Whisper
Collection
Whisper model fine-tuned on different languages • 10 items • Updated
Automatic Speech Recognition
Transformers
PyTorch
Safetensors
Italian
whisper
whisper-event
Generated from Trainer
Eval Results (legacy)
# Load model directly
from transformers import AutoProcessor, AutoModelForSpeechSeq2Seq
processor = AutoProcessor.from_pretrained("ALM/whisper-it-medium-augmented")
model = AutoModelForSpeechSeq2Seq.from_pretrained("ALM/whisper-it-medium-augmented")Quick Links
Whisper Medium Italian - Robust
This model is a fine-tuned version of openai/whisper-medium on the mozilla-foundation/common_voice_11_0 it dataset. It achieves the following results on the evaluation set:
- Loss: 0.1388
- WER (Augmented Test): 7.65
IMPORTANT The model has been trained using data augmentation to improve its generalization capabilities and robustness. The results on the eval set during training are biased towards data augmentation applied to evaluation data.
Results on eval set
- Mozilla CV 11.0 - Italian: 6.60 WER (using official script)
Model description
More information needed
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: 64
- eval_batch_size: 32
- seed: 42
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 500
- training_steps: 7500
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer |
|---|---|---|---|---|
| 0.226 | 0.33 | 2500 | 0.2779 | 14.6642 |
| 0.1278 | 1.03 | 5000 | 0.1818 | 10.2049 |
| 0.0304 | 1.36 | 7500 | 0.1388 | 7.5544 |
Framework versions
- Transformers 4.25.1
- Pytorch 1.13.0+cu117
- Datasets 2.7.1
- Tokenizers 0.13.2
- Downloads last month
- 17
Collection including ALM/whisper-it-medium-augmented
Evaluation results
- Wer on mozilla-foundation/common_voice_11_0 itself-reported7.651
- WER on mozilla-foundation/common_voice_11_0 itself-reported6.600
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("automatic-speech-recognition", model="ALM/whisper-it-medium-augmented")