whisper-tiny-en-finetune-minds14

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: 0.6541
• Wer Ortho: 0.3399
• Wer: 0.3383

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: 8
• eval_batch_size: 8
• seed: 42
• distributed_type: multi-GPU
• gradient_accumulation_steps: 2
• 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: 50
• training_steps: 1000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer Ortho	Wer
0.3136	3.57	100	0.4883	0.3640	0.3524
0.0417	7.14	200	0.5146	0.3560	0.3442
0.0066	10.71	300	0.5736	0.3411	0.3353
0.0017	14.29	400	0.6040	0.3455	0.3418
0.0013	17.86	500	0.6226	0.3393	0.3365
0.0009	21.43	600	0.6352	0.3393	0.3365
0.0007	25.0	700	0.6436	0.3399	0.3371
0.0006	28.57	800	0.6492	0.3399	0.3383
0.0006	32.14	900	0.6530	0.3399	0.3383
0.0006	35.71	1000	0.6541	0.3399	0.3383

Framework versions

• Transformers 4.29.2
• Pytorch 1.13.1+cu117
• Datasets 2.12.0
• Tokenizers 0.13.3