This model is a fine-tuned version of Sunbird/asr-whisper-large-v3-salt on an unknown dataset. It achieves the following results on the evaluation set:

Loss: 0.7247
Wer: 0.5862
Cer: 0.2164

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: 0.0001
train_batch_size: 8
eval_batch_size: 8
seed: 42
gradient_accumulation_steps: 4
total_train_batch_size: 32
optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
lr_scheduler_type: linear
lr_scheduler_warmup_steps: 100
training_steps: 2000

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
0.7385	2.7056	500	0.8633	0.6677	0.3449
0.3695	5.4071	1000	0.9962	0.6732	0.3274
0.1829	8.1085	1500	1.1718	0.6779	0.3267
0.0842	10.8141	2000	1.2547	0.6772	0.3272

Framework versions

PEFT 0.18.1
Transformers 4.52.0
Pytorch 2.10.0+cu128
Datasets 3.6.0
Tokenizers 0.21.4

KasuleTrevor/cdli-whisper-lg-l3-lr1e4_rpt

cdli-whisper-lg-l3-lr1e4_rpt

Model description

Intended uses & limitations

Training and evaluation data

Training procedure

Training hyperparameters

Training results

Framework versions

KasuleTrevor/
cdli-whisper-lg-l3-lr1e4_rpt