Hugging Face Transformers

os.environ["WANDB_PROJECT"] = "<my-project-name>"  # nommez votre projet W&B
os.environ["WANDB_LOG_MODEL"] = "checkpoint"  # journaliser tous les points de contrôle du modèle

from transformers import TrainingArguments, Trainer

args = TrainingArguments(..., report_to="wandb")  # activer la journalisation W&B
trainer = Trainer(..., args=args)

run = wandb.init()
trainer.train()  # démarrer l'entraînement et la journalisation vers W&B

# analyse post-entraînement, tests, autre code enregistré

run.finish()

# Créer un nouveau run
with wandb.init(project="amazon_sentiment_analysis") as run:
    # Indiquer le nom et la version de l'Artifact
    my_model_name = "model-bert-base-high-lr:latest"
    my_model_artifact = run.use_artifact(my_model_name)

    # Télécharger les poids du modèle dans un dossier et retourner le chemin
    model_dir = my_model_artifact.download()

    # Charger votre modèle Hugging Face depuis ce dossier
    #  en utilisant la même classe de modèle
    model = AutoModelForSequenceClassification.from_pretrained(
        model_dir, num_labels=num_labels
    )

    # Effectuer un entraînement supplémentaire ou exécuter l'inférence

last_run_id = "xxxxxxxx"  # fetch the run_id from your wandb workspace

# resume the wandb run from the run_id
with wandb.init(
    project=os.environ["WANDB_PROJECT"],
    id=last_run_id,
    resume="must",
) as run:
    # Connect an Artifact to the run
    my_checkpoint_name = f"checkpoint-{last_run_id}:latest"
    my_checkpoint_artifact = run.use_artifact(my_checkpoint_name)

    # Download checkpoint to a folder and return the path
    checkpoint_dir = my_checkpoint_artifact.download()

    # reinitialize your model and trainer
    model = AutoModelForSequenceClassification.from_pretrained(
        "<model-name>", num_labels=num_labels
    )
    # vos arguments d'entraînement ici.
    training_args = TrainingArguments()

    trainer = Trainer(model=model, args=training_args)

    # make sure use the checkpoint dir to resume training from the checkpoint
    trainer.train(resume_from_checkpoint=checkpoint_dir)

# Instancier le Trainer normalement
trainer = Trainer()

# Instancier le nouveau callback de journalisation en lui passant l'objet Trainer
evals_callback = WandbEvalsCallback(trainer, tokenizer, ...)

# Ajouter le callback au Trainer
trainer.add_callback(evals_callback)

# Lancer l'entraînement du Trainer normalement
trainer.train()

from transformers.integrations import WandbCallback
import pandas as pd


def decode_predictions(tokenizer, predictions):
    labels = tokenizer.batch_decode(predictions.label_ids)
    logits = predictions.predictions.argmax(axis=-1)
    prediction_text = tokenizer.batch_decode(logits)
    return {"labels": labels, "predictions": prediction_text}


class WandbPredictionProgressCallback(WandbCallback):
    """Custom WandbCallback to log model predictions during training.

    This callback logs model predictions and labels to a wandb.Table at each
    logging step during training. It allows to visualize the
    model predictions as the training progresses.

    Attributes:
        trainer (Trainer): The Hugging Face Trainer instance.
        tokenizer (AutoTokenizer): The tokenizer associated with the model.
        sample_dataset (Dataset): A subset of the validation dataset
          for generating predictions.
        num_samples (int, optional): Number of samples to select from
          the validation dataset for generating predictions. Defaults to 100.
        freq (int, optional): Frequency of logging. Defaults to 2.
    """

    def __init__(self, trainer, tokenizer, val_dataset, num_samples=100, freq=2):
        """Initializes the WandbPredictionProgressCallback instance.

        Args:
            trainer (Trainer): The Hugging Face Trainer instance.
            tokenizer (AutoTokenizer): The tokenizer associated
              with the model.
            val_dataset (Dataset): The validation dataset.
            num_samples (int, optional): Number of samples to select from
              the validation dataset for generating predictions.
              Defaults to 100.
            freq (int, optional): Frequency of logging. Defaults to 2.
        """
        super().__init__()
        self.trainer = trainer
        self.tokenizer = tokenizer
        self.sample_dataset = val_dataset.select(range(num_samples))
        self.freq = freq

    def on_evaluate(self, args, state, control, **kwargs):
        super().on_evaluate(args, state, control, **kwargs)
        # contrôler la fréquence de journalisation en journalisant les prédictions
        # toutes les `freq` époques
        if state.epoch % self.freq == 0:
            # générer les prédictions
            predictions = self.trainer.predict(self.sample_dataset)
            # décoder les prédictions et les labels
            predictions = decode_predictions(self.tokenizer, predictions)
            # ajouter les prédictions à un wandb.Table
            predictions_df = pd.DataFrame(predictions)
            predictions_df["epoch"] = state.epoch
            records_table = self._wandb.Table(dataframe=predictions_df)
            # journaliser le tableau dans wandb
            self._wandb.log({"sample_predictions": records_table})


# D'abord, instancier le Trainer
trainer = Trainer(
    model=model,
    args=training_args,
    train_dataset=lm_datasets["train"],
    eval_dataset=lm_datasets["validation"],
)

# Instancier le WandbPredictionProgressCallback
progress_callback = WandbPredictionProgressCallback(
    trainer=trainer,
    tokenizer=tokenizer,
    val_dataset=lm_dataset["validation"],
    num_samples=10,
    freq=2,
)

# Ajouter le callback au trainer
trainer.add_callback(progress_callback)

wandb.init(
    project="amazon_sentiment_analysis",
    name="bert-base-high-lr",
    tags=["baseline", "high-lr"],
    group="bert",
)