whisper-tiny-khmer-aug-kcc-v2

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

• Loss: 0.6443
• Wer: 46.0250

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: 16
• eval_batch_size: 16
• 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: constant
• lr_scheduler_warmup_steps: 1000
• num_epochs: 10

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
1.8936	1.0	862	0.8037	61.5162
0.8143	2.0	1724	0.6129	55.2047
0.6306	3.0	2586	0.5592	46.5133
0.52	4.0	3448	0.5522	53.2805
0.4271	5.0	4310	0.5466	44.1696
0.3484	6.0	5172	0.5576	49.7975
0.2951	7.0	6034	0.5875	46.6725
0.2524	8.0	6896	0.6020	47.5767
0.2156	9.0	7758	0.6260	45.0557
0.1878	10.0	8620	0.6443	46.0250

Framework versions

• Transformers 4.44.0
• Pytorch 2.3.1
• Datasets 2.21.0
• Tokenizers 0.19.1