metadata

language:
  - he
license: apache-2.0
base_model: openai/whisper-tiny
tags:
  - hf-asr-leaderboard
  - generated_from_trainer
datasets:
  - ivrit-ai/whisper-training
metrics:
  - wer
model-index:
  - name: Whisper Tiny Hebrew
    results:
      - task:
          name: Automatic Speech Recognition
          type: automatic-speech-recognition
        dataset:
          name: ivrit-ai/whisper-training
          type: ivrit-ai/whisper-training
          args: 'config: he, split: train'
        metrics:
          - name: Wer
            type: wer
            value: 55.88158581116328

Whisper Tiny Hebrew

This model is a fine-tuned version of openai/whisper-tiny on the ivrit-ai/whisper-training dataset.

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: 16
eval_batch_size: 8
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: 10000
mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.973	0.13	500	0.8480	77.6213
0.9024	0.25	1000	0.7710	67.9838
0.8049	0.38	1500	0.7499	66.7384
0.7221	0.5	2000	0.7092	64.7953
0.7464	0.63	2500	0.6939	62.7543
0.7396	0.75	3000	0.6839	62.5261
0.7336	0.88	3500	0.6716	61.2350
0.6118	1.01	4000	0.6512	58.4637
0.6299	1.13	4500	0.6564	60.1721
0.6318	1.26	5000	0.6475	58.8550
0.6315	1.38	5500	0.6361	58.9724
0.6081	1.51	6000	0.6321	57.1596
0.6487	1.63	6500	0.6459	58.5616
0.6481	1.76	7000	0.6298	56.9379
0.5833	1.88	7500	0.6303	57.8965
0.5689	2.01	8000	0.6305	56.1750
0.5223	2.14	8500	0.6335	56.6967
0.574	2.26	9000	0.6248	55.3730
0.5841	2.39	9500	0.6320	55.6273
0.5533	2.51	10000	0.6254	55.8816

Framework versions

Transformers 4.40.0.dev0
Pytorch 2.2.1
Datasets 2.18.0
Tokenizers 0.15.2