B0808/whisper-small-jotest

Whisper Small-FineTuning-allKeywords

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

• Loss: 1.5471
• Wer: 19.2263

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: 3
• 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: 100
• training_steps: 400
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.2091	50.0	50	1.6165	19.6366
0.0707	100.0	100	1.7163	19.2849
0.0	150.0	150	1.5499	19.4607
0.0	200.0	200	1.5489	19.4021
0.0	250.0	250	1.5481	19.2849
0.0	300.0	300	1.5477	19.2263
0.0	350.0	350	1.5471	19.2263
0.0	400.0	400	1.5471	19.2263

Framework versions

• Transformers 4.39.0.dev0
• Pytorch 2.1.0+cu121
• Datasets 2.18.0
• Tokenizers 0.15.2