# Wav2Vec not getting fine tuned by writing a loop using dataloader

**URL:** <https://discuss.huggingface.co/t/wav2vec-not-getting-fine-tuned-by-writing-a-loop-using-dataloader/18745>\
**Category:** Models\
**Created:** [June 6, 2022, 2:06am UTC](https://discuss.huggingface.co/t/wav2vec-not-getting-fine-tuned-by-writing-a-loop-using-dataloader/18745 "2022-06-06T02:06:22Z")\
**Posts on this page:** 1\
**Page:** 1

<div class="post-metadata">

**Author:** ![kamalojasv2000](https://avatars.discourse-cdn.com/v4/letter/k/9e8a1a/32.png) [@kamalojasv2000](https://discuss.huggingface.co/u/kamalojasv2000)\
**Post date:** [June 6, 2022, 2:06am UTC](https://discuss.huggingface.co/t/wav2vec-not-getting-fine-tuned-by-writing-a-loop-using-dataloader/18745/1 "2022-06-06T02:06:23Z")

</div>

```auto
        with torch.set_grad_enabled(True):

            model.train()
            for batch in dataloader:

                input = batch["input_val"].to(device)
                labels = batch["label"].to(device)

                optimizer.zero_grad()

                out = model(input)
                
                loss = loss_fn(out["logits"], labels)

                loss_epoch += loss.detach().cpu().item()
                num_steps += input.shape[0]

                loss.backward()
                
                optimizer.step()

```

here model is

```auto
model = AutoModelForAudioClassification.from_pretrained(
    "facebook/wav2vec2-conformer-rope-large", num_labels=num_labels, label2id=label2id, id2label=id2label
).to(device)

```

I got the exact same results as I got with an untrained model