multi-whisper-playground

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

• Loss: 1.8508
• Wer: 195.4146
• Acc: 12.6021

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: 32
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.05
• num_epochs: 10
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Acc
2.2601	1.0	215	2.1670	210.5479	12.2520
1.8422	2.0	430	1.9717	187.2986	12.2520
1.6574	3.0	645	1.8943	162.3279	12.2520
1.4953	4.0	860	1.8562	187.3132	12.1354
1.4436	5.0	1075	1.8402	173.3153	12.2520
1.322	6.0	1290	1.8362	184.9692	12.4854
1.2386	7.0	1505	1.8372	183.5409	12.6021
1.1252	8.0	1720	1.8430	192.3747	12.6021
1.0449	9.0	1935	1.8459	198.8866	12.6021
1.0509	10.0	2150	1.8508	195.4146	12.6021

Framework versions

• Transformers 4.36.2
• Pytorch 2.0.0
• Datasets 2.16.1
• Tokenizers 0.15.0