metadata

base_model: openai/whisper-small
datasets:
  - arrow
license: apache-2.0
metrics:
  - wer
tags:
  - generated_from_trainer
model-index:
  - name: dialect
    results:
      - task:
          type: automatic-speech-recognition
          name: Automatic Speech Recognition
        dataset:
          name: arrow
          type: arrow
          config: default
          split: train
          args: default
        metrics:
          - type: wer
            value: 0
            name: Wer

dialect

This model is a fine-tuned version of openai/whisper-small on the arrow dataset. It achieves the following results on the evaluation set:

Loss: 0.0001
Wer: 0.0

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: 1
eval_batch_size: 1
seed: 42
optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
lr_scheduler_type: linear
lr_scheduler_warmup_steps: 100
num_epochs: 5
mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.0444	0.2404	1000	0.0239	3.75
0.0115	0.4808	2000	0.0157	1.5385
0.0047	0.7212	3000	0.0844	4.4231
0.0153	0.9615	4000	0.0050	0.6731
0.0192	1.2019	5000	0.0017	0.1923
0.0	1.4423	6000	0.0075	0.8654
0.0	1.6827	7000	0.0002	0.0962
0.0274	1.9231	8000	0.0195	2.2115
0.0	2.1635	9000	0.0179	1.0577
0.0402	2.4038	10000	0.0020	0.0962
0.0161	2.6442	11000	0.0050	0.3846
0.0009	2.8846	12000	0.0048	0.2885
0.0	3.125	13000	0.0031	0.1923
0.0	3.3654	14000	0.0029	0.1923
0.0	3.6058	15000	0.0029	0.1923
0.0	3.8462	16000	0.0043	0.1923
0.0	4.0865	17000	0.0008	0.0962
0.0	4.3269	18000	0.0000	0.0
0.0	4.5673	19000	0.0000	0.0
0.0	4.8077	20000	0.0001	0.0

Framework versions

Transformers 4.41.2
Pytorch 2.3.1+cu121
Datasets 2.20.0
Tokenizers 0.19.1