Préparer des données (PyTorch)¶

!pip install datasets transformers[sentencepiece]

import torch
from transformers import AdamW, AutoTokenizer, AutoModelForSequenceClassification

# Comme avant
checkpoint = "camembert-base"
tokenizer = AutoTokenizer.from_pretrained(checkpoint)
model = AutoModelForSequenceClassification.from_pretrained(checkpoint)
sequences = [
    "J'ai attendu un cours d'HuggingFace toute ma vie.", 
    "Je déteste tellement ça !"]
batch = tokenizer(sequences, padding=True, truncation=True, return_tensors="pt")

# C'est nouveau
batch["labels"] = torch.tensor([1, 1])

optimizer = AdamW(model.parameters())
loss = model(**batch).loss
loss.backward()
optimizer.step()

from datasets import load_dataset

raw_datasets = load_dataset("paws-x", "fr")
raw_datasets

raw_train_dataset = raw_datasets["train"]
raw_train_dataset[0]

raw_train_dataset.features

from transformers import AutoTokenizer

checkpoint = "camembert-base"
tokenizer = AutoTokenizer.from_pretrained(checkpoint)
tokenized_sentences_1 = tokenizer(raw_datasets["train"]["sentence1"])
tokenized_sentences_2 = tokenizer(raw_datasets["train"]["sentence2"])

inputs = tokenizer("C'est la première phrase.", "C'est la deuxième.")
inputs

tokenizer.convert_ids_to_tokens(inputs["input_ids"])

tokenized_dataset = tokenizer(
    raw_datasets["train"]["sentence1"],
    raw_datasets["train"]["sentence2"],
    padding=True,
    truncation=True,
)

def tokenize_function(example):
    return tokenizer(example["sentence1"], example["sentence2"], truncation=True)

tokenized_datasets = raw_datasets.map(tokenize_function, batched=True)
tokenized_datasets

from transformers import DataCollatorWithPadding

data_collator = DataCollatorWithPadding(tokenizer=tokenizer)

samples = tokenized_datasets["train"][:8]
samples = {k: v for k, v in samples.items() if k not in ["idx", "sentence1", "sentence2"]}
[len(x) for x in samples["input_ids"]]

batch = data_collator(samples)
{k: v.shape for k, v in batch.items()}