whisper-small-r22-e

This model is a fine-tuned version of openai/whisper-small on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.2918
• Wer: 21.3875

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: 5
• training_steps: 150
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.3822	0.09	10	0.4255	23.2826
0.2636	0.18	20	0.3321	22.4196
0.2037	0.27	30	0.3279	23.8071
0.1943	0.36	40	0.3177	22.3858
0.2203	0.45	50	0.3109	22.4873
0.193	0.54	60	0.3071	22.9272
0.2096	0.63	70	0.2990	22.6565
0.214	0.72	80	0.3029	22.4873
0.2375	0.81	90	0.2927	21.7259
0.2238	0.9	100	0.2918	22.4196
0.2119	0.99	110	0.2919	22.7580
0.1362	1.08	120	0.2897	22.0135
0.0997	1.17	130	0.2915	21.3029
0.0824	1.26	140	0.2920	21.4382
0.0923	1.35	150	0.2918	21.3875

Framework versions

• Transformers 4.35.0.dev0
• Pytorch 2.1.0+cu118
• Datasets 2.14.7.dev0
• Tokenizers 0.14.1