metadata

license: apache-2.0
tags:
  - generated_from_trainer
metrics:
  - wer
base_model: nandovallec/whisper-tiny-bg-l
model-index:
  - name: whisper-tiny-order
    results: []

whisper-tiny-order

This model is a fine-tuned version of nandovallec/whisper-tiny-bg-l on the None dataset. It achieves the following results on the evaluation set:

Loss: 0.0015
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: 16
eval_batch_size: 8
seed: 42
gradient_accumulation_steps: 16
total_train_batch_size: 256
optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
lr_scheduler_type: linear
training_steps: 150
mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.5799	5.0	5	1.8414	116.3934
0.2647	10.0	10	0.7719	61.4754
0.1199	15.0	15	0.3593	34.4262
0.063	20.0	20	0.1827	18.8525
0.0257	25.0	25	0.0710	4.9180
0.0103	30.0	30	0.0294	0.8197
0.0045	35.0	35	0.0149	0.0
0.0028	40.0	40	0.0094	0.0
0.0019	45.0	45	0.0064	0.0
0.0014	50.0	50	0.0048	0.0
0.0011	55.0	55	0.0038	0.0
0.0009	60.0	60	0.0032	0.0
0.0008	65.0	65	0.0028	0.0
0.0007	70.0	70	0.0025	0.0
0.0006	75.0	75	0.0023	0.0
0.0006	80.0	80	0.0022	0.0
0.0006	85.0	85	0.0021	0.0
0.0005	90.0	90	0.0020	0.0
0.0005	95.0	95	0.0019	0.0
0.0005	100.0	100	0.0018	0.0
0.0005	105.0	105	0.0017	0.0
0.0005	110.0	110	0.0017	0.0
0.0004	115.0	115	0.0017	0.0
0.0004	120.0	120	0.0016	0.0
0.0004	125.0	125	0.0016	0.0
0.0004	130.0	130	0.0016	0.0
0.0004	135.0	135	0.0016	0.0
0.0004	140.0	140	0.0015	0.0
0.0004	145.0	145	0.0015	0.0
0.0004	150.0	150	0.0015	0.0

Framework versions

Transformers 4.38.1
Pytorch 2.1.0+cu121
Datasets 2.18.0
Tokenizers 0.15.2