whisper-tiny

This model is a fine-tuned version of openai/whisper-tiny on the PolyAI/minds14 dataset. It achieves the following results on the evaluation set:

• Loss: 0.5913
• Wer Ortho: 0.2340
• Wer: 0.2243

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: constant_with_warmup
• lr_scheduler_warmup_steps: 50
• training_steps: 500

Training results

Training Loss	Epoch	Step	Validation Loss	Wer Ortho	Wer
1.7357	2.0	50	0.7179	0.2947	0.2412
0.2772	4.0	100	0.4758	0.2404	0.2113
0.081	6.0	150	0.5069	0.2628	0.2282
0.02	8.0	200	0.5289	0.2564	0.2297
0.0044	10.0	250	0.5366	0.2452	0.2251
0.0018	12.0	300	0.5565	0.2404	0.2251
0.0011	14.0	350	0.5668	0.2388	0.2259
0.0009	16.0	400	0.5762	0.2364	0.2251
0.0007	18.0	450	0.5847	0.2348	0.2243
0.0006	20.0	500	0.5913	0.2340	0.2243

Framework versions

• Transformers 4.32.0.dev0
• Pytorch 2.0.1+cu118
• Datasets 2.14.4
• Tokenizers 0.13.3