wav2vec-bert-2.0-even-pakendorf

<!-- This model card has been generated automatically according to the information the Trainer had access to. You should probably proofread and complete it, then remove this comment. -->

wav2vec-bert-2.0-even-pakendorf-0406-1347

This model is a fine-tuned version of facebook/w2v-bert-2.0 on the audiofolder dataset. It achieves the following results on the evaluation set:

• Cer: 0.2128
• Loss: inf
• Wer: 0.5969

Model description

How to use:

from transformers import AutoModelForCTC, Wav2Vec2BertProcessor

model = AutoModelForCTC.from_pretrained("tbkazakova/wav2vec-bert-2.0-even-pakendorf")
processor = Wav2Vec2BertProcessor.from_pretrained("tbkazakova/wav2vec-bert-2.0-even-pakendorf")

data, sampling_rate = librosa.load('audio.wav')
librosa.resample(data, orig_sr=sampling_rate, target_sr=16000)
logits = model(torch.tensor(processor(data,
                                      sampling_rate=16000).input_features[0]).unsqueeze(0)).logits

pred_ids = torch.argmax(logits, dim=-1)[0]
print(processor.decode(pred_ids))

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 5e-05
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 16
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 500
• num_epochs: 10
• mixed_precision_training: Native AMP

Training results

Framework versions

• Transformers 4.41.2
• Pytorch 2.3.0+cu121
• Datasets 2.19.2
• Tokenizers 0.19.1