whisper-base-aug-22-mar-v1

This model is a fine-tuned version of openai/whisper-base on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.2257
• Wer: 64.4100

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: 16
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 32
• optimizer: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: constant
• lr_scheduler_warmup_steps: 1000
• num_epochs: 10

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
1.5377	1.0	123	1.3611	200.3506
1.2504	2.0	246	1.1472	113.7829
0.8795	3.0	369	0.6213	95.1450
0.4906	4.0	492	0.4055	84.1942
0.3405	5.0	615	0.3291	76.9117
0.2688	6.0	738	0.2860	72.3803
0.2195	7.0	861	0.2642	70.0067
0.1822	8.0	984	0.2465	67.0937
0.1535	9.0	1107	0.2324	65.1787
0.131	10.0	1230	0.2257	64.4100

Framework versions

• Transformers 4.50.0
• Pytorch 2.4.1+cu121
• Datasets 3.4.1
• Tokenizers 0.21.1