arielcerdap/whisper-small-fluencybank

Whisper fine-tuned on FluencyBank — openai/whisper-small

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

• Loss: 1.9714
• Wer: 14.7220
• Cer: 8.4574

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: 8e-06
• train_batch_size: 32
• eval_batch_size: 32
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 128
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: cosine
• lr_scheduler_warmup_ratio: 0.1
• training_steps: 2500
• label_smoothing_factor: 0.1

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
1.5202	11.6279	250	1.7446	14.2167	10.7879
1.4329	23.2558	500	1.8143	13.4476	7.7609
1.4254	34.8837	750	1.8473	13.2938	7.7063
1.4211	46.5116	1000	1.9020	13.8651	7.8338
1.4201	58.1395	1250	1.9092	13.8211	7.8611
1.4189	69.7674	1500	1.9383	14.1727	8.1615
1.418	81.3953	1750	1.9574	14.3265	8.2161
1.4177	93.0233	2000	1.9669	14.5023	8.3572
1.4176	104.6512	2250	1.9709	14.5902	8.3663
1.4175	116.2791	2500	1.9714	14.7220	8.4574

Framework versions

• Transformers 4.45.2
• Pytorch 2.10.0+cu128
• Datasets 4.0.0
• Tokenizers 0.20.3