Lacking config attribute of model after finetuning

Intermediate
1

When I’m finetuning my model I’m loosing that config attribute which is in the non-finetuned model. I’m pytorch lightening to define the parameters of the model which looks like this:

class ConvBertForSequenceClassification(pl.LightningModule):

  def __init__(self, n_classes: int, n_training_steps=None, n_warmup_steps=None):
    super().__init__()
    self.ConvBert = ConvBertModel.from_pretrained("sarnikowski/convbert-medium-small-da-cased", return_dict=True)
    self.classifier = nn.Linear(self.ConvBert.config.hidden_size, n_classes)
    self.n_training_steps = n_training_steps
    self.n_warmup_steps = n_warmup_steps
    self.criterion = nn.BCELoss()
    self.config = ()
    self.ConvBert.config.num_labels = 3

  def forward(self, input_ids, attention_mask, labels=None):
    output = self.ConvBert(input_ids, attention_mask=attention_mask)
    output = self.classifier(output.last_hidden_state[:, 0]) # output = self.classifier(output.pooler_output)#output = self.classifier(output.last_hidden_state[:, 0])
    output = torch.sigmoid(output)
    loss = 0
    if labels is not None:
        loss = self.criterion(output, labels)
    return loss, output

  def training_step(self, batch, batch_idx):
    input_ids = batch["input_ids"]
    attention_mask = batch["attention_mask"]
    labels = batch["labels"]
    loss, outputs = self(input_ids, attention_mask, labels)
    self.log("train_loss", loss, prog_bar=True, logger=True)
    return {"loss": loss, "predictions": outputs, "labels": labels}

  def validation_step(self, batch, batch_idx):
    input_ids = batch["input_ids"]
    attention_mask = batch["attention_mask"]
    labels = batch["labels"]
    loss, outputs = self(input_ids, attention_mask, labels)
    self.log("val_loss", loss, prog_bar=True, logger=True)
    return loss

  def test_step(self, batch, batch_idx):
    input_ids = batch["input_ids"]
    attention_mask = batch["attention_mask"]
    labels = batch["labels"]
    loss, outputs = self(input_ids, attention_mask, labels)
    self.log("test_loss", loss, prog_bar=True, logger=True)
    return loss

  def training_epoch_end(self, outputs):
    
    labels = []
    predictions = []
    for output in outputs:
      for out_labels in output["labels"].detach().cpu():
        labels.append(out_labels)
      for out_predictions in output["predictions"].detach().cpu():
        predictions.append(out_predictions)

    labels = torch.stack(labels).int()
    predictions = torch.stack(predictions)

    for i, name in enumerate(LABEL_COLUMNS):
      class_roc_auc = auroc(predictions[:, i], labels[:, i])
      self.logger.experiment.add_scalar(f"{name}_roc_auc/Train", class_roc_auc, self.current_epoch)

  def configure_optimizers(self):

    optimizer = AdamW(self.parameters(), lr=2e-5) #DEFINING LEARNING RATE

    scheduler = get_linear_schedule_with_warmup(
      optimizer,
      num_warmup_steps=self.n_warmup_steps,
      num_training_steps=self.n_training_steps
    )

    return dict(
      optimizer=optimizer,
      lr_scheduler=dict(
        scheduler=scheduler,
        interval='step'
      )
    )

After fine-tuning the model I’m running a function from transformers-interpret library which gives me an error due to the lack of a config attribute. Can anyone see what I’m doing wrong?

here’s a link to a colab: Google Colab