metadata
language:
- en
license: apache-2.0
base_model: openai/whisper-small
tags:
- generated_from_trainer
datasets:
- librispeech_asr
metrics:
- wer
model-index:
- name: SpeechGPT
results:
- task:
name: Automatic Speech Recognition
type: automatic-speech-recognition
dataset:
name: librispeech_asr
type: librispeech_asr
config: clean
split: None
args: 'config: clean, split: train'
metrics:
- name: Wer
type: wer
value: 23.544963481436394
SpeechGPT
This model is a fine-tuned version of openai/whisper-small on the librispeech_asr dataset. It achieves the following results on the evaluation set:
- Loss: 0.5354
- Wer: 23.5450
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: 1e-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: 1
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer |
|---|---|---|---|---|
| 1.0518 | 0.12 | 1000 | 0.7491 | 31.7103 |
| 0.8884 | 0.24 | 2000 | 0.6588 | 27.4003 |
| 0.8061 | 0.36 | 3000 | 0.6177 | 26.4569 |
| 0.8549 | 0.48 | 4000 | 0.5888 | 25.5002 |
| 0.7836 | 0.6 | 5000 | 0.5688 | 25.8939 |
| 0.691 | 0.72 | 6000 | 0.5542 | 24.1574 |
| 0.7044 | 0.84 | 7000 | 0.5429 | 23.5450 |
| 0.7309 | 0.97 | 8000 | 0.5354 | 23.5450 |
Framework versions
- Transformers 4.39.3
- Pytorch 2.2.2+cu121
- Datasets 2.19.0
- Tokenizers 0.15.2