whisper-small-bn-3ds

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

• Loss: 0.0574
• Wer: 8.5038

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: 4
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 8
• total_train_batch_size: 32
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 8950
• training_steps: 28000

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.2337	0.22	2000	0.2298	31.7734
0.1444	0.45	4000	0.1516	22.2382
0.1106	0.67	6000	0.1181	17.2230
0.0933	0.89	8000	0.1005	14.8008
0.0747	1.12	10000	0.0865	12.7184
0.0642	1.34	12000	0.0788	11.6441
0.0613	1.56	14000	0.0715	10.6524
0.0565	1.79	16000	0.0666	9.8807
0.0494	2.01	18000	0.0640	9.5265
0.037	2.24	20000	0.0619	9.0317
0.0348	2.46	22000	0.0600	9.0329
0.0329	2.68	24000	0.0582	8.8407
0.0321	2.91	26000	0.0564	8.4931
0.0227	3.13	28000	0.0574	8.5038

Framework versions

• Transformers 4.28.0.dev0
• Pytorch 1.13.1+cu117
• Datasets 2.10.2.dev0
• Tokenizers 0.13.2