manuth/whisper-large-khmer

Whisper Large V3 Turbo Khmer

This model is a fine-tuned version of openai/whisper-large-v3-turbo on the Common Voice 11.0 dataset. It achieves the following results on the evaluation set:

• Loss: 0.6277
• Wer: 101.0956

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: 8
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 32
• optimizer: Use adamw_hf 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: 500
• training_steps: 4000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.3899	15.8835	1000	0.6277	101.0956
0.0522	31.7550	2000	0.7151	102.6394
0.0051	47.6265	3000	1.0525	104.0090
0.0023	63.4980	4000	1.1451	104.5319

Framework versions

• Transformers 4.49.0
• Pytorch 2.6.0+cu124
• Datasets 3.3.1
• Tokenizers 0.21.0