Traduction (PyTorch)

!pip install datasets transformers[sentencepiece]
!pip install accelerate
# Pour exécuter l'entraînement sur TPU, vous devez décommenter la ligne suivante :
# !pip install cloud-tpu-client==0.10 torch==1.9.0 https://storage.googleapis.com/tpu-pytorch/wheels/torch_xla-1.9-cp37-cp37m-linux_x86_64.whl
!apt install git-lfs

!git config --global user.email "[email protected]"
!git config --global user.name "Your Name"

from huggingface_hub import notebook_login

notebook_login()

from datasets import load_dataset, load_metric

raw_datasets = load_dataset("kde4", lang1="en", lang2="fr")

raw_datasets

split_datasets = raw_datasets["train"].train_test_split(train_size=0.9, seed=20)
split_datasets

split_datasets["validation"] = split_datasets.pop("test")

split_datasets["train"][1]["translation"]

from transformers import pipeline

model_checkpoint = "Helsinki-NLP/opus-mt-en-fr"
translator = pipeline("translation", model=model_checkpoint)
translator("Default to expanded threads")

split_datasets["train"][172]["translation"]

translator(
    "Unable to import %1 using the OFX importer plugin. This file is not the correct format."
)

from transformers import AutoTokenizer

model_checkpoint = "Helsinki-NLP/opus-mt-en-fr"
tokenizer = AutoTokenizer.from_pretrained(model_checkpoint, return_tensors="tf")

en_sentence = split_datasets["train"][1]["translation"]["en"]
fr_sentence = split_datasets["train"][1]["translation"]["fr"]

inputs = tokenizer(en_sentence)
with tokenizer.as_target_tokenizer():
    targets = tokenizer(fr_sentence)

wrong_targets = tokenizer(fr_sentence)
print(tokenizer.convert_ids_to_tokens(wrong_targets["input_ids"]))
print(tokenizer.convert_ids_to_tokens(targets["input_ids"]))

max_input_length = 128
max_target_length = 128


def preprocess_function(examples):
    inputs = [ex["en"] for ex in examples["translation"]]
    targets = [ex["fr"] for ex in examples["translation"]]
    model_inputs = tokenizer(inputs, max_length=max_input_length, truncation=True)

    # Configurer le tokenizer pour les cibles
    with tokenizer.as_target_tokenizer():
        labels = tokenizer(targets, max_length=max_target_length, truncation=True)

    model_inputs["labels"] = labels["input_ids"]
    return model_inputs

tokenized_datasets = split_datasets.map(
    preprocess_function,
    batched=True,
    remove_columns=split_datasets["train"].column_names,
)

from transformers import AutoModelForSeq2SeqLM

model = AutoModelForSeq2SeqLM.from_pretrained(model_checkpoint)

from transformers import DataCollatorForSeq2Seq

data_collator = DataCollatorForSeq2Seq(tokenizer, model=model)

batch = data_collator([tokenized_datasets["train"][i] for i in range(1, 3)])
batch.keys()

batch["labels"]

batch["decoder_input_ids"]

for i in range(1, 3):
    print(tokenized_datasets["train"][i]["labels"])

!pip install sacrebleu

from datasets import load_metric

metric = load_metric("sacrebleu")

predictions = [
    "This plugin lets you translate web pages between several languages automatically."
]
references = [
    [
        "This plugin allows you to automatically translate web pages between several languages."
    ]
]
metric.compute(predictions=predictions, references=references)

predictions = ["This This This This"]
references = [
    [
        "This plugin allows you to automatically translate web pages between several languages."
    ]
]
metric.compute(predictions=predictions, references=references)

predictions = ["This plugin"]
references = [
    [
        "This plugin allows you to automatically translate web pages between several languages."
    ]
]
metric.compute(predictions=predictions, references=references)

import numpy as np


def compute_metrics(eval_preds):
    preds, labels = eval_preds
    # Dans le cas où le modèle retourne plus que les logits de prédiction
    if isinstance(preds, tuple):
        preds = preds[0]

    decoded_preds = tokenizer.batch_decode(preds, skip_special_tokens=True)

    # Remplacer les -100 dans les étiquettes car nous ne pouvons pas les décoder
    labels = np.where(labels != -100, labels, tokenizer.pad_token_id)
    decoded_labels = tokenizer.batch_decode(labels, skip_special_tokens=True)

    # Quelques post-traitements simples
    decoded_preds = [pred.strip() for pred in decoded_preds]
    decoded_labels = [[label.strip()] for label in decoded_labels]

    result = metric.compute(predictions=decoded_preds, references=decoded_labels)
    return {"bleu": result["score"]}

from huggingface_hub import notebook_login

notebook_login()

from transformers import Seq2SeqTrainingArguments

args = Seq2SeqTrainingArguments(
    f"marian-finetuned-kde4-en-to-fr",
    evaluation_strategy="no",
    save_strategy="epoch",
    learning_rate=2e-5,
    per_device_train_batch_size=32,
    per_device_eval_batch_size=64,
    weight_decay=0.01,
    save_total_limit=3,
    num_train_epochs=3,
    predict_with_generate=True,
    fp16=True,
    push_to_hub=True,
)

from transformers import Seq2SeqTrainer

trainer = Seq2SeqTrainer(
    model,
    args,
    train_dataset=tokenized_datasets["train"],
    eval_dataset=tokenized_datasets["validation"],
    data_collator=data_collator,
    tokenizer=tokenizer,
    compute_metrics=compute_metrics,
)

trainer.evaluate(max_length=max_target_length)

trainer.train()

trainer.evaluate(max_length=max_target_length)

trainer.push_to_hub(tags="translation", commit_message="Training complete")

from torch.utils.data import DataLoader

tokenized_datasets.set_format("torch")
train_dataloader = DataLoader(
    tokenized_datasets["train"],
    shuffle=True,
    collate_fn=data_collator,
    batch_size=8,
)
eval_dataloader = DataLoader(
    tokenized_datasets["validation"], collate_fn=data_collator, batch_size=8
)

model = AutoModelForSeq2SeqLM.from_pretrained(model_checkpoint)

from transformers import AdamW

optimizer = AdamW(model.parameters(), lr=2e-5)

from accelerate import Accelerator

accelerator = Accelerator()
model, optimizer, train_dataloader, eval_dataloader = accelerator.prepare(
    model, optimizer, train_dataloader, eval_dataloader
)

from transformers import get_scheduler

num_train_epochs = 3
num_update_steps_per_epoch = len(train_dataloader)
num_training_steps = num_train_epochs * num_update_steps_per_epoch

lr_scheduler = get_scheduler(
    "linear",
    optimizer=optimizer,
    num_warmup_steps=0,
    num_training_steps=num_training_steps,
)

from huggingface_hub import Repository, get_full_repo_name

model_name = "marian-finetuned-kde4-en-to-fr-accelerate"
repo_name = get_full_repo_name(model_name)
repo_name

output_dir = "marian-finetuned-kde4-en-to-fr-accelerate"
repo = Repository(output_dir, clone_from=repo_name)

def postprocess(predictions, labels):
    predictions = predictions.cpu().numpy()
    labels = labels.cpu().numpy()

    decoded_preds = tokenizer.batch_decode(predictions, skip_special_tokens=True)

    # Remplacez -100 dans les étiquettes car nous ne pouvons pas les décoder
    labels = np.where(labels != -100, labels, tokenizer.pad_token_id)
    decoded_labels = tokenizer.batch_decode(labels, skip_special_tokens=True)

    # Quelques post-traitements simples
    decoded_preds = [pred.strip() for pred in decoded_preds]
    decoded_labels = [[label.strip()] for label in decoded_labels]
    return decoded_preds, decoded_labels

from tqdm.auto import tqdm
import torch

progress_bar = tqdm(range(num_training_steps))

for epoch in range(num_train_epochs):
    # Entraînement
    model.train()
    for batch in train_dataloader:
        outputs = model(**batch)
        loss = outputs.loss
        accelerator.backward(loss)

        optimizer.step()
        lr_scheduler.step()
        optimizer.zero_grad()
        progress_bar.update(1)

    # Evaluation
    model.eval()
    for batch in tqdm(eval_dataloader):
        with torch.no_grad():
            generated_tokens = accelerator.unwrap_model(model).generate(
                batch["input_ids"],
                attention_mask=batch["attention_mask"],
                max_length=128,
            )
        labels = batch["labels"]

        # Nécessaire pour rembourrer les prédictions et les étiquettes à rassembler
        generated_tokens = accelerator.pad_across_processes(
            generated_tokens, dim=1, pad_index=tokenizer.pad_token_id
        )
        labels = accelerator.pad_across_processes(labels, dim=1, pad_index=-100)

        predictions_gathered = accelerator.gather(generated_tokens)
        labels_gathered = accelerator.gather(labels)

        decoded_preds, decoded_labels = postprocess(predictions_gathered, labels_gathered)
        metric.add_batch(predictions=decoded_preds, references=decoded_labels)

    results = metric.compute()
    print(f"epoch {epoch}, BLEU score: {results['score']:.2f}")

    # Sauvegarder et télécharger
    accelerator.wait_for_everyone()
    unwrapped_model = accelerator.unwrap_model(model)
    unwrapped_model.save_pretrained(output_dir, save_function=accelerator.save)
    if accelerator.is_main_process:
        tokenizer.save_pretrained(output_dir)
        repo.push_to_hub(
            commit_message=f"Training in progress epoch {epoch}", blocking=False
        )

from transformers import pipeline

# Remplacer par votre propre checkpoint
model_checkpoint = "huggingface-course/marian-finetuned-kde4-en-to-fr"
translator = pipeline("translation", model=model_checkpoint)
translator("Default to expanded threads")

translator(
    "Unable to import %1 using the OFX importer plugin. This file is not the correct format."
)