes
1e19e28 cb568f0 1e19e28 b873cb9 57ffa19 1e19e28 b873cb9 1e19e28 b873cb9 1e19e28 b873cb9 1e19e28 ee0f30d 62b1ca5 57ffa19 62b1ca5 ee0f30d 62b1ca5 |
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 |
from torch.utils.data import IterableDataset
def count_lines(input_path: str) -> int:
with open(input_path, "r", encoding="utf8") as f:
return sum(1 for _ in f)
class DatasetReader(IterableDataset):
def __init__(self, filename, tokenizer, max_length=128):
self.filename = filename
self.tokenizer = tokenizer
self.max_length = max_length
self.current_line = 0
self.total_lines = count_lines(filename)
print(f"{self.total_lines} lines in {filename}")
def preprocess(self, text: str):
self.current_line += 1
text = text.rstrip().strip()
if len(text) == 0:
print(f"Warning: empty sentence at line {self.current_line}")
return self.tokenizer(
text,
padding=False,
truncation=True,
max_length=self.max_length,
return_tensors=None,
)
def __iter__(self):
file_itr = open(self.filename, "r", encoding="utf8")
mapped_itr = map(self.preprocess, file_itr)
return mapped_itr
def __len__(self):
return self.total_lines
class ParallelTextReader(IterableDataset):
def __init__(self, pred_path: str, gold_path: str):
self.pred_path = pred_path
self.gold_path = gold_path
pref_filename_lines = count_lines(pred_path)
gold_path_lines = count_lines(gold_path)
assert pref_filename_lines == gold_path_lines, (
f"Lines in {pred_path} and {gold_path} do not match "
f"{pref_filename_lines} vs {gold_path_lines}"
)
self.num_sentences = gold_path_lines
self.current_line = 0
def preprocess(self, pred: str, gold: str):
self.current_line += 1
pred = pred.rstrip().strip()
gold = gold.rstrip().strip()
if len(pred) == 0:
print(f"Warning: Pred empty sentence at line {self.current_line}")
if len(gold) == 0:
print(f"Warning: Gold empty sentence at line {self.current_line}")
return pred, [gold]
def __iter__(self):
pred_itr = open(self.pred_path, "r", encoding="utf8")
gold_itr = open(self.gold_path, "r", encoding="utf8")
mapped_itr = map(self.preprocess, pred_itr, gold_itr)
return mapped_itr
def __len__(self):
return self.num_sentences
|