whisper-small-khmer-aug-v6

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

• Loss: 0.2209
• Wer: 60.8400

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
0.5473	0.9994	837	0.2368	79.1309
0.2005	2.0	1675	0.1907	69.7422
0.1505	2.9994	2512	0.1775	65.4289
0.1221	4.0	3350	0.1839	65.3802
0.1013	4.9994	4187	0.1888	64.1641
0.0851	6.0	5025	0.1921	62.8507
0.0725	6.9994	5862	0.1960	61.9588
0.0618	8.0	6700	0.2103	62.6074
0.053	8.9994	7537	0.2161	61.3426
0.0474	9.9940	8370	0.2209	60.8400

Framework versions

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