sensai/examples/transformers.ipynb

In [ ]:

from datasets import load_dataset, Features
from transformers import AutoTokenizer, AutoModelForSequenceClassification, Trainer, TrainingArguments
import os
from datasets import ClassLabel, Value

# https://huggingface.co/docs/datasets/loading_datasets.html

DATASET_DIR = os.environ.get("DATASET_DIR", "../input/sensai")
print("DATASET_DIR", DATASET_DIR)

In [ ]:

dataset = load_dataset("holodata/sensai", features=Features(
                {
                    "body": Value("string"),
                    "toxic": ClassLabel(num_classes=2, names=['0', '1'])
                }
            ))
dataset = dataset['train']

In [ ]:

dataset.features

In [ ]:

model = AutoModelForSequenceClassification.from_pretrained("bert-base-cased", num_labels=2)
tokenizer = AutoTokenizer.from_pretrained("bert-base-cased")

In [ ]:

samples = dataset.shuffle().select(range(50000))

def tokenize_function(examples):
    return tokenizer(examples["body"], padding="max_length", truncation=True)

tokenized_datasets = samples.map(tokenize_function, batched=True)
tokenized_datasets.rename_column_("toxic", "label")

In [ ]:

splitset = tokenized_datasets.train_test_split(0.2)
splitset

In [ ]:

training_args = TrainingArguments("test_trainer")
trainer = Trainer(
    model=model, args=training_args, train_dataset=splitset['train'], eval_dataset=splitset['test']
)

In [ ]:

trainer.train(resume_from_checkpoint=True)

In [ ]: