Perrie/whisper-tiny-epoch-10

Whisper tiny epoch test - Perrie

This model is a fine-tuned version of openai/whisper-tiny on the Common Voice 11.0 dataset. It achieves the following results on the evaluation set:

• Loss: 0.4754
• Cer: 57.5895

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: 4
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 4
• 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: 700
• num_epochs: 10

Training results

Training Loss	Epoch	Step	Validation Loss	Cer
0.4064	0.99	700	0.4476	33.9306
0.3595	1.99	1400	0.4218	29.0735
0.2349	2.98	2100	0.4197	25.9553
0.1507	3.97	2800	0.4237	25.9104
0.1003	4.96	3500	0.4309	35.9495
0.0601	5.96	4200	0.4442	29.8821
0.0481	6.95	4900	0.4551	36.7132
0.0296	7.94	5600	0.4640	38.9224
0.0207	8.94	6300	0.4721	47.1645
0.0169	9.93	7000	0.4754	57.5895

Framework versions

• Transformers 4.36.0.dev0
• Pytorch 1.12.1
• Datasets 2.14.6
• Tokenizers 0.14.1