README.md

metadata

language:
  - ko
license: apache-2.0
datasets:
  - haseong8012/child-50k
metrics:
  - wer
  - cer
base_model: openai/small

whisper-small_child50K_cosLR

This model is a fine-tuned version of openai/small on the child-50k dataset. It achieves the following results on the evaluation set:

• Loss: 0.0241
• Wer: 2.2556
• Cer: 0.9159

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: 1.25e-05
• train_batch_size: 16
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 2
• 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: 500
• training_steps: 5000

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
0.1058	0.36	500	0.0708	8.0241	3.2266
0.0577	0.71	1000	0.0446	5.2591	2.1891
0.0296	1.07	1500	0.0309	3.2662	1.3469
0.0301	1.42	2000	0.0279	2.7286	1.1516
0.0299	1.78	2500	0.0245	2.4578	0.9920
0.0219	2.13	3000	0.0262	2.6882	1.2290
0.0148	2.49	3500	0.0219	2.0899	0.9245
0.0141	2.84	4000	0.0281	2.8782	1.1615
0.0097	3.2	4500	0.0210	2.0616	0.8373
0.0109	3.55	5000	0.0241	2.2556	0.9159

Framework versions

• Transformers 4.34.0
• Pytorch 2.1.0+cu121
• Datasets 2.14.5
• Tokenizers 0.14.1