train04

This model is a fine-tuned version of openai/whisper-base on the ksponspeech_04 dataset. It achieves the following results on the evaluation set:

• Loss: 0.4226
• Cer: 14.6674

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: 128
• eval_batch_size: 64
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 500
• num_epochs: 1
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Cer
0.4513	0.13	100	0.4522	15.1448
0.4537	0.26	200	0.4484	14.7932
0.4419	0.39	300	0.4439	14.4669
0.4402	0.52	400	0.4399	14.8185
0.4397	0.65	500	0.4353	14.4394
0.4359	0.77	600	0.4282	14.0848
0.4245	0.9	700	0.4226	14.6674

Framework versions

• Transformers 4.36.0.dev0
• Pytorch 1.13.1+cu117
• Datasets 2.15.0
• Tokenizers 0.15.0