whisper-tiny-khmer-v9-aug-kcc

This model is a fine-tuned version of openai/whisper-tiny on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.5628
• Wer: 42.0574

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: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: constant
• lr_scheduler_warmup_steps: 1000
• num_epochs: 10

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
2.2112	1.0	973	0.8366	88.7844
0.9868	2.0	1946	0.6426	91.4779
0.7825	3.0	2919	0.5717	55.7305
0.6679	4.0	3892	0.5407	45.5946
0.5866	5.0	4865	0.5342	49.5678
0.5223	6.0	5838	0.5202	44.1110
0.4692	7.0	6811	0.5172	44.5772
0.4225	8.0	7784	0.5337	43.5561
0.3808	9.0	8757	0.5476	43.0087
0.3433	9.9902	9720	0.5628	42.0574

Framework versions

• Transformers 4.47.0
• Pytorch 2.3.1
• Datasets 3.2.0
• Tokenizers 0.21.0