whisper-tiny-full-data-v1

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

• Loss: 0.5167

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: 0.0015
• train_batch_size: 32
• eval_batch_size: 8
• seed: 42
• distributed_type: tpu
• num_devices: 8
• total_train_batch_size: 256
• total_eval_batch_size: 64
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 1024
• training_steps: 63840

Training results

Training Loss	Epoch	Step	Validation Loss
1.5098	1.0	3192	1.8031
1.1616	2.0	6384	1.4578
1.0512	3.0	9576	1.3050
1.0219	4.0	12768	1.2848
0.9588	5.0	15960	1.2070
0.8933	6.0	19152	1.1487
0.8523	7.0	22344	1.0769
0.8058	8.0	25536	1.0350
0.7441	9.0	28728	0.9768
0.6941	10.0	31920	0.9111
0.6478	11.0	35112	0.8570
0.6015	12.0	38304	0.8058
0.5514	13.0	41496	0.7537
0.5175	14.0	44688	0.6989
0.4691	15.0	47880	0.6547
0.4343	16.0	51072	0.6162
0.4078	17.0	54264	0.5818
0.3788	18.0	57456	0.5503
0.3571	19.0	60648	0.5276
0.346	20.0	63840	0.5167

Framework versions

• Transformers 4.29.2
• Pytorch 2.1.0a0+gitcc01568
• Datasets 2.12.0
• Tokenizers 0.13.3