Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

Fix datasets #6

Open
wants to merge 2 commits into
base: main
Choose a base branch
from
Open

Fix datasets #6

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
24b8cb5
load only first files of each dataset
poedator Apr 4, 2024
d87c035
black
poedator Apr 4, 2024
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
65 changes: 39 additions & 26 deletions data_converter.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -3,48 +3,61 @@
from datasets import load_dataset
from transformers.utils import check_min_version
from transformers.utils.versions import require_version

check_min_version("4.28.0.dev0")

logger = get_logger(__name__)
require_version("datasets>=1.8.0", "To fix: pip install -r examples/pytorch/translation/requirements.txt")


def convert_wiki_dataset(tokenizer, seq_len = 256):
dataset = load_dataset("wikimedia/wikipedia", "20231101.en", split="train[0:2000]")
def convert_wiki_dataset(tokenizer, seq_len=256):
dataset = load_dataset("wikitext", "wikitext-2-raw-v1", split="train")

def tokenize_function(examples):
return tokenizer(examples["text"], return_tensors='pt',max_length=seq_len,padding=True,truncation=True)
dataset = dataset.map(tokenize_function, batched=True, remove_columns=['text'])
dataset.set_format(type='torch', columns=['input_ids', 'attention_mask'])
return tokenizer(examples["text"], return_tensors="pt", max_length=seq_len, padding=True, truncation=True)

dataset = dataset.map(tokenize_function, batched=True, remove_columns=["text"])
dataset.set_format(type="torch", columns=["input_ids", "attention_mask"])
return dataset

def convert_cnn_dataset(tokenizer, seq_len = 256):
dataset = load_dataset("cnn_dailymail", "1.0.0", split="test[0:2000]")

def convert_cnn_dataset(tokenizer, seq_len=256):
dataset = load_dataset(
path="cnn_dailymail",
data_files={"test": "1.0.0/test-00000-of-00001.parquet"},
split="test[0:2000]",
)

def tokenize_function(examples):
return tokenizer(examples["article"], return_tensors='pt',max_length=seq_len,padding=True,truncation=True)
dataset = dataset.map(tokenize_function, batched=True, remove_columns=['article'])
dataset.set_format(type='torch', columns=['input_ids', 'attention_mask'])
return tokenizer(examples["article"], return_tensors="pt", max_length=seq_len, padding=True, truncation=True)

dataset = dataset.map(tokenize_function, batched=True, remove_columns=["article"])
dataset.set_format(type="torch", columns=["input_ids", "attention_mask"])
return dataset

def convert_c4_dataset_eval(tokenizer, seq_len = 256):
dataset = load_dataset("c4", "en", split="validation[0:2000]")

def convert_c4_dataset_eval(tokenizer, seq_len=256):
dataset = load_dataset("allenai/c4", "allenai--c4", data_files={"train": "en/c4-train.00000-of-01024.json.gz"}, split="train")

def tokenize_function(examples):
return tokenizer(examples["text"], return_tensors='pt',max_length=seq_len,padding=True,truncation=True)
dataset = dataset.map(tokenize_function, batched=True, remove_columns=['text', 'timestamp', 'url'])
dataset.set_format(type='torch', columns=['input_ids', 'attention_mask'])
return tokenizer(examples["text"], return_tensors="pt", max_length=seq_len, padding=True, truncation=True)

dataset = dataset.map(tokenize_function, batched=True, remove_columns=["text", "timestamp", "url"])
dataset.set_format(type="torch", columns=["input_ids", "attention_mask"])
return dataset


def convert_dataset(tokenizer, file_path):
dataset = load_dataset("json", data_files=file_path, split="train")

def tokenize_function(examples):
input_ids = torch.Tensor(examples['input_ids'])
labels = input_ids.clone()
if tokenizer.pad_token_id is not None:
labels[labels == tokenizer.pad_token_id] = -100
ret = {
"input_ids": input_ids,
"labels": labels
}
return ret
dataset = dataset.map(tokenize_function, batched=True, remove_columns=['input_tokens'])
dataset.set_format(type='torch', columns=['input_ids', "labels"])
input_ids = torch.Tensor(examples["input_ids"])
labels = input_ids.clone()
if tokenizer.pad_token_id is not None:
labels[labels == tokenizer.pad_token_id] = -100
ret = {"input_ids": input_ids, "labels": labels}
return ret

dataset = dataset.map(tokenize_function, batched=True, remove_columns=["input_tokens"])
dataset.set_format(type="torch", columns=["input_ids", "labels"])
return dataset