AkshitSaxena1/whisper-tiny_to_american_accent_v2

Whisper tiny american

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

• Loss: 0.2197
• Wer: 9.1536

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: 2
• eval_batch_size: 1
• seed: 42
• optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 500
• training_steps: 3000

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.0285	0.2363	250	0.1990	7.9297
0.0431	0.4726	500	0.2119	9.2225
0.0469	0.7089	750	0.2184	7.8952
0.0589	0.9452	1000	0.2103	8.1538
0.0197	1.1815	1250	0.2149	8.6537
0.0146	1.4178	1500	0.2157	8.5675
0.0990	1.6541	1750	0.2132	8.6709
0.0278	1.8904	2000	0.2125	8.3262
0.0016	2.1267	2250	0.2183	8.8605
0.0054	2.3629	2500	0.2180	8.5158
0.0062	2.5992	2750	0.2183	9.0157
0.0032	2.8355	3000	0.2197	9.1536

Framework versions

• Transformers 5.5.0.dev0
• Pytorch 2.10.0+cu128
• Datasets 4.8.4
• Tokenizers 0.22.2