whisper-tiny-ckb

This model is a fine-tuned version of openai/whisper-tiny on the razhan/common_voice_ckb_16 dataset. It achieves the following results on the evaluation set:

• Loss: 0.2612
• Wer: 0.4780

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: 192
• eval_batch_size: 128
• seed: 42
• distributed_type: multi-GPU
• num_devices: 4
• total_train_batch_size: 768
• total_eval_batch_size: 512
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 200
• training_steps: 600
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.4502	0.72	100	0.4988	0.7166
0.2977	1.45	200	0.3557	0.5859
0.2494	2.17	300	0.3096	0.5315
0.2224	2.9	400	0.2817	0.5008
0.2148	3.62	500	0.2666	0.4819
0.2096	4.35	600	0.2612	0.4780

Framework versions

• Transformers 4.38.0.dev0
• Pytorch 2.0.1
• Datasets 2.16.1
• Tokenizers 0.15.0